Sustainable Management of Cultivated Organic Soils in Switzerland – an Economic and Policy Analysis

Research Collection

Doctoral Thesis

Sustainable Management of Cultivated Organic Soils in Switzerland – An Economic and Policy Analysis

Author(s): Ferre, Marie

Publication Date: 2017

Permanent Link: https://doi.org/10.3929/ethz-b-000213859

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use.

ETH Library DISS. ETH NO. 24408

SUSTAINABLE MANAGEMENT OF CULTIVATED ORGANIC SOILS IN SWITZERLAND – AN ECONOMIC AND POLICY ANALYSIS

A thesis submitted to attain the degree of DOCTOR OF SCIENCES of ETH ZURICH (Dr. sc. ETH Zurich)

presented by

MARIE FERRE

International Master of Science in Rural Development, Ghent University; Master of science with a major in Food and Resource Economics, University of Florida; diplôme d’ingénieur, specialité: ingénieur agronome, Agrocampus Ouest-Rennes

born on 10.06.1989

citizen of France

accepted on the recommendation of

Prof. Dr. Isabel Günther, examiner Prof. Dr. Stefanie Engel, co-examiner Dr. Jetzke Bouma, co-examiner Dr. Adrian Muller, co-examiner Dr. Elisabeth Gsottbauer, co-examiner

2017 1

Acknowledgements

I would first like to thank my supervisors, Stefanie Engel and Adrian Muller for giving me the opportunity to work on this project. I am glad I worked on such relevant and challenging issue: peatlands. The guidance, support, and trust that I received from Stefanie and Adrian all along the PhD project have contributed to its accomplishment. I thank them for making me explore new fields of studies: behavioral economics and efficiency analysis in particular, and for their constant encouragements throughout this journey. I am also very grateful to Elisabeth Gsottbauer who joined the project, and supported my work. I want to thank Isabel Günther for accepting to supervise my thesis, and Jetske Bouma for being in my thesis committee. Finally, I highly appreciated working within this interdisciplinary team, together with the group of soil scientists composed of Jens Leifeld – leader of the project, Moritz Müller, and Cédric Bader. Our meetings were very pleasant and fruitful! I am deeply grateful to Orencio Robaina who made the experiments possible by creating virtual and interactive vegetable farming systems on organic soils. It was such a great collaboration; thanks to Adrienne Grêt Regamey for allowing this. I acknowledge your patience Orencio despite the numerous programming challenges. My stay at ETH Zurich would not have been better without my past colleagues at PEPE: Alex, Alexandra, Christian, Loretta, Marleen, Marlene, Pan, Tim, and Willi, with a special thought to Saraly. During my second and third years, I had the opportunity to get to know and interact with other amazing research groups including the NARP department; in particular Eva, Kathrin, and Jerelee; FORDEV department; I appreciate Claude’s feedback on the experiment and for letting me discover role-playing games, and FIBL; I appreciate Adrian for making it possible for me to work from there, it was a pleasure every time I worked from there! Special thoughts to Bernadette, Matthias, Stefanie, Rebekka, Robert, Bennan, Matthias, Christian, Johan, Simon, Sylvain, Lidvia, Adele, and Emilia! Finally, the Agricultural economics and Policy group: I give thanks to the whole group, including Robert, Röbi, Katarina, Hang, Manuela, Solen, Stefan, Martina, Tobias, Niklas, Ladina, Poorvi, and Sergei for having me during coffee breaks and for the very nice talks. I gratefully acknowledge funding support from the Swiss National Science Foundation and from the Humboldt foundation. I want to thank the directors and staff of the agricultural schools who allowed me and provided me with suitable conditions to conduct experiments with their students (farm apprentices), especially Alexandre Horner, Daniel Bouquet, Cyril Perrenoud, Melinda Benoit, Philippe Girod, Raphaël Gaillard, Pierre-André Odiet, Claude Gerwig, and Moritz Müller. I am also grateful to the experts who facilitated my understanding of the different aspects of the research problem: Stefan Mann, Moritz Müller, Albert Lüscher, Jean-François Jaton, Peter Trachsel, Nathalie Grob, Martin Lichtenhan, Jacques-Yves Deriaz, Willy Stoll, and Edwin Egger. Finally, I want to thank the Descil lab at ETH - Oliver Brägger and Stephan Wehrli especially - for facilitating the organization of the experiments with students. Thanks to Morgane, Xioaya, Alice, Theresa, and Marlene for making my life in Zurich nicer. Despite the distance, yet some people were always there for me. Vero, Yann, Juliette, Bruno, Pimp, VRP, Olivier, Isabelle, CAT, Vincent, DDT, Claire, Mathilde, Mari, Marie T., Severin, Kayo, Marie Q, Matthieu, Mathieu, Audrey thank you all for your friendship, your support and care. I cherish every time we spend together! My deepest gratitude goes to my family for their love and support. I want to thank my husband Peter, my father Marcel, my mother Suzanne, and my siblings Anna, Yvan, and David for being there for me. Peter was my strength despite the distance; I appreciate his understanding, support, and love.

Contents

ABSTRACT ...... 8

GENERAL INTRODUCTION ...... 15 1. Peatlands and climate change ...... 15 1.1. An overview ...... 15 1.2. Climate change mitigation by peatlands ...... 15 2. Conservation measures for organic soils ...... 16 2.1. Organic soils and current policy approaches ...... 16 2.2. Policy design ...... 17 2.3. Agri-environmental payments ...... 18 3. Overview of the thesis ...... 20 4. Research questions and methods ...... 21

CHAPTER 1: Sustainable management of cultivated peatlands in Switzerland: insights, challenges, and opportunities ...... 26 1. Introduction ...... 26 2. Study region and methods ...... 27 2.1. Study area ...... 27 2.2. Methods ...... 29 3. Vegetable production on organic soils ...... 29 3.1. The economics of vegetable production in Switzerland ...... 29 3.2. Requirements for intensive agricultural management on organic soils ...... 30 3.3. Challenges of intensive cultivation on organic soils ...... 30 4. Alternative management practices on organic soils and challenges to their implementation...... 32 4.1. Options for sustainable management of organic soils ...... 32 4.2. Challenges of alternative management practices on organic soils ...... 33 5. Policy instruments to support alternative management practices ...... 35 5.1. Institutional context in Switzerland ...... 35 5.2. Policies to promote sustainable use of organic soils in Switzerland ...... 36 6. Scenarios for the long-term development of organic soils ...... 39 6.1. Scenario 1: Continuing vegetable farming on organic soils ...... 39 6.2. Scenario 2: Rewetting organic soils ...... 40 6.3. Impacts of scenarios 1 and 2 ...... 40 7. Conclusion ...... 41

CHAPTER 2: Incentives for sustainable land use considering cost heterogeneity among farmers: Results from a computerized framed experiment ...... 44 1. Introduction ...... 44

2. Methods ...... 46 2.1. Experimental set up ...... 46 2.2. Behavioral predictions ...... 48 3. Results ...... 51 3.1. Policy performance ...... 51 3.2. Effect of players’ social preferences on behavior ...... 55 4. Conclusions ...... 58

CHAPTER 3: Can agglomeration payments induce sustainable management of organic soils in Switzerland? – A computerized framed experiment ...... 61 1. Introduction ...... 61 2. Experimental design and expected effects ...... 63 3. Results ...... 66 3.1. Environmental effectiveness ...... 67 3.2. Income inequality and cost effectiveness ...... 68 3.3. Effect of side payments on sustainable management ...... 69 3.4. Effect of players’ socio-demographic and preference characteristics on their behavior ... 71 4. Discussion and conclusion ...... 76

CHAPTER 4: External validity of experiments in environmental economics: framing and subject pool effects among students and professionals ...... 79 1. Introduction ...... 79 2. Literature review ...... 80 2.1. Subject-pool effects ...... 80 2.2. Framing effects ...... 81 3. Methodology ...... 81 3.1. Experimental design ...... 81 3.2. Framed and unframed experiment ...... 83 3.3. Subject pools and organization ...... 83 4. Results ...... 84 4.1. Impacts of subject pool ...... 84 4.2. Impacts of framing ...... 87 4.3. Effect of individual players’ characteristics across framings and subject pools ...... 89 5. Discussion and conclusion ...... 92

GENERAL DISCUSSION AND CONCLUSIONS ...... 95 1. Contributions to a more sustainable management of organic soils ...... 95 1.1. Expanding knowledge on a complex socio-ecological issue ...... 95 1.2. Agri-environmental payments for a sustainable management of organic soils ...... 95 1.3. Sustainable development of organic soils ...... 97

2. Contributions to behavioral and experimental economics ...... 98 2.1. An innovative tool of analysis ...... 98 2.2. Understanding cooperation among players ...... 98 2.3. Importance of the choice of subject pool in experimental economics ...... 99 2.4. The role of social preferences and other socio-demographic characteristics ...... 100 3. Limitations of the thesis and future research needs ...... 101 3.1. Exploration of the potential of other types of policy instruments ...... 101 3.2. Assessing the possibility of a sustainable use of organic soils among other actors ...... 102 3.3. Conducting the experiment with a higher number of players per group ...... 102 3.4. Non-binding side payments in the experimental setup ...... 103 3.5. Further investigation of framing and subject-pool effects ...... 103 3.6. Further analysis of behavioral factors ...... 103 3.7. Transferability of the experimental results ...... 104

REFERENCES ...... 109

APPENDICES ...... 123 1. Appendix A – Appendix to chapter 1 ...... 123 2. Appendix B – Appendix to chapter 2 ...... 124 B1. Screen shot of the experiment ...... 124 B2. Experimental instructions – Baseline ...... 124 B3. Players’ payoffs if organic soils are rewetted, for each payment design tested ...... 127 B4. Derivation of behavioral predictions ...... 128 B5. Representation of social preferences-types ...... 130 B6. Representation of players’ characteristics across treatments ...... 130 3. Appendix C – Appendix to chapter 3 ...... 131 C1. SVO test description ...... 131 C2. Experimental design ...... 131 C3. Instructions – Baseline scenario ...... 134 C4. Percentage of H and L players who adopt activity A ...... 139 C5. Social preference types ...... 139 C6. Representation of players’ characteristics across scenarios ...... 140 4. Appendix D – Appendix to chapter 4 ...... 141 D1. Background visualization ...... 141 D2. Side payments ...... 142 D3. Framing and treatment cost effectiveness ...... 142 D4. Framing and income inequalities ...... 142 D5. Characteristics of players ...... 143

List of Tables

Table 1: Main challenges to a possible change in management practice on organic soils in Seeland .. 35 Table 2: Baseline behavioral predictions ...... 49 Table 3: Expected results for payoff-maximizing players...... 51 Table 4: Cost-effectiveness across treatments ...... 53 Table 5: Summary statistics of players’ characteristics across treatments ...... 56 Table 6: Panel random-effect logistic regression on players’ land use choice ...... 57 Table 7: Farm profits of activity B ...... 65

Table 8: Level of the payment in APV ...... 66 Table 9: Logistic random-effect panel regression on players’ adoption of activity A...... 68 Table 10: Comparison of the cost effectiveness of payment schemes ...... 69 Table 11: Logistic random-effect panel regression of players’ vote for rewetting organic soils (first stage) ...... 72 Table 12: Logistic random-effect panel regression of players’ vote for adopting activity A in stage 2- decision ...... 75 Table 13: Experimental set up ...... 84 Table 14: Random effect logistic panel regression on player’s cooperation ...... 86 Table 15: Impact of the subject pool on cost-effectiveness (framed design) ...... 87 Table 16: Test of subject pool effect on inequalities (framed design) ...... 87 Table 17: Random effect logistic panel regression on player’s cooperation among students ...... 88 Table 18: Panel regression on player’s adoption of activity 2 including player characteristics and their interaction with subject-pool effects ...... 90 Table 19: Panel regression on player’s cooperation including players’ characteristics and their interactions with framing effect...... 92

List of Figures

Figure 1: Soil profile of organic soils ...... 18 Figure 2: Distribution of organic soils in Switzerland ...... 28 Figure 3: The Seeland region and the two main areas of vegetable production ...... 28 Figure 4: Pore volume and soil bulk density of degrading organic soils in Seeland ...... 31 Figure 5: Stage 2 payoffs if organic soils are not rewetted (‘Status quo’) ...... 48 Figure 6: Stage 2 payoffs if organic soils are rewetted ...... 48 Figure 7: Experimental decision stages ...... 48 Figure 8: Percentage of players adopting activity A across treatments ...... 51 Figure 9: Percentage of groups adopting activity A across treatments ...... 52 Figure 10: Mean net offers and mean net transferred side payments ...... 54 Figure 11: Profit functions of H and L players over time ...... 64 Figure 12: Players’ payoffs at time t if organic soils are not rewetted...... 65 Figure 13: Players’ payoffs at time t if organic soils are rewetted...... 65 Figure 14: Percentage of players who adopt activity A across treatments ...... 67 Figure 15: Average net side-payment offers among group members across treatments ...... 70 Figure 16: Players’ payoffs if soils are not rewetted...... 82 Figure 17: Players’ payoffs if soils are rewetted, including side payments...... 82 Figure 18: Organization of the experiment ...... 83 Figure 19: Adoption of activity 2 across subject pools in baseline and UA (framed design, dynamic setting) ...... 85 Figure 20: Proportion of players who cooperate in the baseline and in treatment UA across framings, in the dynamic setting (Figure 20a) and in the static setting (Figure 20b) ...... 88 Figure 21: Distribution of the SVO angle among students (456) and apprentices (146) ...... 89

ABSTRACT

Cultivated organic soils or peatlands are an important example of emission hotspots. Intensified agricultural activities requiring drainage of peatland lead to the degradation and ultimately in many regions already to the disappearance of these soils. These drained peatlands in turn emit substantial shares of greenhouse gas. Therefore, their conservation in the form of actively rewetting drained peatlands are an important pillar of climate change mitigation policies. This thesis addresses peatland conservation in Switzerland from an economic and policy perspective. The study region for this thesis is the “Seeland” region located in the canton of Bern and characterized by intensive vegetable farming on organic soils. Chapter 1 contributes to a better understanding of the inherent complexities of the management and conservation of cultivated organic soils and provides an overview of the institutional context of peatland utilization and preservation in Switzerland. This chapter draws upon qualitative expert interviews and a literature review. Chapters 2 and 3 focus on the design and evaluation of economic incentives to promote sustainable management of organic soils. Economic experiments conducted with students and farm apprentices are used to test and compare different policies for peatland conservation. Chapter 4 contributes to the academic debate on the role of framing and use of different subject populations with respect to the external validity of experiments. Chapter 1 provides an overview of the current farming situation in the study region and the historical development of the management of organic soils with respect to the institutional and environmental context. It compares the environmental and economic consequences of two land-use development scenarios, namely: pursuing current management practices and adopting sustainable management practices enabling the preservation of the peat. By doing so, chapter 1 contributes to an understanding of the complexity of the issue and provides concrete insights for the Swiss context. It identifies the main challenges that need to be tackled in order to promote a change in current management practices to a more sustainable use of these soils. These challenges include, amongst others, the high profitability of the soils in vegetable farming and the lack of systematic data on the properties and exact distribution of soils in the region. In conclusion, chapter 1 argues for the need of a long-term and comprehensive vision of the management problem supported by a clear policy framework that is able to facilitate a change in the management of organic soils. Chapters 2 and 3 address the policy question by capturing the management issue of organic soils in an innovative economic experiment that is computerized, framed, and interactive. The experiment simulates the decision situation of vegetable producers on organic soils and particularly accounts for the fact that (i) regulation of the water table on these soils requires cooperation among farmers, and (ii) farmers within the same peatland area differ substantially in the soil production potential, and thus in their opportunity costs of adopting sustainable practices. The experiment investigates the potential behavior of farmers when facing the trade-off between continuing vegetable farming versus adopting sustainable use of organic soil that necessitates restricting soil drainage. It combines several complexities that have not been addressed in the broader literature on the design of PES: (i) a two-stage decision problem consisting of a collective choice and an individual land use decision; (ii) the heterogeneity among farmers in opportunity costs for adopting sustainable land use; and (iii) the option for farmers to make side payments within their group. Two basic scenarios were evaluated: without policy intervention and with an agri-environmental payment scheme allocated in exchange for adopting sustainable use of the soils. These scenarios were tested with a static and a dynamic experiment. The static experiment (chapter 2), conducted with university students, represents the resource management problem in general terms. The dynamic experiment (chapter 3), conducted with farm apprentices, more

8 closely represents the management situation by accounting for dynamic changes in soil productivity and opportunity costs. Both chapters evaluate the effectiveness of three payment designs: a uniform agglomeration payment, a differentiated agglomeration payment, and a uniform individual payment. In the static experiment, the uniform payment corresponds to an average of the conservation costs of participating farmers, whereas in the dynamic experiment, it is based on the initial opportunity costs of players. The differentiated agglomeration payment is aligned to players’ opportunity costs. While the difference between the players’ opportunity costs is fixed in the static setting, opportunity costs of players may vary in the dynamic setting. The thesis evaluates the performance of the alternative agri- environmental payment schemes along three main criteria: environmental effectiveness, cost effectiveness, and distributional effects. Using individual surveys, the thesis also examines the links between players’ behavior in the experiment and their social preferences and other individual socio- economic characteristics. Chapters 2 and 3 show that players are more willing to implement sustainable use of organic soils when a payment scheme is present than when this is not the case. Chapter 2 (static experiments) assumes a constrained budget of the governmental agency allocating the subsidies. A uniform payment scheme that is set as average between the costs of players only provides an incentive to players with low costs to cooperate. Despite this and due to payment redistribution via side payments, the environmental effectiveness of uniform payments does not differ from that of the differentiated payment. However, the differentiated payment appears to be more cost effective. Thus, in the presence of asymmetry between farmers and a limited budget, differentiated payments may be more promising in promoting sustainable practices than uniform payments. Chapter 2 also shows that the uniform agglomeration payment induces a higher rate of peat preserved and is more cost-effective than the uniform individual payment. The experiment in Chapter 3 which reflects the temporal dynamics of peat degradation shows that a constant agglomeration payment scheme which is aligned to the players’ initial opportunity costs is more cost- and environmentally-effective than a variable payment scheme that mirrors the respective opportunity costs of farmers over time. This result can be explained by the fact that in the constant payment treatment, players are incentivized to cooperate early on and are then also able to maintain high cooperation levels over the course of the game. The constant payment also results in lower inequality with respect to players payoffs than the variable payment. Therefore, with respect to the promotion of sustainable use of organic soils in Switzerland, constant agglomeration payments appear to be the most promising option. It is acknowledged that in both experiments a significant share of players do not opt for sustainable usage of organic soils even though they have an economic incentive to do so. Possible explanations for this include the risk associated with each of the agglomeration payment types, the considerable opportunity costs characterizing the management problem, and players’ heterogeneity. Furthermore, the social preferences of players as well as other personal characteristics such as risk aversion, opinion of cooperative approaches, farming identity, and care for personal reputation, are identified as important factors that influence the decision and behavior of players. Finally, Chapter 4 explores the robustness of the results obtained in the previous chapters by comparing the behavior of farm apprentices to the behavior of university students and by using a framed and unframed version of the very same experiment. The framed experiment corresponds here to the introduction of a particular context issue related in this case to farming and to the management of organic soils in particular. Thus, it contributes to the discussion on the external validity of results generated by abstract and unframed experiments generally conducted with university students. Chapter 4 shows that the type of subject significantly affects experimental outcomes, particularly in the more and realistic dynamic version of the experiment. This study shows that cooperation is on average significantly higher among university students than among farm apprentices. Nevertheless, subject pool effects do not appear 9 to affect treatment differences as the ranking of the policy instrument with respect to their environmental and cost effectiveness as well as distributional outcomes is similar notwithstanding who participates. Furthermore, comparing behavior of university students in the framed experiment to their behavior in an unframed (abstract) experiment reveals no important differences along the tested criteria.

KURZFASSUNG

Landwirtschaftlich genutzte organische Böden bzw. Torfmoore sind wichtige Beispiele für Emissions- Hotspots. Verstärkte landwirtschaftliche Aktivitäten, die die Entwässerung von Torfmooren erfordern, führen zu einer Verschlechterung und letztlich in vielen Regionen bereits zum Verschwinden dieser Böden. Diese trockengelegten Torfmoore wiederum setzen einen erheblichen Anteil an Treibhausgasen frei. Daher kann deren Erhaltung über eine aktive Wiedervernässung einen wichtigen Beitrag zur Minderung des Klimawandels leisten. Diese Arbeit befasst sich mit der Erhaltung von Torfmooren in der Schweiz aus ökonomischer und politischer Sicht. Das Untersuchungsgebiet für diese Arbeit ist die im Kanton Bern gelegene, von intensiven Gemüseanbau auf organischen Böden geprägte Region „Seeland“. Kapitel 1 trägt zum besseren Verständnis der komplexen Abläufe bei der Bewirtschaftung und Erhaltung von landwirtschaftlich genutzten organischen Böden sowie der derzeitigen Anbaumethoden bei und vermittelt einen Überblick über den institutionellen Kontext der Nutzung und Erhaltung von Torfmooren in der Schweiz. Dieses Kapitel stützt sich auf qualitative Experteninterviews und einer Literaturstudie. Die Kapitel 2 und 3 befassen sich mit der Konzeption und Bewertung von wirtschaftlichen Anreizen zur Förderung einer nachhaltigen Bewirtschaftung von organischen Böden. Mit Studenten und Auszubildenden in landwirtschaftlichen Betrieben durchgeführte ökonomische Experimente dienen der Untersuchung und dem Vergleich von verschiedenen Strategien zur Erhaltung von Torfmooren. Kapitel 4 leistet einen Beitrag zur akademischen Debatte über die Rolle Darstellungseffekte („Framing-Effekte“) und Stichprobenpopulationen in ökonomischen Experimenten und deren Einfluss auf die Validität der Experimente. Kapitel 1 liefert einen Überblick über die aktuelle Situation in der Landwirtschaft im Untersuchungsgebiet und die historische Entwicklung der Bewirtschaftung von organischen Böden in Bezug auf den institutionellen und ökologischen Kontext. Darin werden die ökologischen und ökonomischen Konsequenzen (in Form der Kosten und Nutzen) zweier Szenarien der Entwicklung der Landnutzung auf diesen Böden verglichen: a) Beibehalten der derzeitigen Bewirtschaftungsmethoden und b) Anwendung nachhaltiger Bewirtschaftungsmethoden, die die Erhaltung der Torfmoore ermöglichen. Damit trägt Kapitel 1 zum Verständnis der Komplexität des Themas bei und liefert konkrete Einblicke für den Schweizer Kontext. Es werden darin die wichtigsten Herausforderungen ermittelt, die bewältigt werden müssen, um Veränderungen bei den derzeitigen Bewirtschaftungsmethoden hin zu einer nachhaltigeren Nutzung dieser Böden zu fördern. Diese Herausforderungen beinhalten u. a. die hohe Wirtschaftlichkeit der Böden beim Gemüseanbau und das Fehlen von systematischen Daten über die Eigenschaften und genaue Verteilung der Torfböden in der Region. Abschliessend wird in Kapitel 1 die Notwendigkeit eines langfristigen, umfassenden Konzepts zur Nutzung oder dem Schutz der verbleibenden Torfböden dargelegt, welches von einem klaren politischen Rahmen unterstützt werden muss, der Veränderungen bei der Bewirtschaftung von organischen Böden erleichtern kann. In Bezug auf die Notwendigkeit von Massnahmen, die speziell auf diese organischen Böden ausgerichtet sind, werden in dieser Arbeit Agrarumweltzahlungssysteme zur Förderung alternativer Methoden auf organischen Böden untersucht. Die Kapitel 2 und 3 befassen sich mit dieser Massnahmenproblematik mit einem innovativen ökonomischen Experiment zur Bewirtschaftung von organischen Böden, welches computergestützt und interaktiv ist. In diesem Experiment wird die Entscheidungssituation der Gemüseproduzenten auf organischen Böden simuliert. Es umfasst zwei Schlüsselaspekte: (i) Die Regulierung des Grundwasserstandes auf diesen Böden erfordert die Zusammenarbeit von Landwirten, und (ii) die Bodenprofile variieren sehr stark innerhalb der einzelnen Parzellen. Das Produktionspotenzial der Böden von Landwirten im gleichen Torfgebiet unterscheidet sich erheblich 11 und damit auch ihre Opportunitätskosten für die Anwendung nachhaltiger Verfahren. Im Experiment wird das mögliche Verhalten der Landwirte untersucht, wenn diese zwischen der Fortführung des Gemüseanbaus und der Einführung der nachhaltigen Nutzung von organischen Böden, die eine Einschränkung der Bodenentwässerung erfordert, abwägen müssen. Dabei werden verschiedene komplexe Aspekte kombiniert, die in der allgemeinen Literatur zur Gestaltung von Agrarumweltzhalungen nicht behandelt werden: i) ein zweistufiges Entscheidungsproblem bestehend aus einer kollektiven Entscheidung und einer individuellen Landnutzungsentscheidung, ii) die Heterogenität der Landwirte bei den Opportunitätskosten für die Einführung nachhaltiger Landnutzung, und iii) die Möglichkeit für Landwirte, Kompensationszahlungen innerhalb ihrer Gruppe zu leisten. Es wurden zwei Basisszenarien untersucht: ohne politische Eingriffe und mit einem Agrarumweltzahlungssystem, das im Gegenzug für die Einführung der nachhaltigen Nutzung von Böden umgesetzt wird. Diese Szenarien wurden in einem statischen und einem dynamischen Experiment untersucht. Das statische Experiment (Kapitel 2), das mit Universitätsstudenten durchgeführt wurde, stellt das Problem der Ressourcenbewirtschaftung dar und steht im Einklang mit der Literatur zur experimentellen Ökonomik. Das dynamische Experiment (Kapitel 3), das mit Auszubildenden in landwirtschaftlichen Betrieben durchgeführt wurde, stellt noch genauer die Bewirtschaftungssituation durch Aufnahme von Veränderungen bei der Bodenproduktivität im Laufe der Zeit sowie eine dynamische Entwicklung der Opportunitätskosten dar. In dieser Arbeit wird das Potenzial von drei Zahlungsmodellen untersucht: eine einheitliche Agglomerationszahlung, eine differenzierte Agglomerationszahlung und eine einheitliche Einzelzahlung. Im statischen Experiment entspricht die einheitliche Zahlung den durchschnittlichen Erhaltungskosten der teilnehmenden Landwirte, wohingegen im dynamischen Experiment die einheitliche Zahlung auf den anfänglichen Opportunitätskosten der Spieler basiert und somit konstant ist. Die differenzierte Agglomerationszahlung ist auf die Opportunitätskosten der Spieler abgestimmt. Während der Unterschied zwischen den Opportunitätskosten der Spieler im statischen Experiment fix ist, können die Opportunitätskosten im dynamischen Experiment variieren und die Zahlung der Spieler mit kurzfristigem Produktionspotenzial verringert sich. In dieser Arbeit werden die alternativen Agrarumweltzahlungssysteme anhand von drei Hauptkriterien beurteilt: ökologische Effektivität, Kosteneffizienz und Verteilungsgerechtigkeit. Um ein besseres Verständnis für die Entscheidungen der Spieler zu erhalten, werden in dieser Arbeit unter Verwendung von Einzelbefragungen der Teilnehmenden die Zusammenhänge zwischen den Ergebnissen der Experimente und den sozialen Präferenzen der einzelnen Personen sowie sonstigen persönlichen sozioökonomischen Merkmalen untersucht. Die Experimente ermöglichten es, wertvolle Erkenntnisse über die mögliche Entscheidung der Landwirte in Bezug auf die Bewirtschaftungsmethoden für organische Böden und ihre Antworten auf die verschiedenen Zahlungssysteme zu sammeln. In den Kapiteln 2 und 3 wird dargelegt, dass die Spieler eher bereit sind, eine nachhaltige Nutzung von organischen Böden umzusetzen, wenn ein Zahlungssystem vorhanden ist als wenn dies nicht der Fall ist. In Kapitel 2 (statische Experimente) wird von einem begrenzten Budget für die Regierungsbehörde, die für die Vergabe von Zuschüssen zuständig ist, ausgegangen. Ein einheitliches Zahlungssystem, das auf den durchschnittlichen Kosten der Spieler basiert, bietet nur Anreize zur Zusammenarbeit für Spieler mit niedrigen Kosten. Trotzdem und aufgrund der Umverteilung der Zahlungen mittels Kompensationszahlungen unterscheidet sich die ökologische Effektivität der einheitlichen Zahlung nicht von der ökologische Effektivität der differenzierten Zahlung. Allerdings erscheint die differenzierte Zahlung kostengünstiger zu sein. Daher können differenzierte Zahlungen vor dem Hintergrund der Asymmetrie zwischen den Landwirten und einem begrenzten Budget vielversprechender bei der Förderungen nachhaltiger Methoden sein als einheitliche Zahlungen auf Basis der durchschnittlichen Erhaltungskosten der teilnehmenden Landwirte. In Kapitel 2 wird zudem dargelegt, dass die einheitliche Agglomerationszahlung dazu führt, dass mehr Torfmoore erhalten werden und dieses System kostengünstiger ist als die einheitliche Einzelzahlung. Sollten die differenzierten Zahlungssysteme auf Widerstand gegen die Umsetzung stossen und die 12 politischen Entscheidungsträger ein einheitliches Zahlungssystem favorisieren, kann eine Agglomerationszahlung effektiver sein. In Kapitel 3 wird die Dynamik der Bodendegradierung beschrieben und dargestellt, dass ein konstantes, auf die ursprünglichen Opportunitätskosten der Spieler abgestimmtes Agglomerationszahlungssystem kostengünstiger und ökologisch effektiver zu sein scheint als ein variables Agglomerationszahlungssystem, das die jeweiligen Opportunitätskosten der Landwirte im Laufe der Zeit widerspiegelt. Dieses Ergebnis erklärt sich hauptsächlich dadurch, dass bei der konstanten Zahlungsmethode die Landwirte frühzeitig zusammenarbeiten und die Koordinierung bei der Einführung nachhaltiger Landnutzung fortsetzen. Das konstante Zahlungssystem führt zudem zu einer geringeren Ungleichheit beim Einkommen als das variable Zahlungssystem. Bei der Förderung der nachhaltigen Nutzung von organischen Böden in der Schweiz scheinen konstante Agglomerationszahlungen eine vielversprechende Option zu sein. Allerdings hat sich in den Experimenten ein erheblicher Anteil der Spieler nicht für die nachhaltige Nutzung von organischen Böden entschieden, obwohl sie einen wirtschaftlichen Anreiz dafür hatten. Dies ist eine weitere wichtige Erkenntnis dieses Kapitels. Eine mögliche Erklärung hierfür wären das mit jedem der Agglomerationszahlungsarten verbundene Risiko, die erheblichen Opportunitätskosten, die charakterisierend für dieses Problem sind, und die Asymmetrie zwischen den Landwirten, die einige Verhandlungen bei den Kompensationszahlungen zwischen den Landwirten der gleichen Gruppe hervorruft. In diesem einzigartigen Umfeld scheint es, dass starke Anreize erforderlich sind, um Veränderungen bei den Bewirtschaftungsmethoden für organische Böden zu bewirken. Zudem wurden die sozialen Präferenzen der Akteure sowie sonstige persönliche Merkmale wie z. B. Risikoaversion, Einstellung zu kooperativen Ansätzen, die landwirtschaftliche Identität und die Sorge um den persönlichen Ruf als Schlüsselfaktoren ermittelt, die die Entscheidung und das Verhalten der Spieler beeinflussen. Abschliessend wird in Kapitel 4 die Robustheit der in den vorhergehenden Kapiteln erhaltenen Ergebnisse erörtert, indem das Verhalten der Auszubildenden in landwirtschaftlichen Betrieben mit dem Verhalten der Universitätsstudenten verglichen wird und sowohl eine kontextreiche als auch eine kontextlose Variante desselben Experiments angewendet wird. Somit trägt es zur Debatte über die externe Validität der durch die abstrakten, in der Regel mit Universitätsstudenten durchgeführten Experimente erzielten Ergebnisse in vielen solchen Untersuchungen bei. Weiters wird in Kapitel 4 dargelegt, dass die Art der Versuchspersonen das Verhalten erheblich beeinflusst, vor allem bei der komplexeren sowie realitätsnahen dynamischen Variante des Experiments. Die nachhaltige Nutzung organischer Böden ist im Durchschnitt deutlich höher bei den Universitätsstudenten als bei den Auszubildenden in landwirtschaftlichen Betrieben. Dennoch scheinen die gruppenspezifischen Unterschiede („subject pool effects“) die relative Bewertung der politischen Massnahmen hinsichtlich ihrer ökologischen Effektivität und Kosteneffizienz sowie Gerechtigkeit bei den Experimenten mit unterschiedlichen Versuchspersonen ähnlich ist. Zudem lässt ein Vergleich des Verhaltens von Universitätsstudenten im besonders kontextualisierten Experiment mit ihrem Verhalten in einem kontextlosen (abstrakten) Experiment keine wesentlichen Unterschiede bei den untersuchten Kriterien erkennen. Zudem hat die Darstellung, d.h. die Einführung eines bestimmten Kontextproblems - in diesem Fall im Zusammenhang mit der Bewirtschaftung von organischen Böden - keinen Einfluss auf die Art und Weise, wie bestimmte Merkmale der Spielern die Entscheidungen beeinflussen.

GENERAL INTRODUCTION

1. Peatlands and climate change

1.1. An overview

Organic soils, sometimes referred to as peat soils, differ from mineral soils in their composition and the conditions under which they are created. They consist of a top layer of peat that consists of at least 30% (dry mass) of dead organic material (Joosten and Clarke, 2002) and an underlying layer of mineral soils. In a pristine state, organic soils develop under water-saturated, anoxic conditions enabling an accumulation of organic matter, the peat. This results in carbon sequestration from the atmosphere. A larger area with a peat layer and peat-forming vegetation is called a peatland. Globally, peatlands cover approximately 400 Mha, which represents about 3% of the world’s land area (Joosten and Couwenberg, 2009). The perceived benefits obtained from the exploitation of peat have been praised for centuries. Through extraction, peat from organic soils is exploited in temperate countries for producing fuel or as a growing media in horticulture. These activities represent about 0.1% of the global peatland area (WER, 2013). In some cases, this resource contributed significantly to the regional economy and to better livelihoods (Bonn et al., 2016; Chapman et al., 2003). Today, some European countries have forbidden the extraction of combustibles derived from peat. There is also an increasing use of substitutes for peat-based substrates in horticulture. Besides extraction, the provisioning services of organic soils are used for agricultural activities. Globally, about 14-20% of peatland area are under agricultural management (cf. International Peat Society1). While high-altitude peatlands (bogs) are being used mainly as grasslands, low-altitude peatlands (fens) are generally used for more intensive agricultural activities. After drainage, organic soils can be turned into cultivable lands that are suitable for certain types of agricultural production such as vegetable farming. Organic soils are light, easily cropped and do not require as much preparation as mineral soils. About 50% of peatland areas in Europe have been or are utilized for production activities or peat extraction (Joosten, 2010; WEC, 2013). The productive use of peat for agriculture is also a recent development trend in tropical countries, in Indonesia and Malaysia in particular, which together possess the third largest surface area of peat in the world after Russia and North America (WER, 2013). In these two South Asian countries, an increasing share of peatland area is deforested or burnt and then drained for the establishment of plantations such as palm oil trees. Despite the beneficial properties of peatland, which enable attractive and profitable activities in agricultural and industrial sectors, these diverse uses of peat are not sustainable and are associated with severe negative side effects. Most notably, organic soils constitute the most efficient terrestrial stock of carbon. They only cover 3% of the world’s land area but contain about one third of the global terrestrial carbon stock. This corresponds to twice the stock of carbon captured in the total global forest biomass (Joosten and Couwenberg, 2009). Therefore, the preservation of organic soils is highly relevant in the current context of climate change and climate change mitigation targets.

1.2. Climate change mitigation by peatlands

The process of carbon sequestration in peatlands is extremely slow. Under optimal conditions, organic soils form at rates of, on average, 0.4 mm per year (Loisel et al., 2014). Due to this very low formation

1 Web link: http://www.peatsociety.org/peatlands-and-peat/peatlands-and-climate-change, accessed 28.12.2016. 15 rate, peatlands essentially represent a non-renewable resource. Most importantly, any disturbance of organic soils halts the accumulation of organic matter and reverses the process: the soils convert from a carbon sink into a carbon source. Soil drainage in particular results in the compaction of the peat and the exposure of the organic matter to the atmosphere. The activation of soil microbial activity leads to the oxidation of the organic material that has been conserved for millennia. Thus, the material is lost through greenhouse gas emissions into the atmosphere and into water streams. The resulting soil loss results, in turn, in soil subsidence.

Globally, about 10% of the peatlands are degraded, and responsible for 2 Gt of CO2 emissions per year, which is “equivalent to 6% of all global anthropogenic CO2 emissions” (Joosten 2009; Banwart et al.,

2012). In the European Union, 75% of the CO2 emissions from agricultural land use are derived from degraded peatlands (Bonn et al., 2014). In many European countries such as the United Kingdom and Germany, less than 20% of organic soils remain in near natural conditions (Bonn et al., 2016). Measures to support climate change mitigation in the agricultural sector would therefore greatly benefit from harvesting the potential of mitigating emissions from cultivated organic soils. Utilizing the mitigation potential of peatlands requires the preservation of still-intact organic soils and in particular the restoration of degraded peatlands. In the past years, a growing number of scientific articles and books on restoration initiatives focus on peatlands and explore the link between peat degradation and land management practices. Yet, societal awareness of the consequences of peat degradation is very low. The concrete policy challenges and research questions with respect to the promotion of the conservation of organic soils addressed in this thesis are therefore highly relevant and timely. This thesis investigates the effectiveness of different economic incentives to promote sustainable management practices and reach conservation objectives. Other policy approaches, such as regulatory instruments aiming at restoring natural peatland are not part of this thesis.

2. Conservation measures for organic soils

2.1. Organic soils and current policy approaches

Until 2013, emissions from degraded and degrading peatlands under agricultural management had been largely ignored in national and international climate discussions (Regina et al., 2015; Dunn and Freeman, 2011). The Land Use, Land-Use Change and Forestry (LULUCF) framework, as part of the Kyoto Protocol, had mainly focused on forest biomass. There are several potential reasons for the late interest of policy makers and businesses for the preservation and restoration of peatlands. Those include the lack of data on the distribution of peatlands and on their use, and the complex relationship between rewetting degraded peatlands and its effect on carbon dynamics (see Roulet, 2000 for an example in Canada). The LULUCF now also gives attention to the peatlands used for agriculture and provides guidelines and recommendations to develop incentive schemes for the restoration of organic soils (Regina at al., 2015). Since the agreement made at the Durban COP-17 in 2011, LULUCF highlights peatlands as important resources in the mitigation of anthropogenic greenhouse gas emissions. Most importantly, it approved carbon savings from the restoration and rewetting of peatlands as a climate change mitigation strategy. Therefore, in the second commitment period of the Kyoto protocol (2013- 2017) and under Article 3.4 of the Kyoto Protocol “Wetland Drainage and Rewetting”, peatlands can be used to meet GHG emissions reduction targets at national levels (Bain et al., 2012; IPS, 2014; Bonn et al., 2014). Yet, only very few countries have implemented peatland conservation policies so far. The preservation of organic soils and the provision of peatland-related ecosystem services can be sustained or enhanced through a range of policy instruments: classic regulation (e.g. prohibition of 16 drainage), direct state control (i.e. via state owned land), financial instruments (e.g. taxes, subsidies), creation of carbon markets, and information provision (Reed et al., 2010). This thesis tests the potential of agri-environmental payment schemes, a type of financial instrument, to incentivize farmers to adopt conservation practices on organic soils. Under Pillar II of the EU Common Agricultural Policy, funds are available since 2008 to promote actions related to peatland restoration. Also, designations such as Special Areas of Conservation and Special Protection Areas, as well as Birds and Habitats Directives and the Natura 2000 network facilitate the protection of peatlands (Reed et al., 2010). However, the rare cases that use payments to incentivize peatland restoration (e.g., the Upland Entry Level Scheme in the UK; Reed et al., 2010) address extensively-used peatlands. This thesis concerns intensively-cultivated organic soils that are highly profitable. It does so by utilizing a case from Switzerland. Up to now, in most of the countries worldwide and in Switzerland in particular, there are no measures to effectively address the degradation of these intensively-cultivated organic soils and its multiple consequences. This thesis contributes to fill this gap by investigating agri-environmental payment schemes that could allow the preservation of such organic soils.

2.2. Policy design

The design of policy schemes in areas of intensively cultivated organic soils necessitates considering three major factors. First, intensive drainage is required on such soils. Preventing further degradation of these soils and associated negative environmental externalities can only be realized by diminishing the intensity of drainage on the land, and thus raising the water table (Schrier-Uijl et al., 2014; Jauhiainen et al., 2016). This is in conflict with conventional agricultural production because reducing drainage negatively affects farm productivity. There is a growing interest in potential uses of rewetted organic soils. However, farm profits generated by vegetable production (the current land use in the study region) and potential profits from extensive land use on rewetted organic soils differ considerably (such by a factor of 20) (Wichmann, 2016). This makes the promotion of sustainable use of organic soils very difficult without strong external intervention. The second key feature concerns the organization of the drainage systems, and the pumping stations in particular. Those are organized such that several farmers are sharing one drainage system. Therefore, opting for an alternative land use on this area through raising the water table is only possible if farmers cooperate among each other and agree on modifying the regime of water table regulation. A third key feature is the spatial variation of soil profiles. According to experts from the study region, soils highly vary in terms of thickness of the peat layer, state of degradation of the peat, and nature of the sediment that is underneath the peat layer (Figure 1). The thickness of the peat layer varies substantially across places, which translates into different time horizons of production on the peat across farmers, varying from less than 10 to more than 60 years. The underlying mineral soil layer is the key determining factor for the potential of agricultural activities after the peat layer has been lost. More specifically, within a same area, some farmers possess a fertile underneath mineral soil layer and therefore long-term production potential for their current land use. Others have a poor underneath mineral soil layer and therefore only short-term production potential for their current land use, linked to the thickness of the peat layer. This spatial variation in the suitability of the underlying mineral soil for the current land use leads to an asymmetry in the long-term opportunity costs of conservation among farmers. Opportunity costs are higher if the quality of the underneath mineral soil is higher.

Peat layer

Underneath mineral soil ?

Figure 1: Soil profile of organic soils Photo credit: Jens Leifeld. Note: This is a profile from Witzwil (Bern) in Seeland: a cropland on a so-called “Halbmoor” (Swiss classification for a Histosol that has mineral interlayers). Below about 140 cm is the underlying sediment.

Such heterogeneity implies that incentives among farmers with regard to undertaking a certain activity or enrolling in a certain program differ. This requires consideration by policy makers. Heterogeneity in opportunity costs is also an issue in other contexts of resource management. The effectiveness of agri- environmental schemes is sometimes questioned in reference to insufficient differentiation of payments between farmers and insufficient reflection on local conditions (Hodge and Adams, 2013). Considering such differences is of particular relevance when heterogeneous land users need to cooperate with each other. By exploring the combined setting of heterogeneous land users and the need for cooperation among them, this thesis contributes to understanding the specific resource management problem of peatlands in Switzerland and to the more general literature on agri-environmental payment design.

2.3. Agri-environmental payments

Agri-environmental payments are a form of payments for environmental services (PES). PES are promising incentives to promote ecosystem conservation activities (Engel et al., 2008). They seek to internalize what would otherwise be an externality. They are defined as a voluntary transaction where a well-defined environmental service is being ‘bought’ by at least one service buyer from at least one service provider if, and only if, the latter secures provision of the services (Wunder, 2005). PES programs are increasingly established in developing and developed country settings for the provision of sustainable ecosystem management (Engel et al., 2008). Two main issues regarding the design of PES are of particular interest and are investigated in this thesis. The first issue concerns payment schemes that enable to foster coordination among land users in the provision of environmental services such as “agglomeration payments” and “agglomeration bonus”. The second issue concerns the heterogeneity of conservation costs, which include opportunity costs for adopting an alternative activity. In theory, the minimum level of the payment should be at least equal to the opportunity cost of the farmer for adopting the conservation activity (Engel et al., 2008). In reality, the level of PES is commonly set as the average cost of the potential participants to the PES program (Wunder et al., 2008; Ezzine de Blas et al., 2016). Such average payments do not account for heterogeneities in opportunity costs among land users. Therefore, differentiated payments have been proposed as an alternative. This section reviews the literature that examines these two issues. For some ecological indicators, spatial coordination of conservation activities on farmlands increases the level of environmental benefits generated by these activities (Banerjee et al., 2014). Group-based

18 payments are relevant schemes to enroll proximate parcels that have high ecological values or to incentivize the creation of contiguous habitat network in the landscape (ibid). The concept of an agglomeration bonus was introduced by Parkhurst et al. (2002) as an incentive for spatial coordination of conservation areas. The agglomeration bonus consists of a base participation component and a bonus component. The bonus amount is a function of the number of farmers undertaking the joint activity. Drechsler et al. (2010) introduced the agglomeration payment, i.e. a payment for which the allocation is conditional on all farmers of the group undertaking the conservation activity. Using a theoretical approach, Drechsler et al. (ibid) find that agglomeration payments are on average more cost effective than spatially homogeneous payments that are allocated to land users regardless of the location of the habitat area being enrolled to the program. That is, agglomeration payments achieve a higher level of environmental services for a given budget. This result is confirmed by Wätzold and Drechsler (2010) who use a modelling approach comparing the performance of an agglomeration payment, an agglomeration bonus and a spatially homogenous payment under different types of economic and ecological parameters. Also, they find that the agglomeration bonus never significantly outperforms the other two options with regard to cost effectiveness (Wätzold and Drechsler, 2010). In this thesis, we therefore focus on the design of an agglomeration payment rather than on an agglomeration bonus.2 In practice, typical uniform payments are often perceived as more equitable and politically more acceptable (Pascual et al., 2010). However, conservation costs often vary among land users. Thus, uniform payments provide higher rents to low-cost producers and can therefore raise fairness issues among participating agents to the program (Pagiola et al., 2005). Uniform payments are also criticized for targeting, in some cases, unproductive lands that do not provide the expected environmental benefits (Osterberg, 1999). Therefore, differentiating payments according to land users’ conservation costs can potentially translate into significant gains in cost effectiveness of the PES program (Wünscher et al., 2008; Alix-Garcia et al., 2008; Armsworth et al., 2012). However, the design of differentiated payments is likely to generate additional costs as obtaining information on the ecological and economic aspects of ecosystem service provisions of individual farmers is generally associated with large transaction costs (Pascual et al., 2010). Nevertheless, these additional costs appear to be worth bearing as they appear to be significantly lower than the surplus payment that would be given to the farmers in the situation of a uniform payment scheme design (Armsworth et al., 2012). Heterogeneity in conservation costs has been theoretically considered in the design of group-based payments by Wätzold and Drechsler (2014) and Drechsler et al. (2010) who then also allow for side payments in their models.3 According to Drechsler et al. (2010), side payments appear to be relevant in settings where the agglomeration payment does not cover the participation costs of all land users. When payments are conditional on coordination among land users and based on average opportunity costs, some land users benefit excessively from the payment scheme while others tend to be undercompensated. This can be overcome by side payments. Thus, by relying on payment redistribution through side payments, a reasonable level of the payment is permitted, which increases the cost effectiveness of agglomeration payments (ibid; Wätzold and Drechsler, 2010). These previous

2 The literature on the agglomeration bonus is relevant. The selection of the social optimum by land users in presence of such an agglomeration bonus is on average facilitated by communication of information about the choice of the direct and indirect neighboring land users (Banerjee et al., 2014 in a laboratory experiment). This information contributes to reducing uncertainty in the choice of strategy and increasing the likelihood of successful coordination among land users (ibid). Group size of landowners is also recognized as a factor promoting successful coordination under such a payment scheme (Banerjee et al., 2012). The effect of these factors on players’ choices depend, however, on the type and features of the experiment, and on the players’ strategic environment (Banerjee et al., 2014). 3 According to Kimbrough and Sheremeta (2012), side payments are a means for two agents characterized by different incentives to commit to a situation where both agents would be better off if they are able to reach an agreement and share payoffs. Side payments are either binding or non-binding, and enable parties to solve a potentially conflicting situation. 19 theoretical analyses consisted however of simulations. This thesis experimentally investigates the potential of differentiated agglomeration payments to promote conservation activities.

3. Overview of the thesis4

The study region presented in this thesis is a western region of Switzerland called “Seeland”. It is characterized by a large proportion of farmlands on organic soils, which are historically drained and managed under intensive vegetable cultivation. The thesis consists of four chapters. Chapter 1 provides an overview of the current situation of vegetable producers in the Seeland region, and contributes to a better understanding of the challenges and potential for a change towards more sustainable management practices on these organic soils. It also highlights and compares the diverse implications of development on organic soils in two extreme scenarios: pursuing current management with intensive drainage versus switching to peat-preserving land uses after stopping the drainage. Most importantly, it highlights the need to take actions on these soils and to make well-informed decisions regarding their future use. Chapters 2 and 3 address policy options that aim to reduce the negative impacts of current management practices on organic soils. The two chapters capture the organic soil management issue in an innovative framed (contextualized) and computerized economic experiment that depicts the decision situation of vegetable producers on organic soils in the Seeland region. The study investigates the behavior of farmers when facing the land management trade-off of continuing vegetable farming versus adopting more sustainable management practices that require drastically diminishing the intensity of drainage of the soils. Both chapters test a situation in which there is no policy intervention and a situation where an agri-environmental payment scheme is introduced in exchange for adopting the sustainable practice on organic soils. Chapter 2 analyzes and compares the effects of three types of payments in a static experiment with university students. The payment types vary in whether they are uniform or differentiated and whether they are agglomeration payments or based on individual land use. Chapter 3 describes the testing of two types of agglomeration payments in a more complex dynamic experiment conducted with farm apprentices. The two agglomeration payment designs vary in whether they are constant and therefore uniform across farmer types or variable and thus potentially different across farmers. The dynamic experiment depicts the management problem more closely by depicting the temporal dynamics of the impact of land use decisions on soil productivity. Chapter 4 discusses the robustness of the results obtained in the previous two chapters by comparing the behavior of farm apprentices to the behavior of university students in the dynamic and static experiments. Second, it explores the importance of the experimental framing for behavioral outcomes. Specifically, the chapter compares the results obtained in chapters 2 and 3 in the case of a highly contextualized experiment to the results of an unframed (abstract) experiment. By analyzing the impacts of framing and sample population, the chapter contributes to the debate on the external validity of findings generated by experiments conducted with university students in abstract laboratory environments. A more detailed overview of the research questions and methods in the four chapters is provided below. The results are summarized and discussed in the conclusion chapter of this thesis.

4 This work is part of the “Sustainable management of organic soils in Switzerland” project, embedded in the Swiss National Research program 68, “Soil as a resource” [grant number 143145)] funded by the Swiss National Science Foundation. 20

4. Research questions and methods

4.1. Research question 1: What are the social, economic and environmental implications of alternative management scenarios in the study region?

In Europe, the few incentives targeting the conservation of agriculturally managed organic soils focus on extensively used areas. Some examples include the Site of Special Scientific Interest framework in England (UK) that gives legal protection to extensive areas of peat habitats (Defra, 2009) and the MoorFutures Program in Germany that enables companies to invest in peatlands restauration projects in order to offset their carbon emissions. Another example is the German “moorschonende Stauhaltung” program. This very newly developed payment scheme incentivizes farmers to diminish the intensity of drainage on their soils. In all these cases, foregone profits of farmers for adopting a higher level of extensification on their organic soils are low. Land use conversion is therefore economically reasonable. The issue is more delicate where organic soils are used for very profitable uses such as vegetable production. As a necessary step to the development of potential conservation measures on intensively- used organic soils, the thesis aims at providing a better understanding of the environment of farmers operating on these soils, using the case of the Seeland region. Productive use of peatlands in intensive agriculture is prioritized over their conservation despite the harmful environmental consequences of this management and the uncertainty in the long-term productive capacity of these soils. Chapter 1 aims at elucidating the social, economic and environmental implications of alternative management scenarios on organic soils. Thus, using secondary data as well as data collected through interviews with regional experts, the chapter provides the socio-economic and cultural context for the Seeland region and an overview of the situation associated with organic soil degradation under agricultural activities. It then exposes the challenges and obstacles that could appear with a potential change of management practices on organic soils and contributes to the understanding of potential alternative future scenarios of agricultural development on these soils. These considerations are highly relevant for the design of policy instruments, including payment schemes as studied in this thesis or other types of instruments. The chapter also highlights the potential feasibility and difficulties related to the implementation of different policy and market instruments aimed at promoting sustainable uses of organic soils.

4.2. Research question 2: Can agri-environmental payment schemes incentivize a more sustainable use of organic soils?

Economic incentive approaches such as environmental subsidies and taxes have been argued to be more cost-effective than command-and-control approaches in addressing environmental degradation (Perman et al., 2003). They also have the advantage to allow more flexibility and thus better adapt to heterogeneous situations. Among these economic incentives, the thesis hypothesizes that agri- environmental payments, which can be seen as a form of environmental subsidies (Engel et al., 2008), are the most feasible options to promote a sustainable use of organic soils. This is due to three main reasons. First, with regard to the enhancement of environmental services and public positive externalities, subsidies may be more easily accepted in the agricultural sector as compared to environmental taxes or command-and-control approaches. Second, the issue is politically sensible due to the high profitability of the soils, which means that land use restriction options or taxes would be difficult to implement. Third, the spatial variability in the suitability of the underneath mineral soil layer and thus in the future production potential among farmers may lead to fairness issues. All farmers possess equal immediate opportunity costs for switching to sustainable land use on their soils but differ

21 in their long-term opportunity costs as well as in the extent to which they can provide the environmental outcome (spatial variability of the peat layer thickness and hence the carbon stock). Therefore, policy instruments like taxes that are based on the polluter-pays principle are likely to face strong farmers’ opposition with regard to distributional concerns. Furthermore, this study does not aim at cost- minimizing policy but puts emphasize on the protection of the peat for which the value is very high. Using an experimental and behavioral economics approach, chapters 2 and 3 provide an analysis of policy options that can enable a change in management practice on organic soils and thus overcome the high opportunity costs of adopting measures for sustainable use of organic soils. Both chapters explore the effect of selected agri-environmental payment schemes on the adoption of a sustainable use of organic soils. They present a baseline in which farmers are in a situation without external incentives, and compare the resulting behavior to a situation where an agri-environmental payment scheme promoting sustainable use of organic soils is introduced. For this, the two chapters use an experimental economics approach. The experimental design captures the key components of the decision situation of vegetable producers on organic soils in a computerized and framed experiment and analyze the potential behavior of these farmers using a ten-round decision procedure. The experimental economics literature comprises studies that use visualized simulation games, which contextualize the experiment by incorporating the socio-ecological context more explicitly and accounting for the dynamics of some of the socio-ecological ecosystem parameters, e.g., Janssen et al., 2010; Cardenas et al., 2013. Some of these experimental tools have an educational purpose (see Janssen et al., 2014 for a review). Yet, these studies use tokens and cells or magnets, and present a lower level of visualization than the present study. In the experiments, players are placed in pairs that represent groups of farmers whose farmlands are located within the same drainage system perimeter. To reflect the asymmetry among farmers in their future production potential from the soils, each pair of players consists of a farmer who has a high- quality underneath mineral soil layer and a farmer who has a low-quality underneath mineral soil layer.5 The farm profits of the latter are therefore affected by peat degradation resulting from vegetable farming. Consequently, farmers differ in their incentives to adopt sustainable use of these soils. The experimental design is such that the two farmer profiles have the same initial thickness of the peat layer (about one meter of peat). Under vegetable farming, they would therefore meet the underlying mineral soil layer at the same time. This simplification is necessary in order to disentangle the various effects of the experiment. The executed experiments thus contribute to the identification of a policy design that could be successful in promoting a sustainable use of organic soils. They also contribute to an understanding of human behavior in this unique setting, and to the literature on PES design by combining payment differentiation and agglomeration payments and comparing these to more conventional PES designs.

o Research question 2.1: Which agri-environmental payment design is best suited to promote sustainable peatland use when budgets are constrained? First, the thesis hypothesizes that agglomeration payments that are differentiated based on opportunity costs are a promising policy approach for promoting cooperation between land users in the adoption of a sustainable peatland use that requires a change in the drainage system. The design of the payment options analyzed in chapter 2 accounts for the fact that policy budgets are often limited. Payments that are designed uniformly among farmers and aligned with the highest opportunity costs may not be

5 The thickness of peat layers and nature of the underlying soils are uncertain to most farmers in the region (see chapter 1). Yet, this spatial variation in production potentials is a key factor to change of land management practices by farmers. The lack of knowledge on production potentials is mainly due to important constraints at farm management level. However, this uncertainty is, in practice, easy to solve and would allow to decrease information asymmetries among farmers. Hence, the experiments assume players that are heterogeneous in their production potentials, and that are aware of it. 22 feasible because they are too costly. This is particularly true in the context of this study as the opportunity costs of farmers of switching from conventional intensive to sustainable use of organic soils are very large. Therefore, the design in chapter 2 considers a uniform payment that is based on average opportunity costs. Such average payment design is also commonly used in practice as it is often thought to be easier to implement than differentiated payments (Wunder et al., 2008). Although such a uniform payment would induce farmers’ costs to be compensated to different degrees, the need for farmers to cooperate in order to obtain the agglomeration payment may help solve the social dilemma if farmers can redistribute payments among them. The experimental setup therefore allows explicitly for side payments between farmers. The second hypothesis here is that a uniform agglomeration payment based on average opportunity costs could also be effective in promoting sustainable use of organic soils because farmers will make use of the possibility for payoff redistribution via voluntary side payments. To summarize, chapter 2 compares the effectiveness and efficiency of three types of agri-environmental payment schemes: a uniform agglomeration payment, a uniform individual payment (executed regardless of the land use decision of the other member in the group), and an agglomeration payment that mirrors players’ opportunity costs. In this chapter, experiments are conducted with university students.

o Research question 2.2: Can agglomeration payments induce sustainable management of organic soils in a dynamic setting? The thesis hypothesizes that differentiated agglomeration payments may create a particularly high incentive among land users to adopt a sustainable land use. This form of payment scheme is hypothesized to be more environmentally and cost effective than uniform agglomeration payments. Chapter 3 adds to chapter 2 by simulating the dynamic effect of farmers’ decisions on the state of the peat layer and consequently on farm profit over time. This implies that the incentives of farmers for the adoption of a sustainable land use may vary with time. Indeed, the heterogeneity in production potential from the soils is time- and action- dependent. Specifically, the farmer with high-quality underlying soils could keep producing vegetables after the peat layer is degraded. Thus, this farmer’s opportunity costs stay constant over time. By contrast, the farmer with low-quality underlying soils faces decreasing productivity in producing vegetables over time as the peat layer gets thinner. While incorporating this aspect in the experimental design adds complexity to the experiment, it captures the resource management problem more closely. Chapter 3 focuses on the effectiveness and efficiency of two types of agglomeration payments. The first policy type is based on the opportunity costs of farmers in the first time period. Given that those are equal for both farmers, this payment remains constant over time. This treatment captures the idea that policy design is often based on current conditions and ignores the dynamics of opportunity costs over time. The second payment treatment follows the dynamic evolvement of opportunity costs: it starts out equal, stays constant over time for the farmer with the high quality soil, but decreases over time for the farmer with the low quality soil. We refer to the two treatments as constant and variable agglomeration payment. The constant payment implies that farmers with the low-quality underlying soil are excessively compensated as the peat gets degraded and farmers’ opportunity costs differ. This may lead to bargaining between farmers because of the fact that the agglomeration payment is conditional on group adoption of the sustainable land use. Therefore, the experimental set up allows again for side-payments among group members. This time, the experiment is conducted with a sample population similar to the actual farmers concerned with the issue, namely farm apprentices from the western part of Switzerland.

4.3. Research question 3: Do the type of subject and experimental contextualization affect experimental results and policy implications?

Chapter 4 explores the external validity of outcomes generated by abstract laboratory economic experiments that are conducted with university students, and therefore the robustness of the results obtained in chapters 2 and 3. Several studies question the extent to which results obtained from experimental studies undertaken with university students can be extrapolated to the general population and to the field (Harrison and List, 2004; Exadaktylos et al., 2013). This question is especially important if experimental results are used to support policy recommendations. The thesis hypothesizes that it is relevant to conduct such experiments with subjects that are similar to the ones concerned by the issue addressed. Using both static and dynamic experiments, chapter 4 contributes to this body of literature by comparing the behavior of university students to the behavior of farm apprentices in the framed experiment. A second type of common criticism of experimental economic studies refers to the fact that the tasks offered to the participants are usually very abstract and tend not to reflect the issue addressed. Thus, behavior in the field, which is influenced by a multiplicity of contextual factors, may be incorrectly represented in those experiments. Again, this is of particular importance if the results are used to derive policy recommendations. Thus, the thesis hypothesizes that the framing of the experiment with the actual context could affect players’ behavior and potentially increase the accuracy of players’ decisions as predictors of behavior. Based on experiments conducted with university students, chapter 4 compares the results of an unframed (abstract) and a framed (contextualized) experiment, in both static and dynamic experimental settings.

CHAPTER 1 Sustainable management of cultivated peatlands in Switzerland: insights, challenges, and opportunities1

1. Introduction

Peatlands provide a diverse range of ecosystem services, most importantly provisioning services (peat as a substrate), and regulating services including climate regulation via carbon sequestration and storage, water regulation and water purification (Bonn et al., 2016).2 Globally, about 30% of the terrestrial organic carbon pool is stored in peatlands, while they cover only 3% of the global land area (Parish et al., 2008; Yu et al., 2010; Scharlemann et al., 2014). Waterlogged and thus anoxic conditions are a prerequisite for these unique ecosystems to prevent the rapid decomposition of accumulated organic matter (Rydin and Jeglum, 2013). Organic soils build over millennia, at rates of 0.4 mm/year, before reaching their equilibrium height, which approximately corresponds to the level of the water table (Loisel et al., 2014; Clymo, 1984). By acting as a carbon sink, peatlands represent a huge carbon store. Considering the long-time scale, peat is seen as a non-renewable resource (Joosten, 2007). However, many peatlands either no longer provide these vital services or their provision is highly threatened. In Europe, around 50% of peatlands have been or are utilized for agriculture, forestry or peat extraction (Joosten, 2010; WEC, 2013). Agriculture varies from extensive pastures (Oleszczuk et al., 2008) to intensive cultivations: vegetable production in the UK and Switzerland, maize grown for fodder and biogas generation in Germany, or grassland for dairy farming in the Netherlands. Conventional agriculture on peatlands requires drainage to regulate the soil water table to a level that allows crop production (Verhoeven and Setter, 2010). While drained organic soils offer good physical conditions for crop production (FAO, 2016), drainage causes soil subsidence from both physical compaction and the increased microbial respiration of the organic matter that is exposed to the atmosphere (Everett, 1983). The peat that has been conserved for millennia is oxidized, which results in huge greenhouse gas

(GHG) emissions. Emissions from peatlands drained for arable farming are on average 34 t CO2- eq/ha/year and thus, per unit area, almost 60 times higher than emissions due to soil carbon loss from managed mineral soils (Leifeld, 2013; IPCC 2014). Together with the instances of peat fires, global emissions from peat correspond to two thirds of emissions from global deforestation (Leifeld, 2013).

Globally, degraded peatlands contribute to about 6% of all anthropogenic CO2 emissions (Joosten, 2009). This illustrates the particular mitigation potential when taking action on drained peatlands in the context of anthropogenic GHG emissions. Protecting peatlands is not only necessary for climate change mitigation. The loss of soil also challenges their productive use (Verhoeven and Setter, 2010). At some point in time, depending on the original thickness of the peat deposits, all peat will be oxidized and lost. The thickness of the peat layer in Seeland

1 This chapter corresponds to an article co-written with Adrian Muller (Institute for Environmental Decisions, ETH Zürich), Jens Leifeld and Cédric Bader (Agroscope ART Reckenholz-Tänikon, Swiss Federal Research Station for Agroecology and Agriculture, Zürich), Moritz Müller (Bern University of Applied Science, School of Agricultural, Forest, and Food sciences, Bern), Stefanie Engel (Alexander-von-Humboldt Professorship of Environmental Economics, Institute of Environmental Systems Research, University of Osnabrück), and Sabine Wichmann (Institute of Botany and Landscape Ecology, University of Greifswald). Financial support from the Swiss National Science foundation is gratefully acknowledged. We are very thankful to the experts from the Seeland region who took time to share their views on the issue of managed organic soils. 2 Further ecosystem services provided by peatlands are supporting services that include biodiversity, accumulation of organic matter, nutrient cycling; cultural services that include aesthetic and educational opportunities (MEA, 2005). 26 is largely unknown and is very variable across places. Agricultural production will then have to be continued on the underlying material, which provides different qualities of sediment, some well suited and others unsuitable for intensive agriculture. Especially in coastal regions, subsidence increases the cost of water management, the risk of floods and finally leads to the loss of productive land. Despite the negative implications of soil loss in these productive areas, in most countries no regulations are in place to address this issue. Policies addressing GHG emissions from cultivated organic soils are lacking (Regina et al., 2016). A shift towards peat-preserving extensive use requires higher water tables, which is in conflict with intensive management practices. This is a delicate issue at the farm level but also at political and market levels. A growing body of literature investigates strategies and practices to restore degraded organic soils by rewetting, including the introduction of land use practices that are adapted to high water levels such as paludiculture (e.g. Bonn et al., 2016; Wilson et al., 2011; Gaudig et al., 2013; Joosten et al., 2015; Wichtmann et al., 2011, 2016). Most studies apply a plot-level focus and assess the emission reductions that result from such restoration activities and the potential of biomass production on rewetted sites. Only few studies address the consequences of a range of possible management changes at the farm or landscape level considering the socio-economic context of current intensive agricultural use (e.g. Schaller, 2014; Röder and Osterburg, 2012; Graves and Morris, 2013). We contribute to closing this research gap, focusing on the case of the Seeland region in Switzerland. In Switzerland, most organic soils are already gone; the remaining area covers in total less than 1% of Switzerland’s acreage, of which 90% is degrading (Wüst-Galley et al., 2015; Wüst-Galley and Leifeld 2017). However, these soils still store 30 Mt of organic carbon. If current practices remain the same and the peat is depleted, the remaining carbon in these organic soils has the potential to emit 100 Mt CO2- eq within the next century, plus corresponding amounts of N2O (Wüst-Galley et al., 2015). This is twice of Switzerland’s total annual GHG emissions (FOEN, 2016a). This study is therefore of direct relevance for Switzerland and Seeland in particular, and of interest for other regions where organic soils are intensively managed. To avoid that the natural process of peat loss forces decisions upon society, society itself needs to take action in time initiating necessary changes in a well-governed process. Based on secondary data and interviews with experts knowledgeable about Seeland, we provide an overview of the different facets of intensive agricultural production on organic soils in Seeland and within the wider Swiss context. We then explore alternative future management options for these soils and potential policy instruments for their protection, and conclude.

2. Study region and methods

2.1. Study area

Organic soils are characterized by an upper peat layer of more than 10 cm that contains more than 20% organic carbon by weight (IPCC, 2006). In Switzerland, these soils represent 28,000 ha, which corresponds to less than 1% of the total country area and approximately 2% of Swiss agricultural land (Wüst-Galley et al., 2015; Figure 2). Agricultural use of peatlands has a long tradition in Switzerland. Most peatlands are drained and under varying degrees of management intensity (FOEN, 2014). This study examines the case of the Seeland region (“three lakes region”) in western Switzerland. Two large- scale hydrological correction measures were conducted to prevent flooding around the lakes of Neuchâtel, Morat and Biel, and permitted the development of agricultural lands (OED, 2013; FOEN, 2014). The first correction (1868-1891) established canals and lowered the water level of the lakes.

However, significant soil subsidence necessitated a second correction (1962-1973) where mainly the first constructions were reinforced to further keep the water level low and limit its variations (ibid). The Seeland consists of two main areas: “the great swamp” (“Grosses Moos” in German or “Grand marais” in French) located in the center region and the plain of Orbe in the South of lake Neuchâtel (Figure 3). Furthermore, the Seeland is characterized by large areas of organic soils that are drained and used for vegetable production since more than a century. This region is the country’s largest market gardening area. For many decades, the drainage (‘melioration’) of the peatlands has been highly appreciated for developing economic perspectives for the local population in one of the poorest regions in Switzerland as well as for contributing to the national food production and security. Nowadays, the problems caused by peat degradation are getting more attention. In addition to the threat to peat, many of the drainage systems in this region would require replacement or renovation within the next few years (Zollinger, 2006; Butorin, 2015a). This is a costly procedure and funding for this, which in the past has mainly been public, is uncertain. This issue is not yet clearly addressed among policy actors, one reason being its complexity.

Figure 2: Distribution of organic soils in Switzerland Note: red circle: Seeland region; Source: Country outline © Swisstopo; image adapted from Wüst-Galley et al., 2015. In pink are surfaces for which there is evidence of organic soils in recent times (since 1950, classes I, II, III, IV, V in the report). They are recommended estimates of organic soils and cover 28,000 ha in total. In dark blue are surfaces for which there is only historical evidence of organic soil (pre-1950, classes VI, VII in the report, ca. 7,000 ha). Surfaces for which there is no information about organic soil or for which information is ambiguous are not included (class VIII in the report).

France

Switzerland "The Great Swamp"

The Plain of Orbe

Figure 3: The Seeland region and the two main areas of vegetable production 28

2.2. Methods

The following presents a review of the relevant scientific and policy literature, including governmental and other reports, as well as own data collection including expert interviews. Data from experts was collected using semi-structured interviews conducted in 2015. In total, eleven experts were interviewed (see details in Appendix A). They were selected with the objective to have a diverse range of expertise, from soil scientists to persons working closely to farmers and persons involved in the water regulation systems of the area. Each interview lasted between 1.5 and 3 hours and aimed at a better understanding of the farming situation on organic soils. Depending on the expertise, a number of different aspects were addressed, such as the challenges faced by farmers, their perception with regard to the future development of these soils, their participation to conservation programs, and others. Henceforth, information from the interviews are referred to as “expert name, oral comm.”.

3. Vegetable production on organic soils

3.1. The economics of vegetable production in Switzerland

Among other uses, organic soils support the production of vegetables. Today, Swiss horticultural production is taking place on about 10,000 ha (including all types of vegetables and both outdoor and greenhouse productions). This corresponds to 0.6% of the total agricultural land in Switzerland (FSO, 2016; Swisscofel, 2014). Yet, horticulture contributes to 14% of the economic value of the total agricultural output in Switzerland (Swisscofel, 2014). The main vegetable crops produced are carrots, iceberg lettuce, onions, broccoli, spinach and beans. Domestic vegetable production represents about 50% of the domestic demand. During the peak production season, when domestic supplies of vegetables meet national demand, imports have high customs duties imposed (Chevalley, 2012; USP, 2015). The majority of vegetable producers belong to a cooperative either linked to Coop or to Migros that together represent more than half of the retailer market (Summermatter, 2012; Mann et al., 2011). For each product, the harvests of the farmers are collected through a system of “platforms of commercialization”. Each platform is monitored by one farmer that other farmers join to fill the order (volume) placed by the retailer (Trachsel, oral comm.; Swisscofel, 2016). Vegetable producers are under big pressure by the retailers (UMS, 2017; Müller, Trachsel, Grob, oral comm.). Price variability is high and producers need to comply with tight deadlines in delivery and with high quality requirements at product and packaging levels (USP, 2015). This leads to a frequent surplus in the supply of these products (Chevalley, 2012). Major structural change occurred in the sector: the number of vegetable producers declined from approximately 4,500 in the year 1996 to 3,100 in 2013, whereas the total area devoted to vegetable production on lands rose from approximately 6,500 to 8,200 ha (CCM, 2016; Swisscofel, 2014). Vegetable producers have developed several strategies to remain competitive: i) farm growth: farmers compete for additional land; ii) farm specialization: vegetable producers have significantly reduced the number of different types of vegetables grown and exchange land with cereal farmers during the winter (Grob, oral comm.); and iii) increased farm mechanization. Compared to other farmers, vegetable producers are characterized by a high level of entrepreneurship and reliance on external labor force (Mann, Lüscher, oral comm.). This economic environment generates high inequalities in farmers’ incomes (Müller, Trachsel, oral comm.), as big farms have a tendency to get bigger while smaller farms may not afford expanding their land. More than 20% of the Swiss vegetable production is generated in Seeland - an area of 2,500 ha with about 500 vegetable producers (Spavetti, 2011), and to a large extent on drained organic soils. In this region, farmers often possess a natural mix of both organic and mineral soils on their farms, which is 29 beneficial as they can balance the risk between the different types of vegetables produced on farm along the year (Lichtenhahn, oral comm.). Differences in the structure and water retention capacity of these two soil types would ideally go along with different field-management approaches (ibid). Yet, mineral and organic soils are often managed the same way (Trachsel, oral comm.). This is partly due to the fact that the proportion and distribution of the different soils is not well known (ibid).

3.2. Requirements for intensive agricultural management on organic soils

Water management is the key factor for productivity on peatlands. In Seeland, drainage systems consist of gravity systems complemented by drains connected to pumping stations. Their efficiency depends on soil slope, the age of the channels, depth, direction, and distance from each other, but also on soil structure and hydraulic conductivity (Lüscher and Chappuis, 1984; FAO, 2005). The systems are built with very deep main drains so that only the “local drainage systems” need to be replaced over time. Due to former events of flooding, peat layers in these fens (groundwater connected low peatlands) often include layers of poorly permeable sediments that prevent water from infiltration (Lüscher, oral comm.). To overcome this, water permeable gravel is locally inserted in the drains (ibid). Drainage systems are usually owned by the municipality but maintained by a syndicate (the ‘Flurgenossenschaft’ cooperative) (ibid). They should be cleaned every two years depending on the extent of deposits inside the pipes. The pumping station is managed by an employee or a farmer. Individual farmers have generally no control over the water table.3 A general assembly of all farmers served by a joint pumping station (ca. 6 to 50 farmers on 500 to 1,000 ha) decide once a year on the water table of the following year. The decision is mainly based on empirical observations as farmers aim at avoiding water-saturated soils (ibid). Any modification of the water regime therefore requires unanimous agreement and cooperation among farmers. The investment cost for a drainage system depends on the local conditions and the state of the collectors. It is estimated to cost approximately 14,000 CHF/ha plus an additional 5,500 CHF/ha for a complete system with pumping station (Lüscher, Jaton, oral comm.). This total cost is amortized in 30 to 50 years depending on the depth of the drains, subsidence rates, and maintenance (ibid). The maintenance costs (high pressure rinse; excluding costs of electricity) amount to about 80 CHF/ha/year (Lüscher, oral comm.). The investment cost of the drainage systems is shared between the canton (usually more than 50%), the Confederation, and private landowners. The shares of contributions vary across cantons. There is currently a lack of clarity on the future funding for renewing these systems (ibid).

3.3. Challenges of intensive cultivation on organic soils

3.3.1. Consequences of intensive agricultural management practices on organic soils The acreage of organic soils has decreased from 135,000 ha initially to 28,000 ha today (Wüst-Galley et al., 2015, Figure 2). This loss is quasi irreversible. This highlights the urgency to take action on the remaining peat area. Current volumetric soil losses in organic soils used for agriculture vary between 10 and 20 mm/year in the Seeland (Leifeld et al., 2011). The total subsidence of the peat layer results from the combination of three processes: two physical processes that are primary consolidation of low density peat layers at the beginning of drainage as a result of loss of supporting pore water pressure followed by shrinkage resulting from evaporation (Figure 4), and the biological oxidation and corresponding mass loss of the peat material itself (Kasimir-Klemedtsson et al., 1997). Subsidence results in a decreasing

3 Zmoos enterprise in canton Neuchâtel developed a system involving a sluice gate in order to isolate a system (e.g., a parcel) from the others. That would allow a change in water management and thus production on selected lands without affecting the others. However, such a system is not currently used or maintained by farmers. 30 distance between groundwater level and soil surface over time, requiring a continuous correction of the water table in order to maintain crop production.

Figure 4: Pore volume and soil bulk density of degrading organic soils in Seeland

Note: Numbers below 80 cm depth represent undisturbed peat. Towards the surface (degraded peat), pore volume declines to values also found in mineral soils. At the same time, bulk density of soil sharply increases. Source: Leifeld et al. (2011).

While drainage increased provisioning services in form of vegetable crop production, other services such as water regulation at the watershed level, carbon sequestration, habitat function and peatland- specific biodiversity have been lost. The management of these soils with the gain of high value generation in agricultural products induced a shift from providing positive to negative environmental externalities (e.g. GHG emissions, carbon and nitrogen leaching, reduced retention function for pollutants, risk of local flooding due to soil subsidence). Despite this evidence, no major change in the management of the soils has been undertaken to counter the process of degradation of peat. The implications of soil subsidence and loss of peat for agricultural production are also already visible. Based on the interviews, they are as follows:  Production conditions deteriorate depending on the underlying sediments that become increasingly important with lower peat levels. While the soil working substrate usually consists of the first 40 to 50 cm of soil, in some locations there is either less than 40 cm of peat left or the peat is completely oxidized already.  With subsidence, the risk of inundation increases, posing a risk for agricultural use of organic soils. In some places, the draining intensity is no longer sufficient to lower the water level for efficient intensive agricultural production (Butorin, 2015a). Often, subsidence is spatially very heterogeneous, making field traffic with machineries more challenging.  In several locations in Seeland, the drainage systems are no longer fully functional and require replacement, which is normally expected 40 or 50 years after installation.

 Peat degradation and drainage may increase the vulnerability of soils to wind erosion with corresponding adverse effects.

3.3.2. Actions undertaken by farmers to counter subsidence and peat degradation In order to maintain the productivity level of the soils in Seeland and remain in the market, farmers cope with the issues resulting from peat degradation in various ways (a Porta, 2016). To diminish the overall rate of degradation, some areas implement more moderate draining in seasons outside the sowing and harvesting periods. This also results in diminishing costs of active draining (Miéville-Ott and Praz, 2004). Some farmers possess the equipment to dig local water drains in order to increase drainage efficiency where needed. Others work on raising the ground level and leveling out the terrain through covering the current soil with the addition of allochthonous (external) mineral soil (e.g. the Magadino plain in Southern Switzerland and the canton of Saint Gallen) (Lüscher, oral comm.; Sprecher, 2015; FOEN, 2015). This method resembles peatland melioration techniques implemented in the Netherlands and Germany 150 years ago (Göttlich and Kuntze, 1980). Yet, it is a risky procedure. Depending on the original soil profile and the type and amount of material imported, the results are mixed, e.g. it can lead to higher rates of subsidence via compression of the underlying peat (Müller, Trachsel, oral comm.), but it may also reduce further subsidence by peat oxidation (Zeitz, 2016). Deep ploughing peat and mineral layers up to 1.5 m and then mixing both layers in the top 30 cm has also been conducted to increase the potential of production. This operation was conducted in the domain of Witzwil, canton of Bern, and led to stabilized soil surfaces (Butorin, 2015b). This technique, which is expensive, may be seen as relevant when the peat layer is thin (40 cm, up to 1 m) and the underlying material is of good quality

(e.g. sandy sediment) (Lüscher, oral comm.). However, its long-term effects and impact on CO2 emissions are uncertain. The results from the interviews seem to indicate that most of the farmers prioritize overcoming seasonal production challenges over the reduction of long-run environmental impacts resulting from soil degradation. Farmers’ actions are therefore short-term responses to the negative impacts of soil subsidence for their own agricultural productivity. The issue of soil loss is not addressed in the long term and farmers operating on these soils will eventually have to change their practices (SSSS, 2015). There is a growing concern with regard to the future of vegetable farming in Seeland (Butorin, 2015a; Sprecher, 2015). Some farmers think that evening out the soil would enable a more moderated water table regulation and thus a higher level of soil humidity, and a lower rate of degradation of the peat. Indeed, it would avoid setting the water table according to a few location points for which the ground level is lower than average (Egger, Stoll, oral comm.). Furthermore, some farmers state that they would like to invest in long-term development strategies to stop the reduction of peat and preserve the soil potential (ibid, oral comm.). Yet, no systematic measures have been developed to tackle the peat degradation and soil loss and there is so far no widely acknowledged solution.

4. Alternative management practices on organic soils and challenges to their implementation

4.1. Options for sustainable management of organic soils

Preservation of peat and efficient reduction of CO2 emissions are only achieved by rewetting, and thus restoring anoxic conditions for the organic matter (Bonn et al., 2016). When rewetted, organic soils can no longer support intensive arable production such as vegetables. Land uses that can suit high water tables and can thus halt peat loss are receiving growing attention, e.g. by the FAO (Biancalani and

Avagyan 2014). The concept of paludiculture, from the latin ‘palus’ for swamp, refers to the harvest of biomass on water-saturated peatland sites (Wichtmann and Joosten, 2007; Joosten et al, 2016; Wichtmann et al., 2016). On fens, it includes the cultivation of a diverse range of plants species such as reed (Phragmites australis), sedge (Carex spec.), and cattail (Typha spec.) that can be used for energy generation, building purposes, and roofing, alder (Alnus glutinosa) for high-value timber production or short rotation coppice with willow (Salix spec.) or alder as well as medical plants (Abel et al., 2013, Wichtmann et al., 2016). On bogs, cultivating peat mosses (Sphagnum) as constituent of growing media for professional horticulture is a promising option (Gaudig et al., 2014). Extensive high-quality livestock production is another possibility on rewetted organic soils, including highland cattle, and water buffalo that adapt well to high water table, which allows to reduce peat loss (Sweers et al., 2013). Under current temperate climatic conditions, known peat preserving land use is at farm level much less economically profitable than current vegetable production. However, there are a growing number of studies that investigate the potential of productive activities on wetlands or rewetted organic soils and their economic perspectives (e.g. Schäfer and Joosten, 2005; Sweers et al., 2014; Wichmann and Köbbing, 2015; Wichmann, 2016). Collecting medical plants from natural peatlands is a widespread tradition, but methods for cultivation have yet to be developed in many cases (Abel, 2016). The cultivation of different medical plants was tested in Northeast Germany identifying e.g. Meadowsweet (Filipendula ulmaria), Valerian (Valeriana officinale) and Butterbur (Petasites hybridus) as economically promising (Kersten et al., 1999). Cultivating plants for pharmaceuticals and cosmetics on rewetted peatlands may be an option for the intensively used Seeland region, too, since it keeps high employment rates and addresses markets promising high revenues. Moreover, several edible plant species can grow in wet peatlands such as Celery (Apium graveolens) and Wild Rice (Zizania palustris) (Abel, 2016). Under a future climate change scenario with higher temperatures, also the cultivation of rice (Oryza sativa) could allow for reducing subsidence and may be an option to explore (Hatala et al., 2012). Despite the generation of GHG emissions from rice fields, rice production could permit the preservation of peat during a specific time of the year; more research is necessary though. A study by Graves and Morris (2013) in the UK analyzed economic differences between organic soil restoration, conservation options, and continued arable production, under different climate change scenarios. They indicate that climate change increases the relative “advantage of peatland restoration and conservation options over the longer term”. Furthermore, Musarika et al. (2017) showed that raising the water table to −30 cm not only reduced GHG emissions considerably but also increased productivity of vegetable crops like radish.

4.2. Challenges of alternative management practices on organic soils

We detail here five main challenges to a sustainable use of organic soils in Seeland, which we then summarize in Table 1. First, at farm level, the high profitability of current management practices on these soils, which translates into substantial opportunity costs for switching to sustainable management practices, is a major barrier to the consideration of such alternative practices. Organic soils in Seeland are indeed highly fertile (Mann, 2005; Peringer, 2014). For instance, in situations like the construction of public infrastructure that necessitate compensating farmers for the loss of land, the Swiss tax commission may, depending on the interests of the parties involved, attribute a higher value to organic soils than to mineral soils (Jaton, Deriaz, oral comm.). Yet, if the land is sold and placed on the land market, the value of the farm is usually defined within an average range of market prices where acreage rather than soil value is considered (Müller, Mann, oral comm.), thus providing little incentive for long- term considerations of soil degradation. The fact that organic and mineral soils are equally valuable on the land market is likely a result of the scarcity of land and the existence of high competition among farmers to extend their acreage. Based on data from AGRIDEA and from the ‘Centrale Suisse de la 33 culture maraîchère et des cultures spéciales’ (Swiss center for vegetable production and special cultures) (Möhring et al., 2012), the average gross margin for vegetable production, including direct payments, is estimated at around 14,000 CHF/ha/year (Agridea). We compare this with the alternative of growing Common Reed as a possible peat-preserving activity on rewetted organic soils. Reed cultivation is currently seen as one of the most viable land uses on rewetted organic soils. Its utilization has a long tradition in Europe; efficient harvesting machinery is available and a market demand for thatching reed exists (Wichmann and Köbbing 2015). Yet, it would be necessary to analyze its market potential for the case of Switzerland specifically. The gross margin of Common Reed production varies from about 190€/202CHF (biomass used as energy generation by combustion) to 608€/650CHF/ha/year (if used as a construction material) (Wichmann, 2016).4 In contrast to the gross margin of biomass production that does not include any subsidies, the gross margin of vegetable production includes direct payments (on average equal to 1,200 CHF/ha/year; AGRIDEA). This is a common situation in Switzerland and in Europe, which distorts the value of drainage-based agriculture on these soils. But even when accounting for the differences in subsidies the gross margins of these different production systems differ by a factor of 20. A second challenge to sustainable use of organic soils is the economic working environment of vegetable producers. Pressure from retailers, time constraints, and important investments at collective (i.e. drainage syndicate) and individual levels (i.e. machinery, buildings) are the main factors dictating farm management choices (Mann et al., 2011; Miéville-Ott and Praz, 2004). Farmers have therefore very little opportunity to reconsider practices, explore potential adaptation to soil subsidence at the farm level, and elaborate on long-term planning (Müller, Deriaz, Trachsel, oral comm.). Yet, some farmers would like to be given chances to think through long-term adaptation strategies on these soils. Third, from a cultural and social perspective, the Seeland region, also called “Gemüseland” (“land of vegetables”), carries a strong identity and pride that were shaped by its historical management. This region was poorly developed 150 years ago, leading to a decrease in population. The conversion of peatlands to farmland required very strong investment efforts by the past generations to reach today’s production potential. Vegetable producers are still perceived today as hard working people. It is very difficult for them to imagine an extensive farming of these highly productive soils, which they consider to be best valorized through intensive production (Mann, Trachsel, oral comm.; Miéville-Ott and Praz, 2004). Fourth, detailed plot or farm-specific knowledge on the state of the peat in those cultivated areas are lacking (last soil survey from the 1970s (Verein Seeland.biel/bienne, 2016)). First, the spatial distribution of organic soils is unknown, which makes any policy measure development difficult (Müller, oral comm.). Second, the thickness of the remaining peat layer in Seeland varies considerably in the region from 0 m to 3 m due to the varying conditions under which it was formed, microbial activity, and past management practices (see Leifeld et al., 2011; Bader et al., 2017). This translates into different time horizons of production on the organic soils among farmers. Again, precise information is missing and therefore it is rather unclear when the remaining peat layer will be depleted as it already occurred at many sites in Seeland (Butorin, 2015b). This is partly due to the high variability and uncertainty in emissions factors from these soils and peat densities, which depend on drainage and land management. Third, the soil profile is usually not studied. Farmers commonly focus on the first 30/50 cm of soil (the working substrate) and current soil fertility and yield are on average not affected by peat degradation (Trachsel, Müller, oral comm.). In places where peat layers are more than one-meter thick

4 The gross margin is the total income derived from the farming system minus the variable costs incurred in the enterprise. The total income comprises the sales and direct payment received by the farm. The variable or direct costs are costs that are assignable to several specific activities. Net income (or profit) is gross margin minus fixed costs. The fixed costs are not activity-assignable and correspond to the general management of the farm (e.g. investment, buildings). 34 and where drainage is still effective, the visibility of the problem is therefore low. Thus, little is known about the properties of the soils and the nature of the underlying mineral soil layer, which varies from heavy soils and clay to silty, sandy sediments with very different levels of fertility (ibid, oral comm.). It is therefore uncertain which plots will become of low fertility after all peat is gone, which plots will retain high fertility for vegetable production, and if there are any measures to improve soil fertility after peat exhaustion. In addition, the ground levels in Seeland are heterogeneous and uncertain (some may already be below the level of the lakes). It is therefore not yet clear how reducing drainage intensity would affect the soils spatially, which further complicates the design of specific measures to address this issue. Intensively used organic soils, characterized by deep drainage, tillage and high fertilizer applications, are hot spots of GHG emissions. Their rewetting has large mitigation effects and generates benefits that exceed conservation costs (Schaller, 2014; Röder & Osterburg 2012). Efforts for rewetting and adapted peatland management should therefore concentrate on sites where a large carbon stock still remains. In response to these various uncertainties, the association ProAgricultura Seeland is launching a project (2016-2020) that aims at mapping the organic soils in part of the Seeland region (the “Grosses Moos” in particular) (Verein Seeland.biel/bienne, 2016). Fifth, a change of land use on the organic soils in Seeland implies to elaborate solutions for the transfer of the corresponding volume of vegetable production. One the one hand, Switzerland aims at maintaining the level of domestic production. On the other hand, a significant part of the country’s agricultural land is already used intensively. This is therefore an additional challenge to the adoption of sustainable land use in Seeland. We address this point further in a next section.

Table 1: Main challenges to a possible change in management practice on organic soils in Seeland

Main challenges Implications High profitability of soils High opportunity costs of adopting alternative practices Difficult economic working Difficult to integrate new practices environment for producers Strong regional identity for Intensive production perceived as best valorization of organic soils vegetable farming - Unknown spatial variability in soil production potentials and remaining carbon stock Lack of data related to soil - Difficulty in designing adequate policy measures promoting alternative management profile and soil properties of the soils Transfer of the volume of Finding a suitable way to produce this volume elsewhere, in Switzerland vegetable produced in Seeland

5. Policy instruments to support alternative management practices

5.1. Institutional context in Switzerland

With the 1987 Rothenthurm’s initiative, peatlands of national importance are now protected. This includes 1,500 ha of bogs and 20,000 ha of fens (Klaus, 2007). However, not all of these so-called peatlands actually accumulate peat. Many of the fens were classified as peatland from a botanical, not from a soil science perspective (FSO, 2004; FOEN, 2002). Moreover, only a small share of these

“preserved” peatlands are undergoing restoration. The rest is degraded and is still emitting CO2. Many of the degrading organic soils that are agriculturally used are not included in the list of protected sites (ibid). Among those, the small-scale (isolated) non-degraded peatlands (e.g. less than 0.5 ha) are highly 35 threatened by intensive management practices (Klaus, 2007). Because of strong hydrologic perturbations, those are often irretrievably lost (ibid). Swiss agricultural policies, implemented at a cantonal level, include protection measures on farm-lands that aim to ensure that the “land is used for sustainable farming” (OECD, 2010), preserving soil fertility over generations, and enhancing landscape biodiversity. Since 1998, the allocation of direct payments to farmers is conditional upon regulation in farming operations and the adhesion of farmers to certain ecological requirements (i.e. system of cross-compliance) (FOAG, 2017). The ‘Proof of Ecological Performance’ (PEP) includes a balanced use of fertilizers, an adequate share of ecological compensation areas, a regular crop rotation, adequate soil protection, a targeted and limited use of pesticides, an appropriate nutrient balance, and appropriate livestock farming (FOAG, 2015; Joerin, 2007; OECD, 2010). The measures for soil protection mainly address the risks of soil erosion and compaction, and soil composition in polluting minerals (FOEN, 2013).5 For some of these ecological requirements, vegetable producers are not subject to the same regulations as other farmers. For instance, they are only required to set aside 3.5% of their land as ecological compensation areas while other farmers are subject to a minimum level of 7% (AGRIDEA, 2014; OECD, 2010).6 Beyond these ecological standards, additional ecological services are incentivized by targeted agri-environmental payments (e.g. payments for organic farming (Schader et al., 2008)). Across all farmer types, 95% fulfill the cross-compliance requirements (Joerin, 2007). However, the localization of these areas follows the logic of economic optimization, which does not enable the achievement of environmental objectives. The current level of enforcement of ecological measures is reported as insufficient for reaching the desired results in terms of biodiversity and landscape (Jahrl, 2012). In general, farmers see themselves as producers and do not like setting land aside that is “doing nothing” (Mann, Jaton, oral comm.; Howley et al., 2015). This feeling seems to be even stronger for vegetable producers on organic soils, for whom regulation of ecological compensation areas are not widely accepted (Grob, oral comm.). Scarcity of land and time for vegetable production combined with high economic pressure make it difficult for vegetable farmers, and for the small ones in particular, to subscribe to voluntary agro-ecological programs and thus release productive soils for agri- environmental measures (ibid). Thus, the cross-compliance requirements are unlikely to sufficiently address the risk of soil degradation (FOAG, 2017; Günter, 2002; Regina et al., 2016). While Switzerland has statutory benchmark values for soil and water erosion (Verheijen et al., 2009; Prasuhn et al., 2013), no measures address the issue of soil loss on cultivated organic soils. The problem of degradation of intensively used organic soils is recognized by policy makers, but there is no legal instrument specifically designed for their conservation on farm land (Havlicek, oral comm.).

5.2. Policies to promote sustainable use of organic soils in Switzerland

5.2.1. Payments for Environmental Services Payments for environmental services (PES) are a potential option to preserve organic soils and the remaining stock of carbon on farm lands (Klaus, 2007). They are seen as a reward in exchange of producing ecosystem services. A PES is defined as a voluntary transaction where a well-defined environmental service (i.e. sustainable use of organic soils) is being ‘bought’ by at least one service buyer from at least one service provider (here, the vegetable producers) if, and only if, the latter secures provision of the services (Wunder, 2005). This means that the farmer adopts alternative management practices on organic soils and the peat is effectively preserved. The level of payment needs to at least

5 Art. 6, al. 2, OSol; Art. 6, al. 1 ; OSol. Art. 27, al. 1, LEaux (FOEN, FOAG, 2013) 6 Ecological compensation areas take various forms, e.g. extensive meadows, pastures, cultivated strips free of fertilizers and pesticides (OECD, 2010). 36 compensate the opportunity cost of farmers for converting land to another use (Engel et al., 2008), i.e., the net loss in profits from giving up potential economic activities on rewetted organic soils. Through precise information on the localization of organic soils and their relevance for society in terms of remaining carbon stock, it will be necessary to calculate the overall cost of such measure for Switzerland. A challenge in the design of PES for sustainable use of peatlands lies in the fact that rewetting involves a collective decision of the farmers sharing a drainage system. A form of PES that can address such need for coordinated action among several farmers is called agglomeration payments (Drechsler et al., 2010). Agglomeration payments are payments made conditional on coordinated activities. In chapter 3, alternative policy options promoting sustainable management practices on organic soils were tested in a computerized, interactive economic experiment capturing the key aspects of the problem of intensively cultivated organic soils. The experiment was specifically designed within the frame of this project and is contextualized based on the situation in Seeland. It was conducted with farm apprentices from the Western region of Switzerland. Thereby, information was gathered on how farmers would decide between pursuing vegetable farming versus adopting sustainable practices on organic soils. In addition to a baseline scenario without policy intervention, two types of agglomeration payment designs that promote the adoption of sustainable management practices were tested. Chapters 2 and 3 show that agglomeration payments indeed have the potential to promote the adoption of sustainable management practices on organic soils. There are different potential funding sources for such payments, which are linked to different service buyers, including the state, buyers of carbon certificates, and environmental organizations. Those options are discussed next.  Public agri-environmental payments Public agri-environmental payments funded by the government would take the form of a subsidy allocated to vegetable producers in exchange for adopting sustainable use of organic soils. It is also important to highlight that vegetable production is currently heavily subsidized by the Swiss government. Moreover, payments for non-food biomaterials production such as Miscanthus were still in place a few years ago but they have been cancelled with the Agricultural Policy 2014-17, except for organic production of these crops. In Germany (state of Brandenburg), a very recent program for peat conserving water retention (“Moorschonende Stauhaltung”) offers a payment to farmers for keeping the water table high (10 cm in winter to 30 cm in summer below the soil surface). However, it mainly targets extensively used grasslands. In Switzerland, the existing direct payments from the government targeting either the promotion of landscape quality or the sustainable use of natural resources could in principle fit the goal of preservation of organic soils.7 This would require defining and paving for specific measures targeting the use of organic soils.  Carbon credits Within the voluntary carbon market, the introduction of certificates being generated by the rewetting of drained peatlands is another potential tool to compensate farmers and promote the preservation of organic soils. Such a carbon certificates program is already established in Germany with the MoorFutures© program (Joosten et al., 2015) and in the UK with the development of the Peatland Code (IUCN, 2015). Such a program offers companies or intermediary organizations the opportunity to invest in the rehabilitation of degraded peatlands and thereby obtain carbon credits, which they can use to voluntarily offset their carbon emissions (Van Beukering et al., 2008). In practice, intermediaries either buy land from farmers or pay farmers for adopting peat-restoring measures such as modification of the soil water regulation regime. Seiler (2014) explored the potential of such a voluntary market in Switzerland, investigating the willingness of different companies to pay for carbon credits from more sustainable peatland management using analogous crediting conditions as under the MoorFutures

7 At the level of the European Union, Pillar 2 of the Common Agricultural Policy (CAP) is the key instrument to stimulate land use change on cultivated organic soils (Glenk et al., 2014). 37 program. The obstacles to the implementation of such a program were reported to include high opportunity costs associated with rewetting and the very low public awareness of this issue (ibid). Regarding the compliance market, carbon offsets from land use change can currently not be used in mandatory Swiss climate mitigation activities. However, since the Paris agreement (COP 21), discussions in the Federal Council to allow for this are under-way (FOEN, 2016b).  Investments by organizations Investments by environmental organizations could also enable the compensation of the opportunity costs of farmers for rewetting organic soils. Some organizations are specialized in investing in climate protection projects. One possible investor is KliK (Swiss Foundation for Climate Protection and Carbon Offset). KliK is a foundation that operates as a carbon offset group for mineral oil companies responsible for selling fossil motor fuels for consumption, and handles the fulfilment of the legal carbon offset obligation for CO2 emissions resulting from the use of such fuels. It invests the funds put at its disposal in Swiss climate protection projects that have been demonstrated to be effective and fully meet the provisions of the Swiss CO2 laws. If Swiss regulations would include sustainable peatland management in the list of permitted mitigation activities, investments by KliK could become an option.

5.2.2. Product labelling The ideal reward for farmers in exchange for providing environmental services is in form of higher prices for their products (Mann, Trachsel, oral comm.). In respect to this farming identity (Burton et al., 2008), an alternative incentive could therefore consist of valorizing certain products through the creation of labels or certification (e.g. “peat-free vegetables”) (Joosten and Clarke, 2002). For example, vegetables grown on mineral soils in Seeland or products generated from an extensive use on rewetted organic soils could be subject to higher prices through a “Seeland label” (mentioned by Trachsel in Pick- up (2003); Dahms and Schäfer, 2016). Such a gain in price or market share would, however, be unlikely to fully compensate the farmer’s loss of profits on the organic soils. Compensation payments would thus still be warranted. Furthermore, the creation of such a label would need to be well thought through; some labels for low carbon-footprint products already exist (e.g. “Climatop” products in Migros).

5.2.3. Other policy options A downside of the already mentioned options is the necessary time required for their implementation in terms of policy design and implementation, establishment of business partnership (for carbon credits), and/or creation of market demand. This may not match the nature of the issue of managed organic soils, which is urgent and needs rapid action. More fundamental changes in the economic incentive system may be warranted. The system of direct payments indirectly encourages current drainage based farming. Redesigning environmentally harmful subsidies is therefore most urgent (Wichmann et al., 2016). Other options include cross-compliance measures, such as making allocation of direct payments conditional on decreasing the intensity of drainage. Also, an environmental tax (the polluter-pays principle) may be implemented. For example, farmers could be charged a tax for a drainage intensity above a certain level. Further options include a “service buyer” that buys or leases the land from farmers. The service buyer could be either the state or environmental organizations (e.g. KliK). Such leasing or sale contracts may enable immediate conservation effects on these soils, while more durable solutions could be developed over time. However, this option would contradict with the idea of farming identity that is strongly present in Seeland, which may also complicate the procedure for setting the price of the land. An additional option could consist of reducing or stopping the public funding for the drainage systems. Notably, the extraction of peat for the horticultural sector is now forbidden in Switzerland but peat is still imported for this very same use. Similar leakage could occur in case Switzerland decides to prevent prohibit agricultural activities causing degradation of peat, which is an additional potential regulation.

Such a reduction in production could come at the expense of increasing imports of vegetables in the short term. Besides, changing farmers’ behavior, including their acceptance of voluntary incentives such as PES, requires understanding their motivations beyond pure profit maximization (Ryan et al., 2010; Knowler & Bradshaw, 2007). For instance, in the project “Mit Vielfalt punkten - Bauern beleben die Natur” (“Scoring the Biodiversity”; Birrer et al., 2015), farmers highlighted the importance of pro- environmental motivations, revenue security, work-load, and societal gratitude in the creation or implementation of ecological measures (Jahrl, 2012). Such motivations are likely to play a role for vegetable farmers, too. Chapters 3 and 4 analyze the potential importance of personal characteristics on the decision by future farmers to pursue vegetable farming versus adopting sustainable practices on organic soils, with and without agri-environmental payments. The results show that the opinion of cooperative approaches, the level of patience, the willingness to take risks, reputational concerns and social preferences are factors that influence the decision of farmers significantly. Consideration of those aspects may thus be important in the design of payment schemes

6. Scenarios for the long-term development of organic soils

The current situation appears to be such that debate and consideration of alternative management practices will not be undertaken until organic soils are depleted, so that farmers can no longer pursue the current practice and are forced to adapt their practices. In coordination with the concerned farmers, other actors such as policy makers need to take responsibility and engage society in discussing the issue of degradation of organic soils in order to evaluate the potentials of alternative practices. As highlighted above, if preserving the peat and the ecosystem functions related to it are prioritized by society, several options and mechanisms are available. A societal debate about such options requires an understanding of the possible scenarios of development of organic soils and their implications. In this section, two scenarios for development of organic soils in the study region are discussed.

6.1. Scenario 1: Continuing vegetable farming on organic soils

Considering the ongoing process of soil loss, pursuing cropping on these sites can only be guaranteed by “meliorating” subsiding soils. This mainly translates into the addition of external mineral soil in order to maintain a soil surface level allowing for crop production. Yet, in many places the site characteristics may not allow soil melioration. In the absence of such operation, organic soils degrade and eventually completely disappear. The nature of the underlying mineral soil layer will shape the type of production that would be feasible after the peat layer is exhausted. Farmers with underlying soils of poor quality will face a more difficult situation and will probably not be able to pursue the current activity unless comprehensive soil melioration measures are put into place (Verein Seeland.biel/bienne, 2016). The machinery investment, very specific to vegetable farming, may also be a major constraint to any transition to alternative production (Trachsel, oral comm). In cases where the underlying soil cannot suit intensive agricultural activity, the only option will likely be an extensification of land use (Lüscher, oral comm.). In any case, drainage systems need to be maintained and replaced (Zollinger, 2006), which is very costly (cf. section 3.2). In the absence of melioration techniques, drainage will still be necessary when the peat is gone. First, given the low soil surface levels, drainage will be required to enable intensive agricultural production on the underlying sediments if such is still possible, and to prevent flooded areas in case the soil surface is below the level of the lakes. Second, considering the dependence of the farmers in the

39 watershed on soil water regulation, poor-quality underlying sediments may still be drained to enable production on other lands despite not being profitable. The main consequence of this scenario is the loss of a precious resource in areas that will in any case have to modify their production activities eventually and where management costs are very high. Long- term need to be considered, including when the peat layer disappears in some locations. Whether alternative land uses are adopted now or later when soil degradation forces such a change, it will imply substantial restructuration and changes at farm system levels with social implications. First, areas where the share of organic soils on farm-land is high will be highly affected as their economy depends on agriculture. Second, small farms do not have the capacity to adopt a new production system without external support. Third, it will be difficult for farmers to pass on their farm if the underlying soil does not allow the current practices.

6.2. Scenario 2: Rewetting organic soils

We also consider a scenario wherein organic soils are rewetted in Seeland in order to preserve the remaining peat. As mentioned above, various management options, to preserve the peat are possible depending on the location, such as paludiculture (Schumann and Joosten, 2008; Kotowski et al., 2016). From the social perspective, stopping intensive farm management and benefiting from PESs may be particularly relevant in places where the peat layer is deep. This requires identifying those areas. Thus, a significant contribution to the GHG emission reduction goal of the Swiss Federal Office for Agriculture (emission reduction of one third by 2050 relative to 1990; FOAG, 2013) could be achieved.

Measures allowing the offset of CO2 emissions from soils used for agriculture, including managed organic soils, are currently being discussed (FOEN, 2016b). A couple of factors need to be considered to increase the feasibility of this scenario. First, public awareness on the issue of organic soils is currently very low. Increasing awareness of the ecosystem services lost by peatland degradation would likely increase the willingness to support peatland conservation or restoration. The increase in demand for environmentally friendly products over the past years may favor such a change (Chevalley, 2012). Also, retailers are working on reducing their GHG emissions and on the development of more environmentally-friendly products (e.g. “Climatop” products). This context may be favorable for eliciting society’s preferences regarding the future management of organic soils in Switzerland. Furthermore, young farmers are well educated and have a high level of awareness of environmental and longer-run productivity issues. A positive correlation between farmers’ education and the adoption of environmental programs has been reported (Mann, oral comm.). Young farmers may be less experienced compared to the older ones but see themselves more as enterprise managers. They are highly interested in being profitable and are open to innovation and adoption of new practices. Finally, another encouraging factor is the strong potential for adaptability among vegetable producers, often described as the most innovative farmers (Miéville-Ott and Praz, 2004). As mentioned earlier, agglomeration payments were identified as promising policy options to incentivize cooperation among farmers as required for rewetting and adopting sustainable use of organic soils (chapter 3).

6.3. Impacts of scenarios 1 and 2

Whether organic soils are rewetted or the peat is depleted (implying that part of the lands can no longer be used for intensive production), both scenarios lead to important structural changes. Assuming a 50/50 mix of mineral vs. organic soils in Seeland, the volume of vegetables on organic soils would correspond 40 to about 10% of the domestic production. The replacement of this volume that can no longer be produced on organic soils when these soils are either depleted or rewetted needs to be studied. Potential options to relocate this volume include an increase of imports, an intensification of production at other places in Switzerland, and an investment into greenhouses. The first option may not be preferred as Switzerland aims at protecting domestic production (cf. the protective customs on imported goods). The second option could imply to increase intensity of production on mineral soils and on areas that are less suited for vegetable production with corresponding potential adverse effects such as increased erosion. In general, this option is likely to be difficult as a significant part of the Swiss agriculturally used area is already used intensively. With respect to the third option, the Netherlands, Germany, France, and Spain are particularly advanced in this technology; it is highly profitable. In order to ensure a low carbon footprint, technologically advanced greenhouses, e.g., with soil-less substrates, could be an option (Muller et al., 2016). This would require analyzing the water requirements, the carbon footprint, and potential negative externalities generated by vegetable production in greenhouses, as well as solutions for overcoming those. An advantage of greenhouses is that they could ensure that the “know-how” of vegetable production is kept within Switzerland (Trachsel, Grob, oral comm.), although vegetable production in greenhouses partly requires different or additional techniques. In practice, these greenhouses could be built on mineral soils outside Seeland in order to avoid imposing a low water table on organic soils. A more visionary option that is already initiated in the Netherlands consists of floating greenhouses (Beytes, 2011; Talbot, 2007). Thus, greenhouses could be placed on organic soils with high water table, which would allow maximization of the land use and maximization of the preservation of the peat. Elaborating on the concept of vertical farming (Banerjee, 2014), storey-building floating greenhouses (several vertical production layers) would be extra further innovative step. In all this, it has to be emphasized that rewetting the organic soils results in conservation of the carbon and thus the mitigation service of these soils. It does however not restore the original ecosystem of a peatland, which would need several centuries to rebuild, if at all. Additional factors would need to be evaluated: the societal acceptance of a higher number of greenhouses in the landscape, the training of farmers for such an alternative way of production, an analysis of the types of vegetable which can actually be produced this way, and implications for and acceptability of a change within the farm system (investments required, working the whole year). Several experts consider greenhouses as a promising option (Jaton, Trachsel, Grob, oral comm.). Some of the farmers may indeed look at it positively. However, this requires detailed planning, including consideration of broader landscape/territorial impacts in terms of land use allocation, accounting for the agricultural, industrial, and urban areas (Jaton, oral comm.). To realize optimal effects from such a territorial reorganization, it would necessitate a precise understanding on the soil heterogeneity distribution.

7. Conclusion

Organic soils are hotspot resources in the current context of climate change. However, in some regions, the short and medium-term economic benefits of utilizing organic soils are currently prioritized over the environmental benefits from their conservation. With this study, we have enabled a better understanding of the context, the main drivers, and the complexity of a potential change in farm management on these highly profitable organic soils. First, we observe that the loss of peat in intensively farmed areas is already causing substantial difficulties for some farmers. However, there is no long-term adaptation planning at the farm level nor is there an investigation into the production potential of the soil in the future. This is explained mainly

41 by the difficult and uncertain economic environment of the producers. Second, at the policy level, there is a crucial lack of measures to specifically address the external effects on society induced by the degradation of organic soils in intensively used areas and to preserve the remaining peat and therefore, the carbon stock. This seems to be mainly due to a lack of data on the distribution of soils and their properties. Projects just starting (cf. ProAgricultura) aim at filling this gap in the next few years. Finally, the general awareness of the public on this issue and its negative consequences at environmental and social levels is very low. If the current situation continues, the peat will irreversibly disappear. This will substantially contribute to national GHG emissions and affect the farming situation. Indeed, in some locations, the underlying soil does not suit current agricultural activity. Thus, there is a crucial need to take conscious and informed decisions on the future use of these soils: further pursuing the intensive drainage and production activities, shifting to alternative land uses that permit the preservation of the peat, or studying other possible alternatives such as soil melioration measures that might allow continuous use of the soils but possibly without preserving the peat. The former scenario (status-quo) will imply further investments in the drainage system and a restructuring of farmlands once the peat has disappeared. This includes the continued costs of drainage wherever the underlying mineral soil still allows some profitable agricultural activities. The latter scenario (shift to a sustainable use of organic soils) implies the development of policy measures incentivizing farmers to change the way they are using their organic soils. Different types of policy options exist. Both scenarios eventually imply significant structural change: the status- quo scenario later, the sustainable-use scenario sooner. A change in land use on these soils at some point is very likely and will depend on site conditions. It therefore seems societally preferable to change now, thereby avoiding environmental impacts and costs of drainage system renewal/maintenance. A societal debate and reflection is needed on the future use of organic soils that essentially are a non-renewable resource. Such a debate should consider all ecosystem services provided by peatlands and the trade-offs between them. This stresses the need for detailed information on the distribution of the organic soils and depth of the peat layer, not only for the Seeland region but for whole Switzerland, to identify the nationally most important peatlands where highest societal benefits can be generated. Moreover, it should involve the potentially concerned vegetable producers as well as other farmers. Studies eliciting the preferences of the Swiss population on these ecosystem services and the willingness of society, farmers and politicians to preserve these soils are needed.

CHAPTER 2 Incentives for sustainable land use considering cost heterogeneity among farmers: Results from a computerized framed experiment1

1. Introduction

Agri-environmental payments are increasingly popular for incentivizing farmers to adopt sustainable land use practices. They are a form of payments for environmental services (PES) (Engel et al., 2008). A number of PES design issues have been discussed in the literature, including environmental effectiveness and equity impacts in particular (for a recent review see Engel, 2016). Two issues of relevance to our study are agglomeration payments and payment differentiation. Agglomeration payments or boni are a promising approach to promote farmer cooperation (Parkhurst et al., 2002; Parkhurst and Shogren, 2007). Such coordination is often required to achieve environmental objectives. Examples include spatial coordination of activities to promote biodiversity, and activities that require farmer cooperation for implementation. Our study presents an example of the latter, where sustainable use of peatlands requires changes in jointly managed drainage systems. Agglomeration payments are paid if and only if all members of a group of farmers adopt the sustainable activity (Drechsler et al., 2007). Agglomeration boni are similar in concept, but involve only part of the total payment to be conditional on group behavior (Parkhurst et al., 2002). Several related studies investigate the performance of agglomeration payments and boni empirically and demonstrate their potential to foster coordination and cooperation between farmers (Banerjee et al., 2012, 2014, 2015). In practice, spatial variation in resource and farmer characteristics lead to different individual costs among land users in the implementation of conservation measures. Wätzold and Drechsler (2014) and Drechsler et al. (2010) theoretically examine the potential of group-based payments accounting for such cost heterogeneity. To our knowledge, however, all empirical and experimental studies on agglomeration payments assume that farmers are homogenous in their conservation costs and other characteristics. A second strand of literature has dealt with heterogeneity of provision costs focusing on the case of individual PES, i.e. payments made conditional on individual land use. Several studies have shown that differentiating payments according to land users’ conservation costs can lead to significant gains in cost effectiveness of the PES program (Wünscher et al., 2008; Alix-Garcia et al., 2008; Armsworth et al., 2012). Such an approach, however, brings its own difficulties. When costs are hidden or difficult to evaluate, asymmetric information between landowners and conservation agents with regard to the costs of supplying environmental services are at the origin of inefficiencies such as informational rents (Ferraro, 2008). One approach to address this asymmetry are procurement auctions that incentivize farmers to reveal their “true” conservation costs (Ferraro, 2008; Baylis et al., 2008). Uniform payment rates are still most common in PES (Wunder et al., 2008; Ezzine de Blas et al., 2016). While uniform payments are often perceived as more equitable (Pascual et al., 2010), others argue that they can be less

1 This chapter derives from a paper co-written with Stefanie Engel (Alexander-von-Humboldt Professorship of Environmental Economics, Institute of Environmental Systems Research, University of Osnabrück) and Elisabeth Gsottbauer (Institute of Public Finance, University of Innsbruck). Financial support from the Swiss National Science foundation and the Alexander von Humboldt-Foundation are gratefully acknowledged. We are very grateful to the students from ETH Zurich and University of Zurich who participated in this study. Thanks are due to Adrian Muller for support and feedback in this project. 44 equitable because they provide higher rents to low-cost producers (Pagiola et al., 2005). Furthermore, in contexts where conservation costs are considerable, payments that both cover land users’ opportunity costs and are designed uniformly, i.e. aligned on the highest opportunity costs, can be too costly. In practice, uniform payments are often based on average conservation costs across potential participants (Wunder et al., 2008). The performance of differentiated vs. uniform payments has not been analyzed for the case when farmers need to coordinate in a conservation program. In settings that combine high conservation costs and the necessity for farmers to coordinate, payments designed as an average across participants’ opportunity costs could potentially cope with the need for uniform payments in situations where budget is constrained. As for individual payments, such design would imply that farmers differ in the extent to which the payment covers their costs from adopting the conservation activity. Farmers with low opportunity costs would be excessively compensated while farmers with high opportunity costs would be insufficiently compensated by the payment. Thus, farmers’ incentives to cooperate would differ. But because agglomeration payments are conditional on the adoption of sustainable practices by all farmers in a group, this heterogeneity would likely trigger negotiations over side payments among farmers. Side payments, also called transfer payments, are voluntary payments that can be made among individuals of a same group in order to compensate for unequal exchanges. Only one recent study has considered the possibility of side payments between land users receiving PES in a more abstract, unframed context and with a focus on how side payments and cooperation depend on initial wealth distribution (Gsottbauer and Engel, 2016). The study presented in this chapter is inspired by the issue of drained former peatland areas in Switzerland, which are characterized by intensive vegetable farming on organic soils leading to soil subsidence and substantial contributions to national greenhouse gas emissions. Organic soils are the most efficient store of terrestrial carbon. In their natural (wet) state, they contribute to mitigate climate change by sequestering carbon from the atmosphere. However, as in our study region, conventional agricultural activity on organic soil requires drainage and thus leads to the degradation of the peat in the upper soil layer above the water table (Joosten, 2009; Bonn et al., 2016). Conventional land use is highly profitable in such areas. Therefore, promoting the adoption of more sustainable practices would necessitate considerable funds to cover farmers’ opportunity costs of conservation. Furthermore, soil production potentials and therefore long-term opportunity costs are heterogeneous among vegetable producers because soil properties spatially vary across different land parcels situated in the same peatland landscape. While, in the presence of peat farmers face similar immediate opportunity costs of adopting a sustainable land use, the production potential influences the long-term vision of farmers on their soils and therefore their incentive for adopting the sustainable land use. It is therefore a key component to the policy design. Moreover, the adoption of a more sustainable (peat preserving) land use implies to rewet the organic soils. This requires farmers to cooperate as they depend on a joint drainage system. Using a framed computerized economic experiment that captures the core components of the management problem of organic soils in Switzerland, we investigate how different agri-environmental payment designs perform in a setting with heterogeneous players and where their cooperation is necessary to achieve the environmental outcomes. We evaluate these policy designs with respect to three policy performance criteria: environmental effectiveness, cost effectiveness and income equity. We test three types of policy instruments: an agglomeration payment that follows respective players’ opportunity costs, a uniform agglomeration payment based on average costs, and a uniform individual payment. The latter is also based on average costs, but not conditional on other players’ conservation action. We examine and compare how players of different cost types behave under each of these payment

45 schemes, both with respect to land use choice and side payments made. We also analyze the impact of individual level socio-demographic characteristics and social preferences on behavior. We find that the three payment types are equally effective in promoting the adoption of more sustainable practices on organic soils, when compared to a situation where no payment is in place. Notably, despite the low incentive of players with high opportunity cost to adopt an alternative land use in presence of a uniform payment scheme, we find many groups succeeding in implementing the cooperative outcome. In these groups, we observe considerable payoff redistribution through side payments aiming to create an incentive for high-cost players to adopt the sustainable land use. Yet, we find that compared to the other two payment treatments, the differentiated agglomeration payment is the most cost effective option but leads to highest income inequality among players. The chapter is structured as follows. Section 2 presents our experimental design and behavioral predictions. Section 3 presents the results of the experiment, and Section 4 concludes.

2. Methods

2.1. Experimental set up

We designed an interactive and framed economic experiment that depicts the core components of the decision situation of vegetable producers operating on organic soils in Switzerland and conducted it with students from Swiss Universities.2,3 The framing of the experiment aimed at embedding players’ decisions in an environment simulating the real study context, the “Seeland” in Switzerland.4 At the outset of the main experiment, we also assessed players’ social preferences with an incentivized Social Value Orientation test (SVO) (Murphy et al., 2011). This test allows us to distinguish three social- preference types of players, namely “individualistic” players who focus on maximizing their own payoff, “inequality averse” on minimizing the difference between their own and the other players’ payoffs, and “joint maximizers” on maximizing the group’s joint payoff (ibid). At the end of the experiment, participants filled out a short survey for more detailed individual level information. The main experimental game was set-up as follows. Participants were provided with a short description about the agricultural management situation on organic soils, and were requested to take the role of vegetable producers operating under a joint drainage system with another producer. Players were then placed anonymously into groups of two: one with high opportunity costs of conservation (H player) and one with low cost (L player). This was motivated by the fact that organic soils differ by the type of the underlying mineral layer, implying different levels of future soil fertility and productive potential for vegetable production. This translates into different players’ long-term incomes that can be made from the conventional land use and therefore into different long-term opportunity costs of switching to an alternative, less profitable activity. We represented the situation via a two-stage decision procedure taking the form of a vote. In stage 1 each group member was asked to vote either in favor of or against rewetting the organic soils. The results of this vote affect the land use options and related payoff matrices in stage 2. In stage 2, participants decided either to adopt sustainable use of the soils (activity A) or to farm vegetables on drained soils

2 The experiment is web based. Technically, the experiment consists of a console interface, from which the experiment is piloted, and players’ interfaces. The experimenter can see in live all players’ actions on the console, which helps to control the evolvement of the experiment. Finally, data from the experiment are immediately available, in a dataset hosted by MongoDB. A screen shot of the experiment is provided in Appendix B1 and instructions can be found in Appendix B2. 3 These Swiss universities include the University of Zurich and the Swiss Federal Institute of Technology in Zurich (ETHZ). 4 The impacts of framing are assessed in chapter 4. 46

(activity B). Activity A requires rewetting, which can only be done collectively due to the joint drainage system. The procedure is as follows: o If at least one of the members rejected rewetting, the area remained drained and vegetable farming was the only possible land use option in stage 2 (i.e. activity B is chosen by default). In this case, H’s farm profit were equal to 4500 and L’s farm profit were equal to 25005 (cf. Figure 5). We refer to this situation in this chapter as ‘the status quo’ because it reflects the current situation in the study area. This term was not used in the experiment. o If both members voted in favor of rewetting, players could choose between the two land use options in stage 2. They could either adopt a sustainable use of organic soils (activity A; with a payoff of 250) or install a costly personal drainage system and continue vegetable production (activity B; profits of 4500 or 2500 as above, but subtracting a personal drainage cost of 220) (cf. Figure 6).6 For this basic decision set-up, we first conducted a 10-round baseline phase. This was followed by a 10- round treatment phase in which we introduced one of three agri-environmental payment policy designs to incentivize the conservation of organic soils. We rematched groups between baseline and treatment phase. The payment designs were as follows: i.) a differentiated agglomeration payment (DA) which accounted for differences in the opportunity cost of conservation of both players. Moreover, it was made conditional on both players pursuing activity A. Specifically, DA was set equal to players’ respective opportunity costs from adopting activity A (4250 for H and 2250 for L) plus a surplus of 50. Thus, DAH = 4300 and DAL = 2300. ii.) a uniform agglomeration payment (UA) which paid the same amount to both players based on an average across both players’ opportunity costs of adopting activity A. Again, this payment was made conditional on both players pursuing activity A. Specifically, UA was designed as the average of DAH and DAL, i.e. UA = 3300. iii.) a uniform individual payment (UI) which also paid the same amount to players undertaking activity A, but regardless of the land use decision of the other group member. The level of UI is also chosen equal to 3300.7 Note that UA and UI have equal level, but differ in their condition of allocation. While UA (like DA) is conditional on collective adoption of the sustainable land use, UI is only conditional on individual adoption of such land use. Implications of the implementation of these agri-environmental payments for the payoff matrix in case of rewetting are included in Figure 6. See Appendix B3 for a representation of Figure 6 for each treatment separately. We implemented each payment design treatment over a sample of 37 groups (i.e. 74 players). The baseline phase was implemented in all 111 groups (222 players). Our setup acknowledges the fact that uniform payments imply that players holding low opportunity costs may be excessively compensated for the adoption of a sustainable land use, while players with high opportunity costs would be insufficiently compensated. This could be counteracted by side

5 H and L farm profits are based on data from AGRIDEA database and are calculated as present value of infinite cumulative gross margin from vegetable production in Switzerland (with a discount rate of 4%). Thus, we present the “average” production potential of players for the future decades. For easier readability, numbers were rounded up and divided by 1000. Their unit is kCHF. 6 It is technically possible to drain single fields but it is not possible to rewet single fields without affecting other parcels. 7 The study did not test a differentiated individual payment treatment. In such a treatment, players would both have an incentive to choose the sustainable land use, regardless of the other’s choice. It would therefore be very easy for them to solve the game and reach the social optimum. The payment treatments selected leave more room for players to influence the game. It was necessary to select treatments that are relevant for the issue addressed and that enable to contribute to the associated literature. 47 payments between players. Therefore, our experimental design allowed side payments, which were implemented as follows. In stage 1, before voting on rewetting organic soils, each player could make a binding side-payment offer to his/her group member in return for the receiving player adopting activity A. The side payment was transferred independently of the land use choice of the player making the offer. Side payments made by player i (i=L, H) to the other player are represented by Si in Figure 6. Thus, ultimately in our experiment, players choose in stage 1 whether to be in the matrix in Figure 5 (status quo, no rewetting) or in the matrix in Figure 6 (rewetting). Only a unanimous vote for rewetting can move players to the matrix in Figure 6. If at least one player votes against rewetting, land use choice is by default activity B for both players. If both players vote for rewetting, land use choice is between activity A or B, and there may be side-payment transfers. We summarize the decision-making procedure for each round in Figure 7 below.

Player H

Activity B: vegetable farming 2500 Player L Activity B: vegetable farming 4500 Figure 5: Stage 2 payoffs if organic soils are not rewetted (‘Status quo’) Note: Payoffs of player H are indicated in bold. Both players pursue vegetable farming by default.

Player H

Activity A: sustainable use Activity B: vegetable farming 250 + UI + APL – SL + SH 250 + UI + SH Activity A: sustainable use 250 + UI + APH + SL - SH 4500 - 220 - SH Player L 2500 - 220 - SL 2500 – 220 Activity B: vegetable farming 250 + UI + SL 4500 – 220 Figure 6: Stage 2 payoffs if organic soils are rewetted Note: Si: side payment made by player i to other player. In the treatment phase: UI (uniform individual payment) = 3300; APi (agglomeration payment to player i). The agglomeration payment is either uniform: UA = 3300, or differentiated: DAi. with DAL= 2300 and DAH= 4300. In the baseline phase: UI=AP=0. Payoffs of player H are indicated in bold.

Proposal for side payment Vote on payoff matrix Land use decision (activity (S) (Figure 5 or Figure 6) A or B)

Figure 7: Experimental decision stages

2.2. Behavioral predictions

We develop behavioral hypotheses for payoff-maximizing players using backward induction (Serrano and Feldman, 2010, 2011). Social preferences can change behavior (Fehr et al., 2002; Fehr and Fischbacher, 2002). We illustrate this by computing equilibrium outcomes with social preferences for 48 the baseline scenario only, but experimentally analyze the effect of social preferences on behavior for all treatments. Complete derivations on our behavioral predictions are provided in Appendix B4. In general, the only outcome of the payoff matrix in Figure 6 (rewetting) that can potentially make both players better off than the status quo (no rewetting, Figure 5) is when both players adopt activity A. Therefore, rewetting can only be an equilibrium outcome if joint adoption of activity A is a Nash equilibrium in stage 2, Figure 6. i. Baseline: agri-environmental payment = 0 In the baseline and in the absence of side payments, the payoff matrix in Figure 6 has the form of a prisoner’s dilemma. That is, activity B is the dominating strategy for payoff-maximizing players, independent of player type. Side payments could in principle change the game structure, but this is possible only if joint payoffs are higher in the equilibrium where both rewet, compared to the payoffs in the status quo (Figure 5). Such ‘gains from trade’ could then be redistributed among players via side payments. Yet, such gains from trade do not exist in the baseline: joint payoffs are highest in the status quo. Thus, we expect payoff-maximizing players to not offer side payments, vote against rewetting, and maintain activity B. The same outcome is expected for players who aim to maximize joint payoffs. However, inequality-averse players would vote for rewetting because this allows for side payments that can equalize payoffs. While there are three strategies that can equalize payoffs, the profit-maximizing one among these is that H pursues activity B while L adopts activity A. This would be accompanied by a side payment from H to L with SH = 2015. Table 2 summarizes these predictions.8

Table 2: Baseline behavioral predictions

Players’ preferences Individualistic Joint profit maximizer Inequality averse 1st stage decision No rewetting; Si =0 No rewetting; Si =0 Rewetting; SH=2015, SL=0 2nd stage decision Activity B Activity B H: activity B; L: activity A H: exhaustion of peat Environmental Exhaustion of the peat Exhaustion of the peat L: preservation of peat effect Total peat remaining: 50%

ii. Differentiated agglomeration payment: APH = DAH = 4300 and APL = DAL = 2300 An agglomeration payment exceeding both players’ conservation costs transforms the land use decision into a coordination game. In our setup, if the peat area is rewetted, there are two possible equilibria in the payoff matrix in Figure 6 (both choose activity A or both choose activity B). Among these equilibria, joint adoption of activity A is not only the Pareto-optimal equilibrium but also yields higher payoffs to each player individually. These payoffs are also higher (by 50 per player) than the payoffs from not rewetting (Figure 5). Thus, there is no reason to make side payments, and we expect that payoff- maximizing players of both types vote for rewetting and then choose activity A. However, doing so involves a risk of coordination failure and yields payoffs that are only by 50 higher for each player than the certain payoff from not rewetting. Therefore, risk-averse players might prefer to not rewet rather than to risk coordination failure.

8 We represent the “extreme” behavioral cases. In the reality of the experiment, aversion to payoff inequalities translates into different (mixed) players’ behavior, e.g. 0 < SH < 2015. 49 iii. Uniform agglomeration payment: APH = APL = UA = 3300 The level of the uniform payment is based on an average of the group members’ opportunity costs from adopting activity A. It therefore excessively covers L’s opportunity costs (3300 > 2500 – 250) while it does not cover H’s opportunity costs from switching to activity A (3300 < 4500 – 250). Therefore, if no side payment is made by L, a payoff-maximizing H player has no incentive to adopt activity A. Moreover, if H chooses B, L is also better off choosing B. This is because the agglomeration payment is lost as soon as one player adopts activity B. Consequently, the game structure in Figure 6 without side payments has the form of a social dilemma. However, L’s payoff under activity A is higher than under activity B, and total payoffs are higher under joint adoption of activity A, also compared to the status quo payoffs in Figure 5. L would therefore be expected to make a side payment to H to incentivize H to adopt activity A. Such a side payment can turn the payoff structure in Figure 6 into a coordination game, analogous to the DA scenario. Moreover, agreeing on such a side payment can increase payoffs for each of the two players compared to the case when they vote against rewetting. As shown in the Appendix B4, this holds as long as the side payment satisfies 950 ≤ SL ≤ 1050. Thus, we expect that a profit- maximizing L player offers such a side payment and that payoff-maximizing players then both vote for rewetting and choose activity A.9 As for the DA case, there is a risk of coordination failure. However, the side payment may serve as a signal from the L player that (s)he aims at the equilibrium where both players choose activity A. This may help to reduce the risk of coordination failure perceived by the H player. Thus, comparing the DA and UA treatments, the UA treatment introduces an additional difficulty of requiring a side payment to achieve the Pareto-optimal equilibrium, but has a lower risk of coordination failure. iv. Uniform individual payment: UI = 3300 The level of the uniform individual payment is the same as for UA. However, the individual payment is made regardless of the other player’s land use choice. This changes the game structure significantly. If no side payments are made, activity A is the dominating strategy for L players, while activity B is the dominating strategy for H players, regardless of what the other player is doing. Thus, there is a unique Nash equilibrium under rewetting, where the low-cost type adopts the sustainable land use and the high- cost type reverts back to the conventional land use. However, while the L player prefers this outcome over the case of no rewetting, the H player is always better off without rewetting. This is because s(he) can thereby save the cost of the individual drainage system. As in UA, the L player can make a side payment to incentivize H to adopt activity A. The expected equilibrium outcome is the same as for UA. That is, with payoff-maximizing players, we expect a side payment from L to H satisfying 950 ≤ SL ≤ 1050. Both players would then vote for rewetting and adopt activity A. However, there is an important difference here to the UA case. Under UI, the side payment turns joint adoption of activity A into a unique equilibrium in Figure 6. Thus, the side payment under UI turns the game after rewetting into a non-dilemma, while in the UA treatment side payments induced a coordination game structure. The difference in game structure is due to the fact that the individual payment is paid out to players choosing A, regardless of the other player’s choice. Hence, under UI, there is no risk of coordination failure in the case of rewetting. We expect that this facilitates cooperation. We summarize our predictions for payoff-maximizing players in Table 3. In summary, we expect that payoff-maximizing players in the baseline scenario do not adopt the sustainable land use and peat is exhausted. For all three agri-environmental payment scenarios and with risk neutral players, the payoff-

9 The exact level of SL depends on players’ bargaining power. Assuming payoff maximizing players, L player would be expected to offer a side payment just above 950 (e.g. SL=951) and player H would have an incentive to accept it. However, such a small transfer implies a higher risk of coordination failure because H would be almost as well off under non-successful coordination. So, H may then not vote for rewetting. Thus, L might also offer a somewhat larger transfer to reduce the risk of coordination failure. 50 maximizing equilibrium outcome is full peat conservation. However, uniform payments require side payments from L to H, and agglomeration payments involve a risk of coordination failure. Moreover, social preferences can affect behavior, both with respect to side payments and land use.

Table 3: Expected results for payoff-maximizing players

Scenario Expected outcome Side payments required Risk of coordination failure Baseline No rewetting; activity B No No DA Rewetting; activity A No Yes UA Rewetting; activity A Yes (950 ≤ SL ≤ 1050) Yes, but mediated by side payment UI Rewetting; activity A Yes (950 ≤ SL ≤ 1050) No

3. Results

Results are presented in the following order. Considering the ten rounds of play, we first provide a policy performance evaluation, which compares all payment treatments along three important dimensions, namely environmental effectiveness, cost-effectiveness and equality with respect to the distribution of incomes. Second, we investigate whether behavior in our treatments is related to an array of individual preferences and other personal characteristics. Unless stated otherwise, average values reported below correspond to the average of the means of the given variable over all ten rounds.

3.1. Policy performance

3.1.1. Environmental effectiveness As our prime measure of environmental effectiveness, we consider individual stage 2 decisions, i.e. players who adopt the sustainable land use (activity A). Figure 8 illustrates our results graphically.10

100 90 80 70 60 50 40 30 20 10

Percentage of players who adopt activity A activity adoptwho players of Percentage 0 1 2 3 4 5 6 7 8 9 10 Rounds Baseline (222 players) UI - Uniform individual payment (74 players) DA - Differentiated agglomeration payment (74 players) UA - Uniform agglomeration payment (74 players) Figure 8: Percentage of players adopting activity A across treatments

10 We tested for significant behavioral differences across the underlying baselines of the three treatments. As expected, no significant difference was found, hence we merged the baseline phase for all conducted treatments. 51

We find that the average percentage of players who adopt activity A across all rounds is significantly higher in all payment treatments (50.8% in DA, 51.8% in UA, and 42.8% in UI) than in the baseline scenario (3.5%; proportion test; p-value = 0.00). There is no significant difference in the rate of adoption of activity A across the three payment treatments. Rates of adoption of activity A under all payment treatments are relatively low considering that adopting activity A is the payoff-maximizing strategy for both players in all treatments. This could be due to the difficulty to coordinate on side payments, the risk of coordination failure, or social preferences. Other experimental studies of group-based payments, which assume homogenous players, find higher rates of cooperation among players (e.g. Parkhurst and Shogren, 2008; Banerjee et al., 2014). Next, we examine adoption of activity A by player type. In the uniform payment treatments, L players are significantly more likely to adopt activity A than H players (test of proportion, p-values equal to 0.0003 and 0.0001). This is plausible because under a uniform payment L players have a considerable economic incentive to adopt activity A. Considering both treatments UA and UI together, 9.5% of the groups comply with our prediction: both players adopt activity A for net SL ≥ 950. Next, we analyze collective outcomes, i.e. the percentage of groups where both players adopt activity A. Figure 9 shows collective outcomes per treatment and round.

100 90 80 70 60 50 40

adopt activity A A activity adopt 30 20 10

Percentage of groups where both group members members group bothwhere groups of Percentage 0 1 2 3 4 5 6 7 8 9 10 Rounds Baseline (111 groups) UI- Uniform individual payment (37 groups) DA - Differentiated agglomeration payment (37 groups) UA - Uniform agglomeration payment (37 groups)

Figure 9: Percentage of groups adopting activity A across treatments

We find that, on average, uniform agglomeration and differentiated agglomeration payments yield collective cooperation rates (45.7% for DA and 50% for UA) that are similar to individual cooperation rates.11 However, uniform individual payments perform worst in terms of collective cooperation (23.2%). This is intuitive because agglomeration payments are designed to promote collective cooperation by being conditional on such cooperation, while individual payments by definition are allocated regardless of others’ land use choice. Under individual payments, we observe a large number of particularly H-type players who vote rewetting, but then continue conventional vegetable farming. Among the groups who rewet in UI, only 36.2% of the H players adopt activity A, compared to 98.7% of the L players. In comparison, on average 90% and 87.5% of the players adopt activity A in DA and

11 Test for significant difference in joint cooperation (proportion test): between UI and DA: p-value = 0.04; between UI and UA: p-value = 0.02 52 in UA, respectively, with no statistical difference between H and L players. Variability in players’ decision in stage 2 in DA and UA is therefore low. The result that, under UI, more H players revert to activity B after rewetting is also consistent with behavioral predictions in the presence of social preferences. Under UI, in the absence of side payments, there is a unique Nash equilibrium in stage 2 in which the H player reverts to activity B and the L player switches to activity A. Such an asymmetric equilibrium exists only with the individual payment. This equilibrium makes L significantly better off than without rewetting (3550 instead of 2500) while it makes H only slightly worse off (losing 220 in form of the cost of an individual drainage system). Thus, in the absence of sufficient side payments the asymmetric equilibrium appears as an option that can reduce payoff inequality at low cost to H. Moreover, the difference in payoffs between H and L in the asymmetric equilibrium is 730 (4280 – 3550). Thus, it is well possible that the asymmetric equilibrium performs best in terms of payoff equalization.

3.1.2. Cost-effectiveness Next, we examine how the different payment designs differ in terms of cost effectiveness. We define cost effectiveness here as the amount of money units spent per peat unit preserved. We calculate it by dividing the total amount of agri-environmental payments actually made over all 10 rounds by the total amount of peat units for which activity A was adopted (i.e. the total number of times either of the players adopt activity A along the 10 rounds). The results are presented in Table 4. We find that differentiated payments perform best (2957 money units per peat unit preserved), while individual uniform payments turn out to be most costly for conservation (3310 money units). While both agglomeration payments present a very similar level of peat preserved, the differentiated agglomeration payment appears more cost effective than the uniform agglomeration payment (2957 vs. 3204 money units per peat unit preserved). This relates to the fact that in DA, the rate of groups where both players adopt activity A is lower than in UA, which means a lower rate of agri-environmental payments actually paid out. In DA, a set of players revert to activity B in stage 2. The two uniform payment treatments perform similarly in terms of cost effectiveness. It seems that payment differentiation has the strongest positive effect on cost effectiveness.

Table 4: Cost-effectiveness across treatments Baseline DA UA UI Total peat units preserved across 10 rounds 75 375 379 316 Percentage of total peat preserved 3.4 50.7 51.2 42.7 Total payments made across 10 rounds (in money units) - 1’108’800 1’214’400 1’046’100 Cost-effectiveness (money units per peat unit preserved) - 2956.8 3204.2 3’310.4

3.1.3. Equality of income distribution Next, we aim at understanding the impact of our payment treatments on income (in)equality. For this, we compare the average level of inequality in payoffs among players using the Gini coefficient. We compute the Gini coefficient for each round based on players’ payoffs. We then compute the average Gini coefficient across the ten rounds. The coefficient varies from 0 to 1, with 0 representing perfect equality between payoffs, and 1 perfect inequality. In the baseline scenario, the average Gini coefficient is equal to 0.16. We find that differentiated payments DA lead to statistically greater inequality than

53 uniform payments (average Gini coefficient for DA: 0.17, for UA: 0.14; for UI: 0.11).12 One reason for higher income inequality in the differentiated payment treatment may be linked to the fact that we only observe few side payments in this treatment. As the differentiated payment is designed to just cover opportunity costs of conservation, initial inequalities in player’s income appear to be more accepted by players. In order to further understand our results related to equality, we take a closer look at outcomes with respect to side payments. Figure 10 shows average net side payment offers from player L to player H. A negative net transfer indicates that the larger offer was made by H. We also report in Figure 10 actually transferred net side payments across all rounds and treatments. Net offers include all offers that were made by players in stage 1 independently of land use decisions in stage 2. Net transfers refer to the offers that were actually executed based on the receiving player adopting activity A in stage 2.

Baseline DA UA UI 600 400 200 0 -200 -400 -600 -800 -1000 Mean net offers Mean net transfers

Figure 10: Mean net offers and mean net transferred side payments Note: Net side payments from L to H player. Negative values indicate that the larger offer or transfer was made by H

In the baseline scenario and under the differentiated payment treatment, side payments are mainly offered from H to L players (negative mean net offer across the ten rounds equal to -77.8 in the baseline scenario and to -110.5 in DA).13 These offers cannot be explained by payoff-maximizing behavior. Rather they suggest that at least some players have inequality-averse preferences (see predictions in section 2.2). Under the uniform payment schemes (UA and UI), side payments are offered from L to H players. This is consistent with the predicted behavior of payoff-maximizing players. However, the level of the side payments offered is considerably lower than expected under payoff-maximization (427.1 for UA and 182.8 for UI, compared to a predicted level of at least 950). This again suggests that inequality aversion is affecting players’ behavior. Side payment offers are on average significantly larger in the uniform agglomeration payment treatment (UA) than in the uniform individual payment treatment (UI) (two sample t test, 95% confidence level). This difference in payoff redistribution may be explained by the fact that the agglomeration payment (UA) involves a risk of coordination failure leading to the loss of agri-environmental payments. Hence, in order to increase the likelihood that H votes for rewetting and does not revert to activity B in stage 2, L may pay a risk premium in form of a higher side payment offer. The results for actually executed

12 Test of significant difference between the mean Gini coefficients of DA and UA: p-value = 0.00; between DA and UI: p- value = 0.00; between UA and UI: p-value =0.00 (t-test). 13 All reported average mean net transfer offers and mean net transfers are significantly different from zero at 95% confidence level based on the mean comparison test. 54 transfers exhibit the same pattern. Actual transfers are also negative for baseline and DA and higher for UA than for UI in the baseline scenario, only 6.3% of all net side-payment offers made among group members are actually executed. Finally, we consider the effect of side payments on the groups for which both members adopt activity A. In the UA and UI treatments, a net transfer takes place in respectively 79.5% and 80.2% of these groups (average net SL equal to 505.3 and 184.0). This confirms that joint adoption of activity A under uniform payments is strongly driven by side payments. In DA, a net transfer takes place in about 77.5% of the groups who jointly adopt activity A (average net SL transfer equal to 139.1), again suggesting that inequality aversion appears to affect behavior. In UI, a net side-payment offer is made in 81.4% of the groups for which L chooses activity A and H chooses activity B (average net SL offer equal to 99.5). The executed offers (from H to L players) correspond to 41% of them.

3.2. Effect of players’ social preferences on behavior

In this section, we address the impact of players’ social preferences on behavior in the game. At the beginning of each experimental session, we elicited social preferences via the SVO slider measure (Murphy et al., 2011). The analysis of the pattern of individual choices across six slider items (dictator- like distribution decisions) enables to compute a score represented as an angle. The theoretically possible angle ranges between -16.26° and 61.39°. This angle represents the extent to which the player gives importance to his/her own income in relation to the outcome of other players (other participants of the experiment). Thus, the angle provides a measure of prosociality with a higher angle indicating stronger pro-social orientation. An angle close to 0° refers to an individualistic player, i.e. one who gives primary importance to his/her own outcome (Murphy et al., 2011). Based on the angle derived for each player, players are categorized as either prosocial (angle ranging between 22.45° and 57.15°) or individualist (angle between -12.04° and 22.45°) (ibid). We observe no significant differences in measured social preferences between our treatment cohorts. The second set of nine slider items enables to identify, among prosocial players, joint profit maximizers and inequality averse players (see Appendix B5 for the distribution of social-preference types across treatments).

3.2.1. Players’ decisions We analyze the effect of social preferences on players’ decisions. Given the low variability of players’ decision in stage 2 in DA and UA, we primarily study stage 1-decisions for this analysis. In the absence of agri-environmental payment, none of the group members has an incentive to rewet the soils if they are payoff maximizers. However, across all rounds, on average 40% of the players vote for rewetting organic soils in the baseline scenario. We find that H players characterized as “prosocial” exhibit a significantly higher rate of vote in favor of rewetting organic soils than the ones characterized as “proself” (mean rate across all rounds respectively equal to 44.9% and 20.6%; two sample t test, p = 0.00). We also find that in both uniform treatments, social preferences significantly correlate with players’ decision for rewetting. In UA and UI, prosocial H players exhibit a higher rate of vote in favor of rewetting as compared to proself H (65% against 45% in UA; 78.8% against 36.2% in UI; p = 0.00 for both treatments). In DA, we find no significant behavioral differences between the different social preference types. This is reasonable because differentiated payments make rewetting the preferred option for both players, while uniform payments – in the absence of sufficient side payments - maintain ‘no rewetting’ as preferred option for H types.

3.2.2. Side payments Next, we investigate the effect of social preferences on side payments. In the baseline scenario, we find that the largest side-payment offers are made by H players characterized as “prosocial”.14 In DA, prosocial H players offer significantly higher side payments than proself H players. This result is intuitive: reducing inequality between players requires a transfer from H (the player with the more favorable production potential) to the L player. This confirms that side payments in the baseline scenario and the DA treatment appear to be motivated by prosocial preferences. In both uniform payment treatments, proself L players make on average significantly higher offers than prosocial L players: the average net SL for groups with proself vs. prosocial L players is as follows: UA: 463.9 (std. 161.1) vs. 379.4 (std. 98.4); two sample t test, p = 0.06; UI: 318.4 (std. 302.7) vs. 61.7 (std. 70.9), p = 0.001). This result is intuitive as well. Proself players wish to maximize their own payoffs. If they assume that the other player also maximizes own payoffs, L needs to make a side payment to induce H to vote for rewetting. This results in unequal payoffs, which can be partly counteracted by lowering the side payment offer. If L players are inequality averse and think that H players are as well, they would thus make lower offers and hope for H to agree to them.

3.2.3. Effect of individual players’ characteristics on behavior We now analyze the effect of players’ risk and time preferences, gender and general attitude and opinions on behavior in the experiment. Players characteristics were collected from a short exit-survey conducted at the end of the experimental sessions. Summary statistics of the variables considered in the analysis are presented in Table 5. Further statistics on additional players’ characteristics are provided in Appendix B6.

Table 5: Summary statistics of players’ characteristics across treatments Mean (St. dev) Variables Description UI DA UA Risk 0 (avoid risk) to 10 (willing to take risk) 4.6 (2.4) 4.6 (2.3) 5.0 (2.3) Patience 0 (very impatient) to 10 (very patient) 5.4 (2.5) 5.8 (2.3) 5.8 (2.6) 1 (don’t care at all about others’ opinion) to 4 Care for reputation 1.5 (0.7) 1.5 (0.7) 1.4 (0.8) (care a lot) Index of 11 items relating to the frequency of 60.4 63.1 62.8 Care for the environment taking environmentally friendly actions15 (14.1) (11.8) (11.7) Knowledge peat 0 (no knowledge about degradation of organic 2.3 (0.8) 2.3 (0.9) 2.4 (0.7) degradation soils in CH) to 3 (great deal) 0 (don’t think that cooperative approaches in Belief in cooperative agricultural management are more successful 1.1 (0.7) 1.0 (0.8) 0.8 (0.7) approaches than other approaches) to 3 (think that they are more successful) Gender (% of females) 0 = Male, 1 = Female 63.5 54 46 Note: The average values do not vary significantly across treatments (t-test, 95% confidence level)

Table 6 below presents results from an econometric analysis of the determinants of individual adoption of activity A. In the baseline scenario without agri-environmental payments (Model 1), we do not find

14 The mean net SL in groups with prosocial H players across the ten rounds is equal to -89.0 (std. 128.2). In groups with proself H players, it is equal to 33.8 (std. 63.6) but not significantly different from 0 (t-test at 95% confidence level). 15 Each item of this index could be answered by 1 (never), 2 (not very often), 3 (quite often), 4 (very often), 5 (always), or 6 (not applicable). We computed the score of the player by calculating a percentage out of the maximum number of points he/she could make (i.e. the highest level of care for the environment) excluding the “not applicable” answers. This enabled a common basis of comparison for all players. 56 any reliable predictor of adopting activity A. It seems that the incentive to remain in the status quo is so strong when there are no agri-environmental payments, that this incentive dominates all other factors.

Table 6: Panel random-effect logistic regression on players’ land use choice (1) (5) (7) (3) VARIABLES Baseline DA UA UI Round 1.091 0.878 0.953 1.013 (0.0877) (0.0885) (0.0860) (0.0734) Experiment H player 0.0381 0.0455 7.38e-06** 0.00208 (0.105) (0.216) (3.70e-05) (0.00800) SVO angle 1.027 0.986 0.989 0.997 (0.0197) (0.0532) (0.0444) (0.0313) Risk 0.975 0.602 0.742 0.919 (0.0937) (0.206) (0.159) (0.166) Patience 1.016 0.455** 0.873 0.799 (0.0995) (0.166) (0.174) (0.142) Opinions and Care for reputation 0.707 0.949 1.040 0.641 preferences (0.178) (0.559) (0.500) (0.257) Care for the environment 0.969 1.020 0.964 0.947 (0.0215) (0.0537) (0.0354) (0.0389) Belief in cooperative approaches 1.156 1.675 0.818 2.819** (0.298) (1.137) (0.515) (1.207) Knowledge peat degradation 1.160 0.576 0.617 0.752 (0.338) (0.291) (0.331) (0.336) Gender (Female = 1) 0.690 0.115* 0.306 0.439 Gender (0.337) (0.136) (0.326) (0.457) SVO angle*H 1.001 1.047 1.116* 1.015 (0.0262) (0.0827) (0.0688) (0.0453) Risk*H 1.130 0.718 1.758* 0.767 (0.161) (0.267) (0.574) (0.184) Patience*H 0.916 0.987 1.356 1.445* (0.122) (0.483) (0.375) (0.319) Care for the environment*H 1.055 1.013 1.094 1.031 Interaction terms (0.0369) (0.0826) (0.0619) (0.0498) Belief in cooperative approaches*H 0.723 3.403 0.347 1.373 (0.290) (4.637) (0.309) (0.898) Knowledge peat degradation*H 0.924 0.683 3.376 0.994 (0.513) (0.549) (3.165) (0.586) Gender*H 0.738 2.161 1.831 0.110 (0.550) (3.384) (2.453) (0.175) Constant 0.0593* 1,287 743.0** 83.41 (0.0894) (5,650) (2,357) (225.1) Model Number of observations 2,200 730 730 740 Number of groups 220 73 73 74 Note: Dependent variable = 1 if player adopted activity A, = 0 otherwise. Reported values represent odds ratios. Standard errors are presented in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered at the group level. To control for group membership, all regression models also include group dummies (not reported). Overall significance of the regression is high (p-values < 0.05).

In the differentiated payment treatment (Model 2), we find that patient players are less likely to adopt activity A. One possible explanation for this outcome is that patient players might first wait and bargain about side payments or to better build an expectation on the other player’s preference type by observing her side payment offers. Impatient players may value these aspects less strongly compared to the possibility of immediately gaining from successful coordination. Furthermore, female players are less likely to adopt activity A than male players. In the uniform agglomeration payment (Model 3), H players are significantly less likely to adopt activity A compared to L players. This is consistent with the fact that H players have no incentive to adopt activity A in the absence of sufficient side payments from their group member. Moreover, we find that 57 the effect of risk is significantly stronger for H than for L players. This is intuitive as in the presence of this type of payment scheme (agglomeration payment), players incur the risk of not receiving the payment in case their partners revert to activity B on rewetted soils. Therefore, risk-averse players may prioritize the status quo situation over a situation with a risk of coordination failure. Furthermore, H players have proportionally less to gain from cooperating than L players. Finally, we find that the effect of prosociality is higher for H than for L players. As H players have initially no incentive to adopt activity A, it appears intuitive that prosocial preferences are more important for their decision to adopt activity A, as compared to L players. This is even more the case when considering that the side payment offers by L are on average below the level that would turn activity A into the payoff-maximizing strategy for the H player. In the uniform individual payment treatment (Model 4), a positive opinion toward cooperative approaches in agricultural management is observed to be a predictor to players adopting activity A. This is intuitive, because in UI, H types have initially no incentive to adopt activity A and L types tend to make side payments lower than what would induce activity A to become payoff-maximizing for the H type. So other motivations, such as a belief in cooperative approaches may be needed to induce H players to choose activity A. In the DA and UA treatment, the payment design already incorporates the idea of a cooperative approach. In UI, it does not, so the variable ‘opinion about cooperative approaches’ may measure a more general preference effect here. Finally, patience has a stronger impact on H players’ adoption of activity A than on L players. This is intuitive: in the absence of side payments H players do not have an incentive to cooperate under the uniform payment. Thus, the more patient an H player is, the more (s)he may tend to wait for an appropriate side payment offer from their partner player. In fact, by rejecting early side payment offers, H players can try to get L players to raise their offers.

4. Conclusions

We performed a framed experiment in order to provide insights into the performance of alternative designs of payments for ecosystem services to increase peat conservation. Agglomeration payments were analyzed as a promising policy option to promote coordination among farmers, and compared to a more conventional individual payment. We tackle a particular challenge in designing payments for environmental services by explicitly including heterogeneity in the opportunity costs for conservation. At present, payment programs often include fixed uniform payments. Differentiated payments have been argued to be more cost-effective, but have to our knowledge not been implemented or tested in the context of agglomeration payments. Our primary focus was to compare environmental and cost-effectiveness as well as equity outcomes of alternative payment designs when landowners have different opportunity costs of soil conservation and cooperation among farmers is required for adoption of the sustainable land use. We find that all payment schemes are equally effective in incentivizing the adoption of a sustainable use of organic soils. Yet, we found that the individual payment facilitates cooperation in rewetting (first stage decision) in combination with some reverting to the unsustainable land use by the high-cost farmers, while the agglomeration payments rather foster coordination in the adoption of the sustainable land use (second stage decision). Although uniform payment schemes only provide an incentive to players with low opportunity costs to enroll in a conservation program, players manage to agree on side payments that redistribute payoffs and induce cooperation by high-cost players. Moreover, with respect to equity in payoffs, we find that a differentiated (agglomeration) payment leads to greater inequality than uniform payments. However, regarding cost effectiveness, the differentiated agglomeration payment performs best. This result 58 confirms PES cased studies that highlight that differentiated payments are more cost-effective than other payment forms. Uniform individual payments are most common in agri-environmental policy. They require less information on the distribution of opportunity costs across farmers and are often thought to involve lower administrative costs. In this sense, it is somewhat comforting that uniform payments are found to perform equally well on environmental effectiveness and better on income equality. However, our study suggests that uniform payments are not only less cost effective. They may also may generate transaction costs by requiring the establishment of contracts between farmers for payoff redistribution. Furthermore, we examined the effect of a range of individual characteristics on players’ decisions. We identified that impatience and belief in cooperative approaches were potential positive predictors to an agreement among group members in rewetting the peat area. However, it is important to note that side payments are an important determinant of players’ decisions and their magnitude is also indirectly affected by players’ personal characteristics. The reduced form regression analysis does not distinguish the direct effect of a variable on land use from the indirect effect via side payments. Furthermore, we find that individual social preferences are an important determinant of the magnitude of side payments. It should be noted that in our setting individual land use change requires collective action because the alternative activity is only feasible with a collective vote for rewetting. In other settings, commonly analyzed in studies on agglomeration payments, individual land use change is possible independently of other land users’ action. Thus, the difference between individual and agglomeration payments is somewhat less pronounced in our setting. Further research should analyze the performance of agglomeration vs. individual payments with heterogeneous farmers in settings where individual land use change is always possible. While our study focuses on the net present value of farm profits, the dynamic evolvement of farm profits as soils get degraded may also affect behavior in a more complex manner. We address this issue in chapter 3.

CHAPTER 3 Can agglomeration payments induce sustainable management of organic soils in Switzerland? – A computerized framed experiment1

1. Introduction

In Europe, organic soils are historically exploited in various manners: extracted for fuel and growing media in the horticultural sector, or used as a support for crops and livestock grazing. These activities require the draining of the soil, which leads to the degradation of the peat (upper layer of organic soils) that is oxidized and disappears (Xintu, 2009). Europe is the world’s second largest hotspot of greenhouse gas (GHG) emissions from the degradation of organic soils (Couwenberg et al., 2011; Joosten, 2009). In Switzerland, organic soils represent less than 2% of total agricultural land (Wüst, 2015) but their preservation would significantly contribute to national goals of GHG emission reductions. Because of agricultural production activities that include intensive drainage, part of these soils are severely affected and are at risk of disappearance in the immediate future. An example for such a situation is the case of a western region of Switzerland called “Seeland”. This region is characterized by large areas of organic soils that have been historically managed for intensive and profitable vegetable production. In addition to the loss of important ecosystem services, peat degradation leads to uncertainties about the future of intensive agricultural production on these soils. Yet, these soils are not the object of specific management regulations. Rewetting by restricting drainage is the most effective way to protect the remaining peat and to reduce GHG emissions (Lunt et al., 2010; Graves and Morris, 2013; Joosten and Couwenberg, 2009). However, such rewetting is in conflict with current land use. Two core aspects are identified as crucial in enabling sustainable management practices on organic soils. First, rewetting by raising the water table on these soils can only occur if all farmers who depend on the same drainage system agree. On average 10 to 40 farms use one pumping station. Thus, cooperation between farmers is necessary. Second, due to differences in past management practices and the conditions under which the peat was formed, organic soils are highly variable with respect to both the thickness of the peat layer and the future suitability of the underlying mineral soil layer to sustain current farming activities once the peat layer is lost. As a consequence, farmers vary in their vulnerability to peat degradation and in their future farm profits from vegetable farming on these soils. Therefore, opportunity costs for switching to sustainable use of organic soils and thus the incentives to do so differ among farmers. Motivated by this concrete issue and using the “Seeland” as the study region, we aimed at identifying policy instruments that are effective in promoting the sustainable management of organic soils in order to prevent their on-going degradation.

There are many instances where the effective provision of ecosystem services requires cooperation among several land users. Another example is the implementation of tree corridors in the agricultural landscape to increase the population of particular types of fauna. Agglomeration payments have been proposed by economists as an approach for promoting cooperation among farmers (Drechsler et al., 2010). Economic experiments (e.g. Banerjee et al., 2011, 2015; Parkhurst and Schrogen, 2007; Bamière et al., 2013) and models (e.g. Bell et al., 2016; Drechsler et al., 2016) have demonstrated their potential. They are therefore potentially promising instruments in fostering sustainable use of organic soils. Building on the concept of an “agglomeration bonus” defined by Parkhurst et al. (2002), the concept of agglomeration payment was introduced by Drechsler et al. (2010) in the context of habitat pattern creation across farmland. These payments are used as an incentive for the spatial coordination of conservation areas. An agglomeration payment is based on the joint activities of multiple farmers and is only paid out if farmers commonly undertake a similar activity, while an agglomeration bonus includes a base participation component and a bonus component - the amount of which depends on the number of farmers undertaking the joint activity. In Switzerland for instance, as part of the agri-environmental scheme on "ecological compensation areas", the ecological network bonus offers an additional payment to farmers if their land belongs to a contiguous habitat network. Existing empirical studies on agglomeration payments or bonuses assume homogenous land users (e.g. Banerjee et al., 2012, 2015). In practice, however, the opportunity costs of adopting a more sustainable land use usually differ among land users. Moreover, unsustainable land use can affect land users to different degrees in their own productivity, implying that opportunity costs may change over time and this varies across farmers. The case of organic soils in Seeland is an example for this: initial opportunity costs are high but when the peat gets degraded, the farmers with lower quality underneath soils face lower productivity and thus lower opportunity costs over time. These heterogeneities in opportunity costs raise additional complexities and questions for the design of effective policies, which we address in this chapter: Should the agglomeration payment be homogeneous or aligned to farmers’ opportunity costs over time? How does land user heterogeneity affect their strategic behavior when faced with a potential agglomeration payment? Different designs imply different incentives and may lead to different behavioral patterns. We compare the effects of two payment schemes: a constant agglomeration payment that pays the same amount to each farmer and a variable agglomeration payment that mirrors differences in farmers’ opportunity cost over time. We capture the key features that characterize the case of intensively used organic soils in a framed and dynamic economic experiment. The experiment represents the decision situation of farmers on these soils. It allows for heterogeneous farmers who differ with respect to the dynamics of how peat degradation affects their future production potential in conventional land use. We analyze the resulting dynamics in the adoption of conventional vegetable production versus more sustainable land use, both with and without an agglomeration payment, and the results depending on payment design. We also analyze how farmers’ social preferences affect behavior and three policy outcomes: environmental effectiveness, cost-effectiveness and income inequality. We conducted our web-based experiment with students of schools of applied agriculture in Switzerland who were to a large extent highly involved in farming activities. This chapter contributes to the literature in two main ways. First, while previous research on organic soils has concentrated on its degradation aspects and on restoration strategies, we address the management of organic soils from an agricultural and economic perspective. We test whether agglomeration payments could resolve this complex resource problem at hand. Second, in reference to the need for “real-world experience with agglomeration payments” (Parkhurst and Shogren, 2007), we contribute by testing agglomeration payments in the context of an innovative highly contextualized

62 experiment, involving participants from the field, and including the option of a variable payment scheme design. We find that both agglomeration payment schemes are effective in promoting more sustainable practices on organic soils. However, the constant payment appears to be more effective than the variable payment in promoting sustainable management. One of the reasons for the better performance of constant payment is that the majority of the players who adopt sustainable land use do so already early on in the experiment, which contributes to the preservation of about half of the peat by the end of the experiment. Another reason may be that, considering the ten time periods, total joint payments are higher under the constant than under the variable payment. We therefore also analyze cost effectiveness, i.e., environmental effectiveness per unit of money spent on payments. We find that the constant scheme is more cost effective than the variable scheme. Moreover, it leads to lower inequality in incomes. Social preferences, willingness to take risks, opinion of cooperative approaches, and personal reputation appear to influence decisions, and therefore policy outcomes. In the subsequent sections, we first describe our experimental design and then present the results. We end with a discussion and conclusion section.

2. Experimental design and expected effects

Based on information collected from interviews with experts of the Seeland region, we developed an experiment in the form of an interactive, computer-game-like representation of the management decision of farmers on organic soils. This tool builds on the highly visualized “framed lab-in-the-field experiment” approach used by Reutemann et al. (2015). The core aspects of this experimental concept resides in the framing of the experiment with the actual context study including the representation of its actual economic data and the time dependence of the decisions. Contrary to Reutemann et al. (ibid), our experiment includes interactions between players. The experiment was conducted with subjects characterized by a strong agricultural background, namely agricultural students from regional agricultural apprenticeship schools in Switzerland, of whom a majority intend to become farmers. In total, we recruited 254 students for the experiment and randomized their assignment to our treatments. The experimental session is composed of three parts. The first part is an incentivized Social Value Orientation test (SVO) used to assess players’ social preferences (Murphy et al., 2011) (see Appendix C1). Using a slider measure, it allows classifying subjects on the basis of a single metric, the so-called SVO angle that ranges between -16.26° and 61.39°, with a higher value reflecting stronger social preferences. This test also enables to distinguish three categories of players: ‘profit maximizers’ that prioritize their own payoffs, ‘inequality-averse players’ that focus on minimizing the difference between their own and the other person’s payoff, and ‘joint maximizers’ that focus on maximizing joint payoffs. The second part of the experiment is the baseline phase (no policy intervention) and the third part is the treatment phase where either the case of no policy intervention is continued or one of the two agglomeration payments is introduced. A socio-economic survey ends the session. We describe here the setup of the experiment and decision situation faced by players. Appendix C2 describes the model underlying the experimental design in detail. Experimental instructions are provided in Appendix C3.2

2 Each phase of the experiment was incentivized (SVO test, baseline phase, treatment phase). For the baseline and the treatment phase, one time period was randomly picked in each one of them at the end of the experiment. The earning in points from these two time periods were exchanged into francs (CHF) at a rate of 10 points = 0.25 CHF. 63

Players are asked to take the role of vegetable producers and are placed anonymously into groups of two, which represent a community depending on a joint drainage system. The baseline and the treatment phases each consist of ten time-periods. Group members were rematched between these two phases. In each time period, players face a trade-off between pursuing the conventional activity (vegetable production) and adopting a sustainable use of organic soils that requires rewetting the soils and enables the preservation of the peat. To reflect the asymmetry among farmers, each group consists of two player types who differ in the type of underlying mineral layer in their organic soils.3 H player has high quality mineral soils suitable for vegetable production, and L player has low quality mineral soils not suitable for vegetable production. Because of different soil productive potentials, H and L have thus different farm profit functions (as shown in Figure 11). Peat degradation does not affect H’s farm profit. Regardless of the number of time periods in which H produces vegetables, his/her farm profit remains constant for this activity. This is because once the peat layer disappears, H can simply continue farming vegetables on the underlying mineral soil. By contrast, peat degradation negatively affects L’s farm profit. The effect depends on the number of time periods previously used for farming vegetables. The closer L gets to the underlying mineral soil layer, the lower his/her farm profit. The sustainable land use provides a low, but constant profit to both farmers. The difference in profit functions translates into heterogeneity in opportunity costs over time. Because the H player possesses long-term production potential in conventional vegetable farming, (s)he faces constant high opportunity costs for adopting the sustainable use. Because the L player possesses short-term production potential in vegetable farming, his/her opportunity costs for adopting the sustainable use diminishes over time as a function of cumulative past vegetable production. Below we refer to the ‘sustainable use’ as activity A, and to ‘vegetable production’ as activity B. Activity A can only be conducted if organic soils are rewetted.

Figure 11: Profit functions of H and L players over time

The decision process can be represented by two stages, which each correspond to a payoff matrix.4 The first stage is the collective decision to rewet organic soils. Group members can first anonymously communicate through chat messages in order to coordinate strategies. Then, each member has to cast a binding vote in favor of or against rewetting soils on both of their farms. If both players vote for

3 We choose a 2 and not a multi-player design. Having for instance four players per group, e.g. two H types and 2 L types, would be more representative of the reality but would highly complicate the design of the experiment. Indeed, the two L and the two H may not have the same social preferences among each other. This would result in noise in the interpretation of the results and in difficulties in disentangling the effects. 4 In the payoff matrices, farm profits from vegetable farming or from the sustainable use are calculated based on actual farm profit data (from AGRIDEA database and University of Greifswald) as the present value of cumulated future farm gross margin over 5 years (with a discount rate of 4%). Numbers were rounded up. The unit of these numbers is kCHF. 64 rewetting, rewetting takes place. If at least one of the group members rejects rewetting, organic soils remain drained. The second stage is the individual land use decision. If rewetting was rejected, the only option for both player types is to remain in activity B, as sustainable use is not feasible on drained soils. In this case, payoffs are as in Figure 12 below. If rewetting was accepted, each player can choose between activities A and B. However, conducting activity B on rewetted soils requires installing a costly personal drainage system. Payoffs are as in Figure 13. The payoff matrices change as the peat degrades. 푖 In this chapter, we define 휋푛,푡 as the farm profit of player i at time t, given a number n of previous time periods in activity B. As explained above, profits from activity B are constant and independent of n for i=H: πH = 800 throughout the experiment. For i=L profit from activity B declines in n due to peat degradation. Table 7 shows the exact parameterization used in the experiment, which was known by players.

H player Activity B: vegetable production 휋퐿 L player Activity B: vegetable production 푛,푡 훑퐇 Figure 12: Players’ payoffs at time t if organic soils are not rewetted. Note: The payoffs of H players are indicated in bold.

H player Activity A: Sustainable use of organic soils Activity B: vegetable production 퐿 L H H Activity A: Sustainable 40 + 퐴푃푛,푡 – S + S 40 + S use of organic soils 40 + 퐀퐏퐇 + SL - SH 훑퐇 - 25 - SH L player 퐿 L 퐿 Activity B: vegetable 휋푛,푡 - 25 - S 휋푛,푡– 25 production 40 + SL 훑퐇 – 25 Figure 13: Players’ payoffs at time t if organic soils are rewetted.

L H i Note: APn,t and AP = agglomeration payment to L and H players respectively. S : side-payment offer made by player i. The payoffs of H players are indicated in bold. Cost of personal drainage system is 25. Constant profit from activity A is 40.

Table 7: Farm profits of activity B n 0 1 2 3 4 5 6 7 8 9 퐋 훑퐧,퐭: farm profit of L for activity B 800 800 800 550 160 0 0 0 0 0 훑퐇: farm profit of H for activity B 800 800 800 800 800 800 800 800 800 800

We test three treatments: (i) no policy intervention (i.e. repetition of the baseline phase), (ii) constant agglomeration payment, and (iii) variable agglomeration payment. By definition, an agglomeration payment is only made if both group members adopt activity A. The level of the payment is chosen to be slightly higher (2% or 10 money units) than the opportunity costs of the respective players to induce cooperation among players (Ferraro, 2008; Van Soest et al., 2013). In the constant agglomeration payment (APC) treatment, the payment is based on opportunity costs in the first time period, which are equal for both players, and remains constant over time. Specifically, APL = APH = 770. In the variable agglomeration payment (APV) treatment, the payment follows players’ evolvement of opportunity costs H over time. Specifically, APv is constant for H (AP = 770), while for L the payment is a function of n.

L When L’s farm profit decreases due to soil degradation from activity B, the payment (APn,t) that (s)he could receive in exchange of adopting activity A decreases by the same absolute amount (see Table 8 below). This means for n > 2, the potential payments for L and H differ.

Table 8: Level of the payment in APV n 0 1 2 3 4 5 6 7 8 9 Payment of L for activity A 770 770 770 520 130 10 10 10 10 10 Payment of H for activity A 770 770 770 770 770 770 770 770 770 770

Our experimental setup acknowledges the increasing asymmetry among players’ incentives for activity A over time, as they continue with activity B. This asymmetry may lead players to bargain their cooperation with their group member and redistribute the gains. We incorporate this possibility by allowing for side payments (Si) between players i in the same group. Such side payments may also be used by players to secure cooperation or to signal a willingness to cooperate. They are implemented as follows. Before voting, each group member can offer a binding side payment to the other member. The payment is only transferred if and only if both members vote to rewet organic soils and on the condition that the potential beneficiary adopts activity A, regardless of the choice of the player making the offer (see Figure 13). Note that under the ‘no policy’ treatment, payoffs for both players are highest without rewetting as long as n < 5. Thus for the first periods, payoff-maximizing players have no incentive to vote for rewetting. For n ≥ 5, the L player could earn higher payoffs in activity A, but this requires rewetting and thus a side payment to the H player. With the agglomeration payment, both players are always better off under activity A because the payment covers more than the opportunity costs. Yet, because L has more to lose than H if rewetting does not take place and rewetting is conditional on a unanimous vote, there may be bargaining between group members over side payments. With the constant agglomeration payment, the bargaining would be over sharing the surplus, which is the difference between the payment and the actual opportunity cost of the player. This surplus is increasing for L player with number of periods in which L conducts activity B. With the variable agglomeration payment, the potential for bargaining and strategic behavior of the H player is even stronger. H knows that L’s potential agglomeration payment also declines as activity B continues and may try to exploit this to obtain a higher side payment. Thus, in summary, we expect that players remain in activity B in the no-policy scenario, but switch to activity A in the two agglomeration payment treatments. However, bargaining may delay adoption of activity A, and we expect this effect to be stronger under the variable payment. Social preferences may also change results because side payments, voting and land use choices could be employed to reduce payoff inequalities or increase joint payoffs.

3. Results

We explore the experimental results as follows. We first examine the environmental effectiveness of the payment treatments. We also analyze the net side payments and the communication content between players to better understand decision-making. We then study the distributional impact and the cost effectiveness of the policy treatments. Finally, we analyze the effect of social preferences and other personal characteristics on outcomes. Unless stated otherwise, in this section average values reported correspond to the average of the means of the given variable over all rounds.

3.1. Environmental effectiveness

A proxy for the environmental effectiveness of our policy treatments is the rate of players who adopt activity A, and at what time. The more rounds players adopt activity A, the more peat is preserved. Players who for instance only switch activity as from n = 5 do not enable preserving the peat on their land. Figure 14 presents this proxy across the three treatments. This analysis is provided in Appendix C4 for H and L players separately. We only consider the outcomes from the baseline treatment and not the outcomes of the preceding baseline phase. We first analyze the effect of the tested policy options on peat preservation as compared to the situation with no policy intervention. Second, we identify the payment scheme that enables the highest level of peat preservation.

100 90 80 70 60 50 40 30 20

Percentage of players who adopt activity A A activity adoptwho players of Percentage 10 0 1 2 3 4 5 6 7 8 9 10 Time-periods Baseline (78 players) Constant agglomeration payment (88 players) Variable agglomeration payment (88 players) Figure 14: Percentage of players who adopt activity A across treatments

Across all time periods, on average 65.6 % (std. 9.2) in the constant agglomeration payment treatment

(APC) and 40.2 % (std. 7.3) in the variable agglomeration payment treatment (APV) adopt activity A as compared to 18.2 % (std. 8.5) in the baseline scenario. In the baseline scenario, a shift of activity on the organic soils mainly occurs when L’s farm profits from activity B start to or threaten to decline. On average 64.4 % of the H players that coordinate in rewetting the soils revert to activity B in stage 2. Despite the adoption of activity A in the later time periods by a significant proportion of the L players, the peat is strongly degraded on both farms. This stresses the importance of a policy instrument to preserve organic soils.

In both APC and APV, the average percentage of players who adopt activity A is significantly higher than in the baseline scenario (proportion test, p-values=0.00).5 A logistic random-effect panel regression on players’ adoption of activity A and the presence or not of an agglomeration payment confirms this result (Model 1, Table 9). It shows that the introduction of the agglomeration payment increases the odds of players adopting activity A. However, H players are less likely to vote for rewetting as compared to L players, and the effect of the payment is significantly weaker on H than on L players. This is in line with H’s farm profit not being at risk of decreasing, implying a weaker economic incentive to adopt activity A.

5 In both APC and APV, the percentage of players who adopt activity A is also significantly higher than in the baseline scenario at each time period (two-proportion test at 95 % confidence level) except in time period 5 (two proportion test, p-value = 0.15). 67

When we compare the results of both policy treatments, we find that on average adoption of activity A in APC is higher than in APV (proportion test, p-value = 0.00). This is also true when considering each time period separately (proportion test, 95 % confidence level). Again, a logistic random-effect panel regression of players’ adoption of activity A in all time periods confirms this result: players are more likely to adopt activity A in APC than in APV (Model 2, Table 9). Moreover, L are more likely to vote in favor of rewetting than H players, which is intuitive as farm profit of L player under activity B declines over time. Considering the first-stage decision, in APC we find that on average 73.4% (std. 6.4) of the groups rewet the soils. This rate is relatively constant across the 10 time periods as can be seen with the standard deviation. In contrast to APV, we find that a majority of players are able in APC to successfully coordinate in activity A during the ten time periods, which enables to preserve the peat. In APV, on average 50% of the groups rewet the soils, which is significantly lower than in APC (proportion test: p

= 0.04). Moreover, in APV 31.6% of the groups do not rewet because one of the player rejects rewetting while the other player votes in favor of it. H player is the player who rejects it in 67% of the case for such situation. Overall, the constant payment scheme appears to be most environmentally effective in incentivizing land use change. The results correspond to our predictions at the end of section 2.

Table 9: Logistic random-effect panel regression on players’ adoption of activity A. (1) (2) Variables Baseline + APC + APV APC + APV Time period 1.13*** (0.03) 1.10*** (0.03)

Treatment: (1): Baseline = 0, APC or APV = 1; (2): APC 118.7*** (67.1) 0.12*** (0.08) = 0, APV = 1 H player 0.22*** (0.11) 0.68** (0.28) Treatment*H 0.32** (0.18) 1.02 (0.59) Constant 0.14*** (0.05) 1.15 (0.51) Number of groups 254 176 Number of observations 2540 1760

Note: To control for group membership, all regression models also include group dummies (not reported). We report odds ratio: * p < 0.1, ** p < 0.05, *** p < 0.01. Standard errors in brackets clustered at the group level.

3.2. Income inequality and cost effectiveness

We use the Gini coefficient to compare the level of inequality in final payoffs among players and to understand the impact of the policy on income inequality. The coefficient varies from zero to one, with zero representing perfect income equality and a value of one corresponding to perfect income inequality. We compute the Gini coefficient in payoffs across all players in each time period and then compute the mean Gini coefficient across all time periods. We find that, in the baseline scenario, the mean Gini coefficient is equal to 0.34 (std. 0.21) and thus significantly higher than in the presence of a payment scheme (t-test: p-values = 0.00). The Gini coefficient in APC is equal to 0.14 (std. 0.02) and in APV it is 0.20 (std. 0.15). In the absence of payment, inequality in income is likely to be due mainly to the decline in farm profit from activity B for L-player types. In the presence of a policy, the payoffs of L players can be as high as those of H players, hence inequalities among players are lower. We find that the level of inequality in income is significantly larger in APV than in APC (t-test, p-value = 0.00). Variation of the level of inequality across time periods is also higher in APV than in APC, as seen from standard deviations. These two findings are explained by the fact that a significant portion of players do not adopt activity A immediately in APV (cf. section 3.1), implying that L players’ payoffs decrease over time. Moreover, in

APV, the potential payment received by L decreases as the layer of peat declines, while it is constant for H. Thus, the more L delays adopting activity A, the lower his/her expected payment from adopting this activity A and the higher the difference of payoff between L and H players. Next, we compare the two policy treatments with respect to cost-effectiveness. This is important because payments are potentially higher under the constant than under the variable payment treatment. Thus, the increase in environmental effectiveness in APC may come at the cost of high public spending. We define cost-effectiveness here as the funds spent per peat unit preserved, which is determined by dividing the total volume of peat preserved by the total payments made.6 The results are presented in Table 10.7 We find that the constant payment treatment preserves most peat, spends most money on policy payments, yet performs best with respect to cost-effectiveness.

Table 10: Comparison of the cost effectiveness of payment schemes

APC APV Peat volume preserved in percentage of the total, in % 45.2 15.7 Total payment spent along the 10 time periods, in monetary units 403,480 208,740 Percentage of the total payment possible, in % 59.5 30.8 Cost-effectiveness: funds spent per peat volume unit preserved 2027.5 3025.2

3.3. Effect of side payments on sustainable management

Figure 15 presents the average net side-payments offers made between group members across all time periods. A positive net H offer refers to the fact that the level of the offer made by H was larger than the one made by L, while a negative net H offer means that the offer made by L was larger than the one made by H. All reported average values referring to side payments are significantly different from zero at 95% confidence level (t-test). The results indicated that side payments are also made in the baseline scenario. We find that the mean net H offer equals 126.0 (std. 89) and increases over the course of the experiment. A potential explanation for this is that while H players have no economic incentive to offer a side payment to their partners, they may use these transfers to diminish payoff inequality. In fact, in 65.6 % of the cases where players rewet and where there is a positive net H offer, we observe that H chooses activity B and L chooses activity A. We find that 46% of the H players make at least one positive net offer to their group member along the ten time periods, and 50% of these H players are characterized as inequality averse.8 For a matter of comparison, only 19.0% of the H players that do not make any positive net offer are characterized as inequality averse (proportion test: p value = 0.04). Furthermore, the mean net H offer is not significantly different from zero in groups where H players are characterized as individualistic or as joint-profit maximizers. The analysis of the chat content of groups that communicate in the baseline

6 We calculate the volume of peat preserved for each player at the end of the 10 time periods. Given that after 5 time periods used in activity B, the peat is exhausted, the volume of peat remaining is a continuous variable taking values from 0 to 5. 0 corresponds to the situation where the player has conducted activity B at least 5 times during the 10 time periods, so there is no peat remaining; 5 corresponds to the situation where the player has not degraded the peat, i.e. he/she has conducted activity A continuously and the initial volume of peat remains. By summing up over all players, we calculate the total volume of peat preserved in each treatment. 7 Under a policy treatment (sample pop. = 88 players), total preservation of the peat would mean to preserve 88*5 = 440 vol. units. The maximum total payment which can be made under each policy treatment (i.e. all players cooperate from the first time period) is equal to 88*770*10 =677,600. 8 Two H players behaved in a purely inequality-averse manner by making a transfer equal to 367 to their group member and then rewetted the organic soils. 69 scenario (74% of the groups) reveals that 24% of those groups mention “equitability” of payoffs in their discussion.

600

500

400

300 Average net H offers offers net H Average 200

100

0 1 2 3 4 5 6 7 8 9 10 -100

-200

-300

-400 Time periods

Baseline Constant agglomeration payment Variable agglomeration payment Figure 15: Average net side-payment offers among group members across treatments

Under the APC treatment, we find that a net side-payment offer takes place in 47.5% of the instances and the mean net H offer is equal to -40.0 (std. 23.0). This is significantly different from the average net offer in the baseline scenario (t-test, p-value = 0.00). Thus, L players make larger offers than H players. We hypothesize that differences in bargaining power between the group members explain this result. This is confirmed by the chat content: we find that transfers from L are mainly requested by H players as a condition for voting in favor of rewetting and adopting activity A.9 It thus seems that a portion of the H players exploit the fact that rewetting is based on a unanimous decision and that farm profits of L decline if they would be forced to pursue activity B. In APC, both group members’ payoffs are better off when adopting activity A, and this is true in the absence of payoff redistribution. Groups for which both players adopt activity A do not make use of side payments in 55% of the cases. In those cases, based on the analysis of the communication content, we find that part of the groups clearly justify their cooperation decision either as a way to maximize each group member’s payoff or as a way to achieve equitability of payoffs.10 However, in 45% of the cases, successful joint cooperation in adopting activity A still goes along with a side payment.

In APV, just like in APC, payoff-maximizing players have no incentive to offer a side payment. We would only predict side payments from inequality-averse H players to L players in cases where players’ potential payments for activity A would differ (i.e. if n ≥ 3). We find in APV that side payments are mainly offered by H to L players: a net side-payment offer takes place in 36% of the time periods and

9 We provide two examples to illustrate this finding (quotes from players, translated from French): In one group, the H player said in time period 2: “we rewet and we both choose activity A and you give me your 770 points […]. Either you give me 770 and you will still earn 40, or I do not vote for rewetting and your profit will drop to zero”. In another group, the H player said: “okay (i.e. we cooperate from time period 4; note from authors based on the conversation content) but you pay me 10”. 10 Cooperating in order to maximize each other’s’ payoffs is observed, for example, through the following quotes: “If we rewet, with the subsidies, we can both earn more”, or “If we choose activity A, we both gain something”. Cooperating in order to equalize group members’ payoffs is observed, for example, through the following quote: “we raise the water table and then choose activity A so that we both earn the same”. 70 the mean net H offer is equal to 112.9 (std. 47.7), which is not significantly different from the mean net offer in the baseline scenario (t-test, p-value=0.37). Moreover, as expected, prosocial preferences seem to affect the level of side payment offers: over the ten time periods, the mean net H offer is equal to 163.8 (std. 80.5) in groups with prosocial H players, while it is equal to 71.8 (std. 83.4) in groups with individualistic H players.11 However, we do not find an effect of time on side-payment offers. Analyzing the effect of side payment on the adoption of activity A in this treatment, we find that a net side payment is executed in 35% of the cases where both players adopt activity A (mean net H transfer equal to 99.2, std. 126). In APV, transfers thus seem to be mainly driven by inequality aversion.

3.4. Effect of players’ socio-demographic and preference characteristics on their behavior

Finally, we explore the role of social preferences and other socio-demographic characteristics on the decisions of players. The mean SVO angle of players is equal to 25.2° (std. 13.9) in the baseline scenario,

25.1° (std. 15.7) in APC, and 25.3° (std. 13.8) in APV. It is not significantly different across treatments (t-test), indicating that our randomization at the individual level was successful. More information on the social preference data is provided in Appendix C5. We also draw on the individual socio-demographic characteristics from the short exit survey (see Appendix C6 for a detailed description of all variables measured). For our analysis we include participants’ care for the environment (Concern for environment), their willingness to take risks (Risk taking) and be patient (Patience), their attention to reputational concerns with respect to the opinion of others (Personal reputation), their general knowledge about degradation of organic soils in Switzerland (Knowledge peat degradation), and their trust in cooperative approaches in agricultural management (Belief in cooperative approaches). A higher score always indicates stronger preferences, e.g., a higher score on the measured willingness to take risks means that this individual is more risk seeking. Finally, we also control for some typical farming household characteristics of our participants including if participants or their parents own a farm (Farm at home) and if so, if the household has parcels located on organic soils (Peat presence).12

Our primary interest is to characterize the players who adopt activity A. However, conducting a regression analysis where the response variable corresponds to the adoption or not of activity A would imply that the players who conduct activity B include two distinct subsets: those who adopted or had to adopt activity B by default because rewetting was rejected, and those who voted in favor of rewetting but then decided to return to activity B. It is therefore important to examine the decisions in each stage separately. For this, we provide regression analyses with the dependent variable being the vote by players to rewet organic soils (first-stage decision) (Table 11) and with the dependent variable being the land use choice of players on rewetted organic soils (second-stage decision, Table 12). We report these regressions for the full sample, the payment treatments together, the baseline, and each payment treatment separately. The regressions report odd ratios. An odds ratio comprised between 0 and 1 indicates a lower likelihood to vote for rewetting (in Table 11) or for adopting activity A (in Table 12) for a higher level of the given variable. An odds ratio greater than 1 indicates a higher likelihood to vote for the same given a higher level of the variable.

11 The difference between these two mean net side-payment offers is significant when considering the standard deviation over the time periods’ average values (t-test, p-value = 0.004), but it is not significant when considering all values across time periods (t-test, p-value = 0.37). 12 These variables were measured as follows: concern for environment: index ranging from 0% to 100%, higher score indicate higher concern for environmental issues; risk taking: score ranging from 0 to 10, higher score indicates greater willingness to take risks; time preference: score ranging from 0 to 10, higher score indicates more patience in general; personal reputation: score ranging from 0 to 10, higher score indicates stronger concern for others’ opinions and therefore for own reputation. 71

Table 11: Logistic random-effect panel regression of players’ vote for rewetting organic soils (first stage) (1) (2) (3) (4) (5) VARIABLES Full sample APC & APV Baseline APC APV Time period 1.155*** 1.119*** 1.233*** 1.114* 1.122*** (0.0310) (0.0362) (0.0625) (0.0622) (0.0445) Treatment: in (1): APC or APV, 17.99*** 0.352 Experiment in (2): APV (9.547) (0.224) Player type (L = 0, H = 1) 0.0515** 0.00924*** 0.488 0.00289* 0.0421 (0.0646) (0.0163) (0.737) (0.00909) (0.0956) Concern for environment 0.998 1.004 0.958** 0.974 1.023** (0.00882) (0.0103) (0.0205) (0.0189) (0.0119) Risk taking 0.898* 0.843** 1.157 0.834 0.921 (0.0497) (0.0689) (0.127) (0.117) (0.0989) Patience 1.012 0.999 0.983 1.191 0.927 (0.0455) (0.0739) (0.0684) (0.160) (0.0720) Opinions and Personal reputation 1.118 0.964 2.217** 0.989 0.929 preferences (0.190) (0.250) (0.705) (0.472) (0.257) Knowledge peat degradation 0.826 1.009 0.831 1.570 0.909 (0.165) (0.255) (0.273) (1.058) (0.214) Belief in cooperative 1.379* 1.445 1.688* 3.700** 1.110 approaches (0.236) (0.365) (0.514) (1.911) (0.263) SVO angle 1.005 0.993 1.027 0.957 1.021 (0.0105) (0.0141) (0.0204) (0.0267) (0.0170) Peat presence 1.034 1.050 0.976 1.134** 1.018 Farm (0.0358) (0.0334) (0.0460) (0.0718) (0.0403) characteristics Farm at home 1.044 1.078 1.749 1.659 0.991 (0.216) (0.280) (0.855) (0.794) (0.290) Concern for environment*H 1.015 1.017 1.044 1.060* 0.987 (0.0134) (0.0159) (0.0295) (0.0339) (0.0193) Risk taking*H 1.146 1.383*** 0.738* 1.616*** 1.130 (0.109) (0.165) (0.129) (0.298) (0.181) Patience*H 1.062 1.201* 0.976 1.149 1.244* (0.0759) (0.129) (0.103) (0.222) (0.143) Personal reputation*H 1.116 1.279 0.590 0.958 2.303* (0.286) (0.464) (0.214) (0.520) (1.002) Interaction Knowledge peat degradation*H 1.267 0.908 1.479 0.922 0.908 terms (0.359) (0.331) (0.787) (0.675) (0.403) Belief in cooperative 0.864 0.714 1.466 0.173** 1.031 approaches *H (0.246) (0.240) (0.662) (0.136) (0.263) SVO angle*H 0.999 1.021 0.951* 1.070** 1.004 (0.0142) (0.0190) (0.0269) (0.0354) (0.0230) Peat presence*H 0.924 0.912* 1.112* 0.768*** 1.009 (0.0487) (0.0465) (0.0709) (0.0685) (0.0510) Treatment*H 0.715 0.831 (0.310) (0.335) Constant 1.405 12.13* 0.215 16.39 0.489 (1.333) (15.94) (0.320) (37.27) (0.857) Model Number of observations 2,480 1,720 760 870 850 Number of groups 248 172 76 87 85

Note: We report odds ratios: *** p<0.01, ** p<0.05, * p<0.1.

For the first-stage decision (Table 11), we find that across all specifications the likelihood of players voting for rewetting organic soils increases over time. In fact, time of play is the strongest and most stable predictor overall (Models 1 to 5). This is intuitive as the incentive of one player (L) to rewet the soils increases as his/her farm profit declines. In addition, players are more likely to vote for rewetting 72 in payment treatments when compared to the baseline scenario (Model 1), which is in line with our predictions and previous findings. Comparing player types, H players are generally less likely to vote for rewetting than L players (Models 1 and 2). This is consistent with our expectations: in the baseline scenario, a profit-maximizing H player has no incentive, at no time, to rewet the soils, while the L player prefers rewetting once the peat is degraded. In the presence of a constant agglomeration payment, the incentive of L player to rewet the soils also increases over time with peat degradation, while that of the H player remains the same. Under a variable agglomeration payment, considering the surplus provided by the payment, the incentive of the L player may not increase as much as under APC over time because payment declines in proportion to the decrease in profits from activity B. This is confirmed by the fact that the effect of player type is insignificant when considering only the APv treatment (Model 5, as compared to Model 4). When considering only the payment treatments (Model 2), we find that the likelihood of voting for rewetting is higher among risk-takers than among risk averse players. There is also an interaction effect between the willingness to take risks and the player type (see the interaction term Risks*H). Therefore, this effect of being risk taker refers to the L players, which is the reference level in our regression. One potential explanation is that if L players focus on maximizing profits from activity B during the first time periods and thus postpone cooperation, they take the risk of seeing their farm profit drop. The more willing the L player is to take risks, the less likely L is to vote for rewetting during the first time periods. With regard to the interaction effect, the effect of Risks for H player is significantly stronger than that of L players. The total effect of Risks on H players can be calculated by multiplying the two coefficients: it is equal to 0.843*1.383 = 1.17. Thus, the more H are risk takers, the more likely they would vote for rewetting the soils. This result matches with our initial expectations: in general, the agglomeration payment type generates a risk for players to vote for rewetting as the other group member could decide to defect in the second-stage decision, which would result in a non-allocation of the payment. This risk is even more important for H than for L players given that from n =3 onwards, the gain which can be made by L from the adoption of activity A is considerably larger than the gain which can be made by H. Next, we consider separate regressions for the baseline scenario and each of the two payment treatments (Models 3, 4, and 5). In the baseline scenario (Model 3), we find that the likelihood of voting for rewetting increases for players who “care for their reputation”. In particular, for the H players this can be interpreted as players who care about how they would be considered by their own group members or the experimenter if they would let their group member’s profits drop without taking any action. In addition, a belief in cooperative approaches appears to be a predictor of cooperating in form of rewetting. Under the constant agglomeration payment (Model 4), H players are significantly less likely to vote for rewetting than L players. As explained above, in contrast to H, the incentive of L players to rewet the soils increases in the course of the experiment as their profits from activity B decline. Furthermore, L players who believe in cooperative approaches in agricultural management are on average more likely to vote for rewetting the soils. The format of the payment (agglomeration scheme) may indeed work more successfully with L players that have a positive opinion of cooperative approaches. The effect of this belief is significantly stronger among L than H. Next, we find that the L players who reported having organic soils on their farmland have a higher likelihood of voting for rewetting. This finding is interesting: these players, who are knowledgeable about the issue of peat loss, may thus be more likely to raise the water table in order to preserve the future production potential of their soils, particularly when assigned the role of having an underlying mineral soil layer of low fertility. The effect of this variable is significantly stronger for L than for H. Finally, prosocial preferences (i.e. players with a higher SVO angle) and a willingness to take risks have a significantly stronger effect for H than for L.

This is consistent with the fact that H have a weaker economic incentive to adopt activity A than L players. In the variable agglomeration payment (Model 5), we find that a higher level of care for the environment significantly increases the likelihood of players voting for rewetting. In addition, the effects of reputation and patience are significantly stronger for H than for L players. Reputation may correlate here with the fact that a portion of the players prioritize vegetable farming over the economically incentivized sustainable land use. This is likely linked to farming identity. Next, Table 12 below presents the results from a logistic random-effect panel regression of players’ adoption of activity A in second-stage decision. Note that this regression only includes the players who rewetted the organic soils. One first observation is that, unlike in Table 11, belief in cooperative approach does not appear among these players as a major predictor of their decision. This is intuitive as stage 1-decision is the stage that primarily contributes to coordination among players. We find that the introduction of a payment scheme has a significantly positive impact on the likelihood that players adopt activity A (see Model 1) which confirms our main treatment effects. Furthermore, the payment effect differs depending on player type (see interaction term Treatment*H, Model 1). The introduction of a payment scheme has a significantly stronger impact on L than on H players with respect to the adoption of activity A. This is consistent with the fact that in the baseline scenario and in contrast with the scenarios with a payment, a large proportion of the players who rewetted the soils revert to activity B, which makes sense in terms of payoff maximization. In the baseline, on average 64.4% and 25.7% of the H and L players, respectively, who rewetted the soils revert to activity B. By contrast, the corresponding percentages are 13.9% and 8.2% in APC, and 24, 2% and 13.3% in APV. In addition, patience appears to be a weak predictor.

We now consider the effect of time period. In the baseline scenario and in APC (Models 3 and 4), adoption of activity A becomes more likely over time, which is consistent with our previous findings and the fact that the incentives for L to adopt activity A increase over time in these treatments. By contrast, in APV (Model 5) the likelihood that players adopt activity A declines significantly over time. As explained in section 2, the decline in L’s payoffs with peat degradation may induce a strong incentive for H players to bargain for higher side payments. Thus, those H players who prefer activity B may only vote for rewetting in later rounds and then cause the effect of time period in Model 5. In fact, at 90% confidence level, we find that H players are in APV less likely to choose activity A as compared to L players. Also, a few players prioritize activity B by reverting to activity B in stage-2 decision. Next, we consider each treatment scenario separately. In the baseline scenario (Model 3), belief in cooperative approaches has a stronger effect on the decision of L than on H players in this stage. In APC (Model 4), the decision over land use on rewetted soils is not significantly influenced by any players’ characteristics. A potential explanation is that in comparison with the variable payment treatment, the constant payment treatment seems to induce a strong incentive among players to coordinate in the adoption of activity A, as reflected in the previous sections. In APV, the odds of players adopting activity A declines with the level of importance given to personal reputation and increases with the presence of peat on farm land and the care for the environment. Moreover, risk takers L are less likely to opt for activity A. On the assumption that “care for reputation” is also related to farming identify (pride in being a producer), L players tend to maximize their profits from vegetable farming, and doing so increases their risk that their potential payment drops. We find that risk takers H are in contrary more likely to adopt activity A. Risk taking H players may be more prone to bargain for higher side payments at the risk of foregoing the benefits of the agglomeration payment.

Table 12: Logistic random-effect panel regression of players’ vote for adopting activity A in stage 2- decision (1) (2) (3) (4) (5) VARIABLES Full sample APC & APV Baseline APC APV Time period 1.033 1.007 1.257*** 1.187* 0.783* (0.0663) (0.0797) (0.109) (0.121) (0.101) Treatment: in (1): APC or 7.729*** 0.0307*** Experiment APV, in (2): APV (5.994) (0.0356) Player type (L = 0, H = 1) 0.480 0.0593 2.678 0.201 0.000496* (1.340) (0.187) (9.415) (1.007) (0.00221) Concern for environment 0.997 1.011 0.941* 0.993 1.048** (0.0173) (0.0186) (0.0338) (0.0275) (0.0207) Risk taking 0.898 0.922 1.102 1.024 0.532** (0.104) (0.144) (0.234) (0.217) (0.133) Patience 1.190* 1.155 1.103 1.390* 1.054 (0.117) (0.160) (0.153) (0.277) (0.183) Opinions and Personal reputation 0.672 0.587 0.756 1.708 0.131*** preferences (0.206) (0.245) (0.352) (1.307) (0.0700) Knowledge peat 1.372 1.889 1.168 1.874 2.559 degradation (0.590) (1.059) (0.672) (2.133) (1.756) Belief in cooperative 0.929 1.235 1.516 2.766 0.407 approaches (0.325) (0.558) (0.663) (2.112) (0.251) SVO angle 1.010 1.010 0.983 0.976 1.019 (0.0178) (0.0212) (0.0313) (0.0293) (0.0287) Peat presence 1.055 1.068 0.958 1.080 1.221*** Farm (0.0530) (0.0593) (0.121) (0.109) (0.0699) characteristics Farm at home 0.614 0.709 0.515 0.645 0.349 (0.259) (0.341) (0.399) (0.545) (0.239) Concern for 0.992 1.004 1.053 1.021 0.970 environment*H (0.0249) (0.0281) (0.0432) (0.0436) (0.0319) Risk taking*H 1.193 1.412 0.800 1.520 2.166** (0.211) (0.308) (0.232) (0.509) (0.753) Patience*H 0.916 1.069 0.973 0.830 1.208 (0.126) (0.204) (0.172) (0.251) (0.318) Personal reputation*H 0.817 0.679 0.687 0.300 1.048 (0.384) (0.388) (0.420) (0.309) (0.815) Interaction Knowledge peat 0.947 0.536 1.018 0.602 0.421 terms degradation*H (0.517) (0.356) (0.785) (0.728) (0.375) Belief in cooperative 1.090 1.054 0.188** 0.638 2.778 approaches *H (0.608) (0.666) (0.145) (0.743) (2.377) SVO angle*H 0.987 1.011 0.962 1.049 1.043 (0.0244) (0.0283) (0.0381) (0.0438) (0.0417) Peat presence*H 0.907 0.916 0.888 0.857 (0.0657) (0.0655) (0.127) (0.0956) Treatment*H 0.140*** 1.144 (0.105) (0.808) Constant 6.658 2.876 12.46 0.161 1,136** (12.78) (7.591) (31.30) (0.562) (3,955) Model Number of observations 1,323 1,069 252 644 425 Number of groups 228 162 65 85 77

Note: We report odds ratios: *** p<0.01, ** p<0.05, * p<0.1.

4. Discussion and conclusion

The computerized framed experiment that we conducted helped to gather valuable insights about the potential decision-making process among farmers with respect to management practices on organic soils and their responses to two alternative agglomeration payment policies. We find that players are more willing to implement sustainable use of organic soils when an agglomeration payment scheme is present than when this is not the case. Without any policy intervention, only the low-production-potential players (L) have an incentive to rewet organic soils in order to pursue an extensive sustainable activity, but only once their profit from vegetable production drops to zero. In terms of environmental effectiveness, the agglomeration payment is therefore a promising approach for promoting sustainable use of organic soils and thus preserve the carbon stock. Nevertheless, the overall cost of such a policy measure may be considerable. We find that the policy effectiveness depends strongly on the design of the payment scheme. A constant agglomeration payment scheme, aligned with initial highest opportunity costs of farmers, turns out to be more environmentally-effective and also more cost-effective than a variable payment that follows the dynamic evolution of opportunity costs. In addition, a constant payment scheme leads to less income inequality among farmers. More equal group outcomes are partly driven by preferences of inequality aversion, as inequality averse subjects more frequently offer side payments aimed at equalizing income differences. Under the constant payment, a large proportion of the players coordinate in the adoption of activity A early on in the experiment, and thus preserve their peat layer. Under the variable payment, the level of asymmetry between players of the same group is larger: in addition to their farm profit from vegetable farming being affected by peat degradation, L players face decreasing potential payments. A large proportion of the players under the variable payment do not achieve or maintain a cooperative equilibrium or delay cooperation, which hampers the preservation of the peat. Therefore, for the case of promoting sustainable use of organic soils in Switzerland, the more common approach of constant payments appears also more appropriate when designing agglomeration payments. This is promising because variable payments are usually also more difficult to implement as they require prior knowledge of the opportunity costs of farmers. In our setting it would require sampling each farm plot to determine the nature of the underlying mineral soil layer and the thickness of the peat layer. Moreover, payments that are differentiated according to opportunity costs are sometimes (falsely) perceived as less equitable. Although this perception can be questioned because equal payments imply unequal rents, the perception often makes differentiated payments politically less feasible. Thus, it is promising that the easier to implement option of constant payments also performs better on environmental effectiveness, cost effectiveness and income inequality in our setting. Notably, in both payment schemes tested, a considerable number of players did not adopt the sustainable land use despite having an economic incentive to do so. This is also one major finding in this study. It is important to understand the potential reasons to this lack of cooperation in order to improve the design of policy instruments and to predict their impacts. The lower level of cooperation could have several reasons. First, agglomeration payments as well as the collective nature of the drainage systems involve a risk of coordination failure. Thus, risk averse subjects, particularly those with high-production- potential in activity B, may prefer to continue that land use and obtain certain outcomes. However, we did not observe such an effect in the above analysis. Second, some high-production-potential farmers may exploit the deteriorating position of the low-production-potential farmers and delay cooperation to bargain for higher side payments. In reality, this could be even strengthened by strategic considerations not reflected in our experiment, such as the potential that the reduction in L types’ vegetable farming potential over time could induce higher vegetable prices and provide a possibility to extend farm 76 acreages for H types when L types go out of business. Third, behavior may be triggered by other non- economic factors, such as farming identity. To overcome these barriers and ensure rapid cooperation among farmers in order to preserve as much of the peat as possible and to prevent delays generated by bargaining processes, the incentive provided by the payment needs to be high. A higher incentive than the one tested (which represented 2% more than the opportunity costs of adopting the sustainable land use) could enable a more rapid coordination process. The speed of coordination between players and the maintenance of cooperation over time is indeed crucial in this context, as every time period in which vegetable production is continued, causes significant carbon emissions and soil degradation. However, in the case of intensive vegetable production, the opportunity costs of switching to extensive land use are considerable. Given that our goal was to identify a policy option that effectively promotes the adoption of sustainable use of organic soils, we assumed in this study an unconstrained budget for the regional governmental agency allocating the subsidies. In reality, this is obviously not the case. Because of budget constraints, the payment may be too low to create a sufficient incentive for farmers to coordinate with each other for the purpose of rewetting the soils and then to adopt an extensive land use. We addressed this point in the previous chapter that tests the design of lower constant agglomeration payments based on average costs. Alternative stricter policy approaches such as environmental regulation may be necessary to ensure the preservation of the peat, as voluntary agri-environmental payments may not be a sufficient incentive to overcome the barriers mentioned above. Finally, policy design should take into account farmers’ opinions and preferences. Risk aversion among farmers, their opinion of cooperative approaches, and concerns about their own reputation have been identified as determinant factors. Our experiment reflects the key features of the management issue on intensively used organic soils. Thereby, it enables a higher level of representation of the reality as compared to the majority of economic experiments that aim at deriving environmental policy recommendations. Yet, our experiment still has limitations regarding the external validity of our findings. First, it abstracts from other potential decisional factors such as uncertainty on the quality of the underlying mineral soil, crop yields, and flooding events. An extension of the experiment could also consider that farmers have to invest in finding out the quality of their underlying mineral soils. Second, while the experiment is conducted with farm apprentices, one could also conduct this experiment with experienced vegetable producers. Third, the decisional situation is evaluated over a 2-hours experiment while the entire process of soil degradation and potential loss of profits would take decades. During such a time, farmers’ incentives to modify management practices on organic soils as well as their social preferences are likely to change, and time preferences may be important additional determinants of behavior.

CHAPTER 4 External validity of experiments in environmental economics: framing and subject pool effects among students and professionals1

1. Introduction

The use of laboratory experiments to study issues in environmental policy has grown in prominence within environmental economics. Yet, the external validity or the generalizability of the results from these laboratory studies has been generally questioned (e.g. Harrison and List, 2004; Exadaktylos et al., 2013). The latter is crucial for studies that aim to inform environmental policy. This paper contributes to this methodological discussion by addressing the effect of framing and subject pool on experimental outcomes. It has become standard in experimental economics to use context-free and neutrally framed instructions in order to retain experimental control (Durlauf and Blume, 2009). For example, many scholars use public good games presenting the game in completely abstract terminology and then draw conclusions with respect to environment-related issues (e.g., Barrett and Dannenberg, 2012; Cinyabuguma et al., 2005). Others choose to use context-loaded instructions where contributions are interpreted as climate change mitigation (e.g., Milinski et al., 2006, 2008; Tavoni et al., 2011). While there has been discussion on whether to use heavily contextualized instructions or not, systematic comparisons are rare. We further develop this point in the next section. Another strand of this literature analyzes the potential bias of using university students in behavioral studies, and focuses on involving instead either “professionals” or randomized non-selected population. While some find that the type of participant strongly determines the nature of the results (e.g. Anderson et al., 2013), others do not find differences between the behavior of students and non-students and recognize the convenience of using students for laboratory experiments (Fréchette, 2014). We designed a visualized experiment that is fully contextualized by a specific agricultural problem. It deals with farmers’ management decisions on intensively-cultivated organic soils. The problem arising is that the draining of organic soils is necessary for intensive agricultural activities but at the same time leads to the loss of the top soil layer and the related production potential. This soil loss leads to substantial negative environmental externalities, particularly greenhouse gas emissions. The experiment aims to test conservation payments as means to incentivize the adoption of a more sustainable use of these soils. We ran separate sessions of the experiment with farm apprentices and with a sample of generic university students. In addition to the contextualized version, we also developed a context-free (unframed) version of this experiment, which we only conducted with university students. We studied both subject-pool and framing effects with a static and a dynamic experimental design. While the static design simplified the farmers’ decision situation considerably, the dynamic design captured more of the actual complexity in the dynamics of soil degradation. We also examined the effect of personal characteristics on behavior within and among the various experimental setups. We find no significant effect of the introduction of a specific context in the decision environment on behavior. The rate of players who adopt sustainable land use is not significantly different in the presence

2. Literature review

2.1. Subject-pool effects

While most experimental studies make use of the usual sample of university students (see e.g. Danielson and Holm, 2007), there is a growing literature analyzing experimental results from different subject pools, including samples of large representative populations and samples of professionals or specialists. Fréchette (2009) provides a meta-study comparing studies including students and professionals in laboratory environments. Some of these studies highlight significant behavioral differences between the two subject pools and others do not. An initial reason why it may not be possible to generalize the behavior of students to field behavior is the possible difference in the distribution of social preferences (e.g. Carpenter and Seki, 2011). A second potential bias to the behavioral generalization is the familiarity of the subject with the experiment, e.g. if players encounter the experiment’s social dilemma in their work milieu, but also the absence of elements of the work environment from the experiment (Fréchette, 2009). In the latter case, professionals may either assume the presence of certain features of their work and behave accordingly, or some of their behavior may only be triggered by specific signals that are not represented in the experiment (Fréchette, 2014). Third, self-selection may cause interpretation issues. Most studies, however, do not find different social inclinations between volunteers and non-volunteers (Falk et al., 2011; Anderson et al., 2013). A few studies involve professionals in games that imitate a particular working environment (e.g. Kagel et al., 2001). Yet, none compares professionals with students in such a contextualized experiment. In the presence of differences between the behavior of students and professionals, professional’s behavior is observed to be slightly further off equilibrium predictions than that of students. For example, Carpenter and Seki (2010), in a study similar to ours, compare the behavior of students to resource users, specifically fisherman, in a public goods game. They find that fishermen contribute significantly more than students do. This is also in line with the broader literature on subject-pool effects which shows that representative populations and professionals usually behave more prosocial than students in typical social preference games (e.g. Fehr and List, 2004; Bellemare and Kröger, 2007; Falk et al., 2011; Exadaktylos et al., 2013; Anderson et al., 2013; Belot et al., 2010). While the literature largely supports the use of students in experiments, involving professionals can provide precious and unique insights (Fréchette, 2009). In particular, there is evidence of a link between the use of different subjects pools and the experimental context provided, which may trigger signals that do or do not matter to the decision-making process of a particular type of subject (Belot et al., 2010; Fréchette, 2014). One such aspect is the framing.

2.2. Framing effects

Framing effect corresponds to a shift in the subject’s decisions or preferences induced by an alternative way of describing a particular situation or problem (Druckman, 2001; Frisch, 1993). A well-known study is the Asian Disease Problem by Tversky and Kahneman (1981), which illustrates the reversal of preferences when the effects of a medical program were presented in two logically equivalent manners. Other types of framing approaches have been tested in experimental economics, (cf. Levin et al., 1998) such as by varying the formulation of an incentive scheme or giving the frame a specific connotation that affects individuals’ social preferences. Most of this literature finds a significant effect of framing on subjects’ behavior (e.g., Tversky and Kahneman, 1981; Hossain and List, 2012; Gächter et al., 2009; Pillutla and Chen, 1999; Elliott et al., 1998a, b; Crisafulli et al., 2008; Rucker et al., 2008). Yet, a few studies find no effect of framing (e.g., Meier, 2006; Rege and Telle, 2004). Most of these experiments use abstract designs and address framing by either playing on the negative/positive connotation of the framing tool, which means that they use a negative or a positive oriented formulation referred to as “valence framing”, or by shaping the wording towards a certain type of behavior, e.g. self-interested vs cooperative, such as in Bernold et al. (2015). Important for this study is the influence of culture. For instance, the introduction of a deliberate rhetorical framing related to the cultural background of participants has been shown to affect behavior significantly (Cronk, 2007). In this paper, we test the effect of framing induced by the introduction of a specific field context. The study by Cronk and Wasielewski (2008) shares similarities with our study. They show that minimal framing (mainly the label of the game is changed) and brief exposure to an unfamiliar social norm can produce strong behavioral effects. An increasing number of authors work on understanding the effect of frame on how subjects view their decision and maximize their utility, and on the creation of norms (e.g., Pillutla and Chen, 1999). A specific experimental context as compared to an abstract experiment may indeed affect internalized social norms of individuals, and/or their interpretation of others’ behavior (Ellingsen et al., 2011). This means that some people may associate a context with a certain kind of behavior (e.g. a primary interest in payoffs in an economic context) (Koneberg et al., 2010).

3. Methodology

3.1. Experimental design

Our experimental design captures the three key components of the decision situation of farmers on organic soils (see chapters 2 and 3 for further details on the framed version of the experiment). First, sustainable use of organic soils is considerably less profitable than current management practice that relies on heavy drainage of these soils. Second, the adoption of sustainable land use requires farmers to cooperate among each other. Several farmers depend on a joint drainage system. Therefore, rewetting the soils, which is necessary for the more sustainable land use, requires unanimous agreement. Third, organic soils spatially differ in their profiles. Some farmers have high quality soils beneath the peat layer and others have poor quality underlying soils. Considering that the peat disappears under intensive land use (due to drainage), farmers are therefore heterogeneous in their soil production potential in the intensive land use, and differ in their opportunity costs of adopting the sustainable land use. In the experiment, players are placed in groups of two. Each group consists of a High-production- potential player (H) and a Low-production-potential player (L). In the experiment, players were referred to as Blue farmer and Yellow farmer. Each player needs to decide between intensive and sustainable use of the soil. The decision procedure, repeated in each round of the baseline and the treatment phase is as follows: each group member first votes in favor or against rewetting the soils (stage-1 decision). 81

This is because rewetting is a necessary step to the establishment of sustainable land use. If at least one player of the group rejects rewetting, both players automatically pursue intensive cultivation and their respective payoffs (πH and πL) are as in Figure 16. If both players vote for rewetting, the drainage is stopped and intensive use of the soils is no longer possible. Each member needs then to decide between adopting the sustainable land use (profit R) and reverting to the intensive land use by installing a personal drainage system at cost C (stage-2 decision). If soils are rewetted, players’ possible payoffs are as in Figure 17. Players need to cooperate to adopt the sustainable land use. Because of asymmetric opportunity costs, they may, however, differ in their incentives to do so, which may trigger negotiation. Players may also want to reduce payoff inequalities. Thus, before proceeding to the vote, each player can make a binding side-payment offer to his/her group member. The offer is only executed if the player to whom the offer was made adopts the sustainable land use. Side payments made by player i (i=L, H) to the other player are denoted as Si and are included in the payoffs in Figure 17.

Player 퐇

Intensive land use 휋퐿 Player 퐿 Intensive land use 훑퐇 Figure 16: Players’ payoffs if soils are not rewetted. Note: Payoffs of the H farmer are indicated in bold print

Player 퐇 Sustainable land use Intensive land use R + UI + APL – SL + SH R + UI + SH Sustainable land use R + UI + APH + SL - SH 훑퐇 - C - SH Player 퐿 휋퐿 - C - SL 휋퐿– C Intensive land use R + UI + SL 훑퐇 – C Figure 17: Players’ payoffs if soils are rewetted, including side payments. Note: In the treatment phase: UI = uniform individual payment, APi = agglomeration payment to player i. The agglomeration payment is either uniform (UA) or differentiated according to players’ opportunity costs (DAi). Si: side payment made by player i to his/her group member. Payoffs of the H player are indicated in bold print.

In the treatment phase, we test one of four conservation payment schemes to foster the adoption of sustainable land use: - A differentiated agglomeration payment (DA) which mirrors the opportunity costs of players for adopting sustainable land use. Following the definition of Drechsler et al. (2007), the agglomeration payment is only allocated to the players if both group members adopt the sustainable land use. - A uniform agglomeration payment (UA) which is also conditional on both players adopting sustainable land use, but pays an equal amount to players. We test two versions of this payment: one is aligned on H player’s opportunity cost and one is set as an average across both players’ opportunity costs.

- A uniform individual payment (UI) which is set as an average across players’ opportunity costs but is only conditional on the individual player’s land use, regardless of the other player’s land use choice. The three payment treatments are also represented in the payoff matrix in Figure 17. We refer to chapters 2 and 3 for an overview of behavioral predictions associated to the treatments explored in this paper. In addition, we conducted this experiment in a “static” and “dynamic” setting. This enabled to further test the validity of our results. In a “static” setting, in line with most economic experiments, the decision procedure is strictly repeated over all rounds (i.e. repeated one-shot decision-making). The “dynamic” setting captures the dynamics of the management problem; it depicts the effect of land use on soil. This includes that H’s farm profit is constant over time while L’s farm profit declines across rounds used for intensive land use (due to of his/her underlying soil layer not being suitable for intensive land use).2 The dynamic setting is described in detail in chapter 3 (see also footnote 2). Finally, the experiment also included an experimental elicitation of social preferences, which was conducted at the outset of the main decision procedure, and a short exit survey. Figure 18 summarizes the timeline of the overall experiment.

3. Treatment phase (test a 1. Elicitation of 2. Baseline phase (no policy intervention) payment scheme that promotes individual social sustainable land use) 4. Exit survey preferences 10 rounds 10 rounds

Figure 18: Organization of the experiment

3.2. Framed and unframed experiment

The basic experimental setting as described in 3.1 refers to the framed (contextualized) version of the experiment. The unframed (context-free) experiment differs from the framed version as follows. First, while the visual background of the framed experiment depicts the agricultural landscape in the Swiss region concerned by the problem of cultivated organic soils and thus even visually reflects the impact of players’ decisions on their land, the background of the unframed experiment is unspecific (see Appendix D1 for screen shots). Second, the wording of the unframed experiment includes no reference to agricultural terms nor farming management decisions. Instead of referring to farmers, participants are simply referred to as players; land use options are referred to as activities.

3.3. Subject pools and organization

We conducted the framed experiment with both professionals and a generic sample of university students (from the Swiss Federal Institute of Technology (ETHZ) and University of Zurich). Professionals in our case are farm apprentices from Swiss agricultural schools, who have a strong background in agriculture and most of them intend to become farmers. Some of these already work on

2 The experiment is calibrated as follows. In the static setting, players’ payoffs in intensive land use are invariant across rounds: πL = 2500, πH = 4500; R = 250, C = 220, DAL = 2300 and DAH = 4300; UI = UA = 3300 (i.e. (2300+4300)/2). In the dynamic setting, H’s profit is constant (πH= 800); L’s profit is negatively affected by intensive land use: it declines and falls to 0 as th L from the 6 round in intensive land use. With πn,t defined as the payoff in intensive land use at time t given a number n of L L L L previous rounds used in intensive land use (n={0,...,9}): πn,t = 800 if n = 0, 1, 2; π3,t = 550 ; π4,t = 160 ; and πn,t = 0 if n = 6, …, 10; R = 40, C = 25; UA = 770. 83 a farm, most often of their parents. In addition, we conducted the unframed experiment with a sample of university students only. Table 13 presents the treatments tested for the purposes of this paper across frames and subject pools.

Table 13: Experimental set up Framed experiment Unframed experiment Apprentices Students Students Baseline (78) Baseline (80) Baseline (76) Dynamic setting UA (88) UA (80) UA (76) Baseline (58) Baseline (222) Baseline (78) UI (30) UI (74) - Static setting DA (28) DA (74) - - UA (74) UA (78)

Note: number of participants in brackets. UI = uniform individual payment, UA = uniform agglomeration payment, DA= agglomeration payment differentiated according to players’ opportunity costs.

4. Results

We derive our results based on two main questions: 1) How is performance of our treatments affected by subject pool and by the experimental framing? and 2) how do the individual characteristics of players affect their decisions depending on subject pool and across framings? For the first question, we examine different performance criteria including environmental effectiveness, cost effectiveness, and effect on income inequalities. We first analyze how the behavior of farm apprentices compares to that of university students and then investigate how the framing affects players’ behavior. In addition to analyzing absolute performance, we also examine whether the subject pool and the type of experimental framing, here the level contextualization, affect the ranking of the payment designs in terms of performance. Henceforth, we refer to the intensive land use as “activity 1” and to the sustainable land use as “activity 2”. We also term the adoption of sustainable land use as “cooperation”. Below, reported average values correspond to the average of the means of the given variable over all rounds.

4.1. Impacts of subject pool

4.1.1. On environmental effectiveness In this study, a higher rate of cooperation means a higher level of the environmental outcome being provided, specifically a larger amount of peat being preserved. We examine cooperation across subject pools. Figure 19 illustrates cooperation across rounds graphically. We find that cooperation rates (i.e., percentage of players who adopt activity 2) of students and apprentices in the baseline treatment are very much similar until round 5, but start diverging from round 6 onwards. In fact, the average percentage over all rounds of players who adopt activity 2 is significantly higher among students (40.3%, std. 5.4) than among apprentices (23.6%, std. 3.2) (proportion test, p-value = 0.02). We further examine this effect with a panel regression on cooperation (Table 14, Model 1). We find that H-player types are less likely to adopt activity 2 than L types, and the effect of the subject pool is significantly stronger for H than for L types. This is one important finding of this study: students seem to more strongly focus on payoff maximization as compared to farm apprentices. In the treatment phase with a uniform agglomeration payment (UA) to incentivize cooperation, we find that the mean percentage of players who cooperate is again significantly higher among students than among farm apprentices (97.3% (std. 3.2) and 65% (std. 3.0) respectively; proportion test, p-value = 84

0.00). We confirm this subject-pool effect in UA with a panel regression on cooperation (see Table 14, Model 2). Students are significantly more likely to cooperate than farm apprentices. One possible explanation is that the behavior of students is more strongly driven by payoffs than that of farm apprentices.3 Thus, students aim at cooperating and maximizing their payoffs from the payment incentive earlier on in the experiment as compared to farm apprentices. Furthermore, the effect of subject pool is significantly stronger on L than on H players. In this treatment, L players have an increasing incentive to adopt activity 2 along the experiment.

100 90 80 70 60 50 40 30 20 10 0 Percentage of players who adopt activity 2activity adoptwho players of Percentage 1 2 3 4 5 6 7 8 9 10 Rounds Baseline, Students (80 players) Baseline, Apprentices (88 players) UA, Students (80 players) UA, Apprentices (88 players) Figure 19: Adoption of activity 2 across subject pools in baseline and UA (framed design, dynamic setting)

We now investigate the impact of the subject pool with the static setting. In the baseline scenario, on average 5.7% of the apprentices and 3.5% of the students adopt activity 2 in all rounds. This cooperation rate is not significantly different (proportion test, p value = 0.44). Yet, we find a weak effect of subject pool on cooperation (Table 14, Model 3). Apprentices appear to be more likely to adopt activity 2 than students in the baseline scenario. In both payment treatments (UI and DA), we find no significant effect of subject pool on the adoption of activity 2; see Table 14, Models 4 and 5.4 The average percentage of players who cooperate among students versus farm apprentices is equal to 42.8% (std. 6.1) vs. 36.3% (std. 7.3) in UI (proportion test, p-value = 0.51) and 50.8% (std. 6.3) vs. 49.3% (std. 14.5) in DA. An analysis of each decision stage separately led to the same findings. In summary, in the dynamic setting, environmental effectiveness differs significantly by sample population in the framed experiment. The effect is less pronounced in the static setting. Regardless of subject pool, all payment design options rank the same in environmental effectiveness.

3 In the experimental economics literature, students behave on average more individualistically than the rest of the population. Here, students behave more cooperatively (less selfishly) than the farm apprentices. One main reason is that, under the treatments tested, cooperation is associated to profit maximization. Second, students seem to be less influenced by the farming background and be less inequality averse, which enable them to cooperate more rapidly in the sustainable land use. 4 In UI, the mean percentage of players who adopt activity 2 is 42.8% (std. 6.1) with students and 36.3% (std. 7.3) with apprentices (proportion test, p-value = 0.51); in DA, it is 50.8% std. 6.3 with students and 49.3% std. 14.5 with apprentices. 85

Table 14: Random effect logistic panel regression on player’s cooperation (1) (2) (3) (4) (5) Variables Baseline UA Baseline UI DA dynamic dynamic static static static Round 1.440*** 1.163** 1.144* 1.035 0.856** (0.0959) (0.0769) (0.0764) (0.0610) (0.0542) Subject pool (0 = apprentices, 1 = 0.657 479.0*** 0.291* 3.830 3.323 students) (0.432) (574.0) (0.178) (4.230) (4.942) H player 0.243*** 0.650** 0.627 0.0944*** 1.415 (0.0135) (0.0238) (0.00510) (0.0316) (0.0555) H player*Subject pool 10.10*** 0.650** 1.334 1.694 1.420 (5.365) (0.137) (0.859) (1.392) (0.454) Constant 0.0208*** 2.262 0.0223*** 0.149 0.333 (0.0164) (1.962) (0.0210) (0.219) (0.794) Number of observations 1,580 1,680 2,800 1,040 1,020 Number of groups 158 168 280 104 102 Note: Odds ratios are reported: *** p<0.01, ** p<0.05, * p<0.1 Robust standard errors clustered at group level in parentheses. Models 3 and 5 weakly fit the data: general p-values of the models respectively equal to 0.09 and 0.06. All regressions include a dummy variable referring to the group membership of the player (not reported).

4.1.2. On the use of side payments and cost effectiveness We analyze the effect of subject pool on side payments between group members (see Appendix D2 for an overview of offered and executed side payments). For this, we examine net side-payment offers from player H to player L (i.e., SH – SL). A positive net H offer refers to the fact that the level of the offer made by H was larger than the one made by L, and a negative net H offer means that the offer made by L was larger than the one made by H. We first examine the experiment conducted in a dynamic setting. In the baseline scenario, we find no significant effect of the subject pool on side-payments offers. In the UA treatment, we find a significant effect of subject pool on side payments. The mean net H side- payment offer is significantly higher among students than among apprentices: it is equal to -250 and - 40, respectively (t-test, p-value = 0.00). However, considering all rounds, 47.5% of the apprentice groups make a side payment offer (either positive or negative net H offer) as compared to 10.2% of the student groups, which is significantly different (proportion test, p-value = 0.00). Note that in this treatment, H types may have an incentive to bargain given that L’s profit decreases in the event of no cooperation. The analysis of the communication content between group members confirms that the proportion of H players who condition their cooperation on a side payment and therefore delay group cooperation is significantly higher among apprentices than among students. Next, we analyze the experiments conducted in a static setting. We find no significant effect of the subject pool on the side payments in these treatments. Next we analyze cost-effectiveness, defined as the total amount of soil units preserved per money unit spent on conservation payments. The distinction in the use of bargaining power between subject pools in the UA-dynamic treatment is also reflected in the cost effectiveness of the payment scheme. We find that cost-effectiveness of the UA payment is higher among students than among apprentices (see Table 15). This result is due to the fact that students cooperate early in the experiment, and thus are able to preserve most of the peat. For the static setting, cost effectiveness does not largely differ between subject pools in both treatments (see Table 15). We also observe that the ranking with regard to cost- effectiveness of the two payment schemes examined in the static setting does not vary with the type of subject: UI appears as more cost effective than DA for both subject pools.

Table 15: Impact of the subject pool on cost-effectiveness (framed design) Peat preserved as Payment made as % of the Cost- N % of the total total payment possible effectiveness Dynamic UA 45.2 59.5 2,027.5 88 Apprentices UI 36 36.3 3,330.5 30 Static DA 49.3 44.3 2,965.22 28 Dynamic UA 93.75 97.5 1,601.6 80 Students UI 42.7 42.8 3,310.4 74 Static DA 50.7 45.4 2,956.8 74

4.1.3. On income inequality Using the Gini coefficient, we analyze the effect of our treatments on income inequality between players and compare these effects across subject pools (see Table 16). The Gini coefficient varies from 0 (perfect equality between payoffs) to 1 (perfect inequality). We calculate the Gini coefficient for each round based on players’ payoffs and then report the average Gini coefficient across all rounds. For each treatment considered, the coefficient differs significantly between subject pools (t-test, 95% confidence level). This is due to the low variation of the variable across the rounds (cf. standard deviation). The highest difference is observed for UA in the dynamic setting: the coefficient is equal to 0.14 with farm apprentices and 0.27 with students. This is consistent with our previous finding: side-payment offers are higher among students than among farm apprentices, hence a higher redistribution of payoffs and a higher level of inequality in incomes among players. However, as for the measure of cost effectiveness, we find that the subject pool does not affect the ranking of policy options with regard to their impacts on income inequalities (see UI and DA treatments - static, Table 16).

Table 16: Test of subject pool effect on inequalities (framed design) Gini coefficient (std.) N Baseline 0.34 (0.21) 78 Dynamic UA 0.14 (0.02) 88 Apprentices Baseline 0.18 (0.02) 58 Static UI 0.13 (0.02) 30 DA 0.18 (0.03) 28 Baseline 0.27 (0.16) 80 Dynamic UA 0.03 (0.01) 80 Students Baseline 0.16 (0.01) 222 Static UI 0.11 (0.00) 74 DA 0.17 (0.02) 74

In summary, and along the three examined performance criteria, the type of subject affects the performance of the payment treatments in their magnitude. However, the effect is less pronounced in the static setting than in the dynamic setting. Moreover, we find that, for the treatments tested in this study, subject pool does not affect the ranking of policy options in their performance. This means that conducting experiments with regular university students still seems to enable identifying most and least performant policy schemes.

4.2. Impacts of framing

Again, we examine the effect of framing on environmental effectiveness among the sample of university students. For this, we compare the percentage of players who cooperate across the two framings, for both the baseline and treatment phases (see Figures 20a and 20b for visual illustration). We also conduct

87 a panel regression on cooperation across all treatments accounting for the framing effect (Table 17). We find no significant difference in decisions of university students, regardless of whether the experiment is played framed or unframed, and in either setting (static or dynamic). Thus, there appears to be no significant effect of framing on decisions of students for these experiments.

100

40 activity 2 activity 20

Percentage of players adoptwho players of Percentage 1 2 3 4 5 6 7 8 9 10 Rounds Baseline framed (80 players) Baseline unframed (76 players) UA framed (80 players) UA unframed (76 players) Figure 20a

100

activity 2 activity 20

Percentage of players adoptwho players of Percentage 1 2 3 4 5 6 7 8 9 10 Rounds Baseline framed (222 players) Baseline unframed (78 players) UA framed (74 players) UA unframed (78 players) Figure 20b Figure 20: Proportion of players who cooperate in the baseline and in treatment UA across framings, in the dynamic setting (Figure 20a) and in the static setting (Figure 20b)

Table 17: Random effect logistic panel regression on player’s cooperation among students (1) (2) (3) (4) Baseline dynamic UA Dynamic Baseline static UA static Round 1.521*** 1.40** 1.034 1.076 (0.101) (0.235) (0.0733) (0.0647) Framing (0=unframed, 0.342 1.21 0.926 0.196 1=framed) (0.247) (3.163) (0.731) (0.261) H player 0.0433*** 1.0 0.620 1.026 (0.0185) (0.404) (0.0894) H player*framing 2.378 1.0 1.398 1.181 (1.487) (0.00002) (0.975) (0.156) Constant 0.0300*** 635.9 0.0102*** 7.130 (0.0208) (2371) (0.0118) (11.94) Number of observations 1,560 1,404 3,000 1,520 Number of groups 156 156 300 152 Note: We report odds ratios: *** p<0.01, ** p<0.05, * p<0.1 Robust standard errors in parentheses. In UA dynamic (Model 2), the variation of the response variable is very low: nearly all players cooperate: on average 97.3% std. 3.2 in the framed and 93.9% std. 3.9 in the unframed design.

We also analyzed the use of side payments and the potential impact of framing on policy cost effectiveness. We find no significant effect in both dynamic and static settings and across all treatments, of framing on side-payments offers (see Appendix D3 for an overview of those results). With respect to the impact on income inequalities, we compare the Gini coefficients for all treatments. Again, we do not observe a noteworthy effect of framing on income inequalities (see Appendix D4 for an overview). In summary, along the three examined performance criteria, the experimental framing does not affect the performance of the payment treatments explored in this study with a sample of university students. Given that each experimental setup for the analysis of framing effect includes only one payment treatment, we cannot formulate conclusions with respect to the effect of framing on the performance ranking of policy options.

4.3. Effect of individual players’ characteristics across framings and subject pools

4.3.1. Distribution of social preferences across subject pools The SVO slider measure that we conducted at the outset of the experiment uses the test proposed by Murphy et al. (2011) and provides us with an individual measure of social preferences. This test enables to elicit the extent to which the player places importance on his/her own income in relation to the outcome of other players. An angle is defined for each player (ranging between -16.26° and 61.39°) and characterizes his/her level of prosociality. The higher the individual’s SVO angle, the stronger his/her social preferences. The average SVO angle is not significantly different between the two subject pools: it is equal to 24.8 (std. 14.8) in farm apprentices and to 24.2 (std. 13.2) in university students (t-test, p- value = 0.64). Yet, the distribution of individual measured SVO, illustrated in Figure 21, varies between students and apprentices: apprentices tend to exhibit higher SVO angles, while the students show a similar peak at higher SVO angles, and an additional second peak at lower angles. The SVO angle also allows categorizing players in distinct types. The most common types are proself (i.e. selfish types) and prosocial (i.e. altruistic types). In line with our observed distributions, we find that the percentage of players characterized as individualistic is significantly higher among students than among apprentices. (45.7% versus 34.3%, respectively; proportion test: p-value = 0.013). This likely explains the higher rate of successful coordination among students than among farm apprentices.

Students Apprentices 33.77 34.25

30.92

19.86

20 15.79 15.75

Percentage 10.96 10.27 8.904

10 5.921 6.164

2.74 1.37 1.096 .6579.8772 .6849

0 -20 0 20 40 60 -20 0 20 40 60 SVO angle Percent Figure 21: Distribution of the SVO angle among students (456) and apprentices (146) Percent Graphs by G

4.3.2. Impact of individual characteristics on players’ decision in link with framing and subject pool We are interested in the influence of social preferences and other characteristics on players’ behavior in relation with the framing and the subject pool. Besides the SVO angle we include further important characteristics which we elicited in our exit survey: environmental preferences (index ranging from 0% to 100%, higher score indicated stronger concern for environmental issues); risk preferences (score ranging from 0 to 10, higher score indicates greater willingness to take risks); patience level (score ranging from 0 to 10, higher score indicates more patience); reputation (score ranging from 0 to 10, higher score indicates more concern for others’ opinions). Appendix D5 presents a summary of all collected characteristics. We conduct a random-effect logistic panel regression on players’ cooperation (i.e. adoption of activity 2), and test for effects of interaction between players’ personal characteristics and subject-pool vs. framing on players’ decisions (see Tables 18 and 19 respectively). First, we analyze the effect of the subject pool (see Table 18). The main findings are as follows. We find in UA-dynamic (Model 2) that reputation is a predictor for player’s adoption of activity 2. We also find an interaction effect between players’ concern for reputation and subject pool. The effect of reputation refers therefore to farm apprentices, which are the reference level in the regression. Because of their difference in cultural background, apprentices may be more sensitive to reputation than regular students, inducing a higher likelihood to adopt activity 2. The total effect of reputation on students is calculated by multiplying the two coefficients, i.e. 0.206*2.903 = 0.60. Thus, the effect of reputation is working opposite for students, possibly highlighting the fact that concern for reputation does not affect profit- maximization strategies among students. Next, we find that the effects of environmental consideration, willingness to take risks, and social preferences are each significantly stronger for students than for apprentices in this very same treatment. One possible explanation may be the familiarity of the player with the context offered in this experiment. Apprentices who are familiar with the farming context and consider it their professional future activity may be less influenced by environmental or social preferences. For this reason, for instance, a high level of care for the environment may induce a relatively higher level of cooperation in students than in apprentices. Interestingly, the intensity of the effect of risk depending on the subject pool varies between UA-dynamic (Model 2) and DA-static (Model 5). In UA and in line with the previous result, the effect of risk on players’ cooperation is stronger among students than among apprentices. In UA, students are more likely to cooperate than farm apprentices. Risk aversion linked to the agglomeration-payment type may be a stronger limiting factor to cooperation among students than among apprentices. In DA, it is the other way around: the effect of risk on players’ cooperation is stronger among farm apprentices than among students. One potential explanation may relate to prosocial preferences. Farm apprentices may pay higher attention to income equalities than university students. Given that successful cooperation facilitates payoff redistribution, farm apprentices may thus be more sensitive to the risk of unsuccessful cooperation than university students. In UI (Model 4), the type of subject does not interact significantly with any players’ characteristics in a way that affects players’ decisions. Next, we analyzed the interaction effects between framing and the characteristics of players (see Table 19). In UA-dynamic (Model 2), we find that the willingness to take risks and the consideration for the environment appear as significantly stronger predictors of cooperation in the framed than in the unframed design. The introduction of a resource management problem (with a farming context) may induce non-economic considerations in the decision-making process (e.g. linked to the environment), which then makes the player more sensitive to risk and to environmental preferences. In the static setting baseline scenario (Model 3), we find that risk takers are less likely to cooperate, and this is true in the unframed design (reference level). We have an interaction effect: the effect of willingness to take risk 90 is stronger in the framed than in the unframed design. The total effect of risks for the framed design is equal to 0.718*1.406 = 1.17. Thus, in the framed design, the more players are risk takers, the more likely they would adopt activity 2. One possible explanation is that in the framed design, the player may not only account for the impact of his/her decision on payoffs but also the impact of the decision on the soils or the environment, which induces different types of risk considerations. Moreover, we find that patient players are more likely to cooperate and this effect is significantly stronger in an unframed than in a framed design.

Table 18: Panel regression on player’s adoption of activity 2 including player characteristics and their interaction with subject-pool effects (framed design) (1) (2) (3) (4) (5) VARIABLES Baseline UA Baseline UI DA Dynamic Dynamic Static Static Static Round 1.463*** 1.156** 1.144* 1.036 0.854** (0.0987) (0.0763) (0.0822) (0.0618) (0.0550) H-player type 0.0352** 0.173 0.0174* 0.000592** 0.141 (0.0509) (0.511) (0.0363) (0.00179) (0.574) Subject pool (0=apprentices, 1=students) 0.867 0.135 0.147 5.874 0.885 (1.211) (0.516) (0.325) (19.53) (4.544) Environment 0.979 1.021 0.999 0.975 0.942 (0.0142) (0.0241) (0.0240) (0.0431) (0.0537) Risks 0.962 0.832 0.790** 0.757 1.238 (0.0996) (0.143) (0.0787) (0.165) (0.286) Patience 1.005 1.037 0.885 0.910 0.792 (0.0816) (0.120) (0.0868) (0.222) (0.227) Reputation 1.038 2.903** 0.385*** 1.071 0.221* (0.311) (1.439) (0.135) (0.683) (0.183) SVO angle 1.010 0.970 1.025 0.988 0.954 (0.0205) (0.0241) (0.0242) (0.0365) (0.0429) Environment*H 1.018 1.021 1.045* 1.069* 0.970 (0.0181) (0.0340) (0.0265) (0.0424) (0.0515) Risks*H 0.860 1.349 1.070 1.083 0.917 (0.115) (0.322) (0.103) (0.203) (0.242) Patience*H 1.002 1.133 0.944 1.042 1.355 (0.112) (0.214) (0.0996) (0.209) (0.419) Reputation*H 0.957 0.225** 1.148 0.508 2.419 (0.320) (0.157) (0.447) (0.335) (1.927) SVO angle*H 0.970 1.033 1.017 1.040 1.040 (0.0224) (0.0342) (0.0204) (0.0429) (0.0605) Environment*Subject pool 1.008 1.087** 0.977 0.976 1.076 (0.0168) (0.0447) (0.0237) (0.0433) (0.0697) Risks*Subject pool 0.983 1.763*** 1.279** 1.198 0.548** (0.101) (0.380) (0.149) (0.245) (0.150) Patience*Subject pool 1.050 0.837 1.147 1.035 0.623* (0.102) (0.194) (0.128) (0.220) (0.167) Reputation*Subject pool 1.103 0.206** 1.746 0.851 1.886 (0.386) (0.140) (0.655) (0.577) (1.299) SVO angle*Subject pool 1.008 1.065* 0.990 1.001 1.045 (0.0229) (0.0384) (0.0244) (0.0395) (0.0439) Subject pool *H player 12.70*** 0.505 0.628 0.598 1.655 (8.034) (0.342) (0.307) (0.605) (1.220) Constant 0.0902 1.300 0.295 11.50 1,957 (0.166) (3.704) (0.666) (42.81) (11,037) Number of observations 1,560 1,670 2,780 1,020 1,010 Number of groups 156 167 278 102 101 Note: Reported odds ratios are: *** p<0.01, ** p<0.05, * p<0.1 Robust standard errors clustered at group level in parentheses.

Table 19: Panel regression on player’s cooperation including players’ characteristics and their interactions with framing effect (students) (1) (2) (3) (4) VARIABLES Baseline UA Baseline UA Dynamic Dynamic Static Static Round 1.519*** 1.380* 1.034 1.076 (0.102) (0.234) (0.0731) (0.0648) H player 0.0591 0.0342 0.0420 4.07e-05*** (0.135) (0.166) (0.0966) (0.000139) Framing (0=unframed, 1=framed) 1.602 3.14e-07** 0.962 1.840 (3.384) (2.08e-06) (2.236) (7.175) Environment 0.991 0.976 0.935*** 0.971 (0.0173) (0.0350) (0.0236) (0.0280) Risks 0.936 0.797 0.718* 0.638** (0.108) (0.200) (0.122) (0.130) Patience 1.091 0.889 1.826*** 0.920 (0.100) (0.291) (0.383) (0.141) Reputation 0.846 0.328 0.983 0.908 (0.334) (0.345) (0.789) (0.543) SVO angle 1.019 0.949 0.995 1.041 (0.0166) (0.0499) (0.0238) (0.0281) Environment*H 0.999 1.077 1.037 1.129*** (0.0221) (0.0654) (0.0311) (0.0483) Risks*H 0.601*** 0.781 1.058 1.771** (0.110) (0.310) (0.131) (0.420) Patience*H 1.123 0.749 0.863 0.950 (0.162) (0.294) (0.109) (0.156) Reputation*H 2.162 1.990 1.232 0.353 (1.036) (2.627) (0.576) (0.274) SVO angle*H 0.979 1.032 1.018 1.055 (0.0336) (0.0724) (0.0229) (0.0461) Environment*Framing 1.001 1.159** 1.047 0.971 (0.0192) (0.0695) (0.0327) (0.0417) Risks* Framing 1.054 2.675*** 1.406** 1.066 (0.143) (0.776) (0.241) (0.247) Patience* Framing 0.959 1.184 0.580** 1.044 (0.0984) (0.448) (0.124) (0.197) Reputation* Framing 1.101 0.429 0.642 1.901 (0.449) (0.514) (0.499) (1.390) SVO angle* Framing 0.987 1.071 1.021 0.972 (0.0183) (0.0631) (0.0260) (0.0325) Framing *H player 1.385 1.594 2.211 1.549 (0.799) (1.600) (1.684) (0.722) Constant 0.0302 746,050** 0.0589 87.09 (0.0774) (4.526e+06) (0.130) (265.1) Number of observations 1,550 1,395 2,990 1,510 Number of groups 155 155 299 151 Note: Odds ratios are reported: *** p<0.01, ** p<0.05, * p<0.1 Robust standard errors clustered at group level in parentheses.

5. Discussion and conclusion

Our comparative analysis reveals that, along the tested criteria, the type of subject taking part in economic experiments may significantly impact behavior of players, while the experimental framing does not appear to have a significant effect on the decisions of players. This is true in the absence and in the presence of a policy promoting a specific activity. Environmental effectiveness and payoff redistribution, and hence cost effectiveness and impact on income inequalities, differ between apprentices and students. Yet, the subject-pool effect is mainly present with the dynamic experimental setting that closely captures the resource management problem. The weaker impact of the subject pool on decisions in the static setting may be explained as follows. The static setting is an abstraction of the

92 actual farming situation and may therefore be perceived by farm apprentices as less connected to the actual agricultural decision situation and therefore may induce a higher level of payoff-maximizing decisions. By contrast, the dynamic setting gives a more real depiction of agricultural management decision and the actual soil degradation dynamics. First, this may trigger questioning among farm apprentices (and not among students) with regard to farming identity, such as “shall I maximize my soil potential by producing vegetables until I am no longer able to or shall I benefit from the payment incentive and stop producing?” Second, the dynamic setting may trigger more emotional responses. For instance, observing soils degrade over time triggers more concerns about the L player. A potential explanation to the difference in behavior between the two populations is the representation of social preferences, which differs between university and agricultural students; this is in line with the literature. Furthermore, we find that the impacts of several players’ characteristics on their decision vary significantly across subject pools and across framing designs. This includes willingness to take risks, social preferences, environmental consideration, and caring about personal reputation. This paper thus highlights the importance of identifying the personal characteristics of subjects that could potentially be affected by the choice of an experimental setup, as this could be a bias factor in the estimation of behavior. In conclusion, subject pool affects the size of policy impacts but tends to not affect the ranking of policy options along the performance criteria. As policy recommendations are based usually on the ranking and general direction of effect and not on its exact magnitude, the type of subject used in the experiment may not matter so much. However, this result needs to be nuanced with the type of treatment investigated. In this paper, we only explored the effect of payment schemes for which there is a collective gain from adopting the sustainable land use. Group members have therefore an incentive to negotiate and cooperate. Another type of treatment could more severely influence the nature of the outcomes across subject pools. This is especially important when the experimental study intends to draw environmental policy implications from the results. Furthermore, the type of subject does affect the magnitude of policy impacts and policy costs, which can make predictions difficult. Two further limitations of our study need to be highlighted. First, there is a need to analyze a larger range of payment treatments in order to conclude on the impact of both subject pool and framing on policy ranking. In this study, this analysis was only possible under the static setting and considering the effect of subject pool. We could not compare the impact on ranking of different policy options in the dynamic setting. Yet, the impact of subject pool seems to be more pronounced for the dynamic setting, which was depicting the actual decision situation more closely. It is thus possible that subject pool may also affect the ranking of policy options in that setting. Second, our analysis of framing effects focused only on university students. Further research is needed on the impact of framing among a professional population: How would the framing affect the behavior of farm apprentices? Would farm apprentices be more sensitive to the framing than university students?

GENERAL DISCUSSION AND CONCLUSIONS

This final section highlights the major results and contributions of this PhD project and reflects on the main outcomes from the thesis. It also highlights limitations of the conducted research and needs for further investigation.

1. Contributions to a more sustainable management of organic soils

1.1. Expanding knowledge on a complex socio-ecological issue

The management of degraded and degrading peatlands has mainly been approached from an ecological point of view. Most studies are merely descriptive and mainly address the loss of important ecosystem services and the substantial greenhouse gas emissions resulting from the degradation of peatlands. Some of these studies also focus on agricultural techniques to restore naturally peat-accumulating areas. The study region explored in this thesis has already lost the key landscape features that characterize a peatland, with regard to flora and fauna especially. This is due to historical drainage of these organic soils and their use for intensive vegetable production. Because of drainage, the area releases substantial GHG emissions. However, the area still carries a significant layer of peat that has the potential, if preserved, to contribute to climate change mitigation. Furthermore, the rapid loss of the peat - 1-2 cm per year - makes the management of these soils a highly sensitive topic for the future of agricultural activities in this area. Despite the urgency of the issue, the current use of these soils is barely reconsidered, which is partly due to the complexity of the problem. This thesis addresses the sustainable management of organic soils from an environmental-economic perspective. It thereby contributes to a more complete analysis of the management problem on organic soils and to the assessment of potential alternative management strategies and the use of new, economic instruments for the preservation of the remaining peat. In the study region, the absence of long-term preservation strategies of the organic soils is due to four main factors (cf. chapter 1): the high profitability of the soils, the difficult economic environment of farmers, the uncertainty on the spatial distribution and properties of soils, and the low awareness of society on this issue. The lack of precise soil data makes the design of policies addressing the negative externalities induced by the degradation of these soils particularly complex. All potential scenarios of development of these soils imply costs, for which the extent varies depending on the time scale considered (short vs. long-term) and the main sector incurring these costs (e.g. private vs. society). A reflection is necessary on the current and future use of these soils before the peat disappears. This implies to discuss policy options that could be suitable to address a change of management practices on these soils.

1.2. Agri-environmental payments for a sustainable management of organic soils

Chapters 2 and 3 of this thesis provide experimental evidence on the potential of economic incentives as policy instruments to promote the adoption of sustainable management practices on organic soils. In the absence of such policy intervention, only some farmers, particularly those who face a short-term decline in their production potential when organic soils are degraded could have an incentive to adopt alternative uses of organic soils. However, this incentive is likely to only emerge after farmers have maximized their profit from conventional land use, and therefore when the peat layer is heavily degraded

95 or depleted. For most farmers, adopting more sustainable practices involves very high opportunity costs. Economic incentives in the form of agri-environmental payments to promote a change in management practices have the potential to overcome this barrier to peat conservation. Given the political influence of the agricultural sector, agri-environmental payments also appear more politically feasible than environmental taxes or command-and-control regulation. In order to assess the potential of agri-environmental payments for promoting a sustainable management of organic soils, the present thesis uses economic experiments to study the effect of alternative agri- environmental payment schemes. Behavior of students and farm apprentices in the laboratory is studied in a situation with and without policy intervention. The design of the alternative policy interventions account for the need for farmers to cooperate in rewetting the peat area and for the asymmetry among farmers in the long-term conservation costs of adopting sustainable management of organic soils. The accounting of both of these key aspects in the design of agri-environmental payments is a major contribution of the thesis. Constant agglomeration payments calibrated according to initial opportunity costs of adopting a sustainable use of the soils appear more environmentally- and cost-effective than variable agglomeration payments that mirror the evolution of opportunity costs over time (cf. chapter 3, dynamic experiment). On the other hand, uniform agglomeration payments set as an average between participants’ opportunity costs appear to be as effective as differentiated agglomeration payments and uniform payments set individually (cf. chapter 2, static experiment). The strong performance of uniform agglomeration payments in face of heterogeneity in opportunity costs is made possible through successful payoff redistribution between low-cost and high-cost farmers via side payments. Agglomeration payments are therefore promising options in promoting cooperation among farmers in rewetting the soils and in adopting sustainable management. The choice between uniform (average) and constant agglomeration payment by policy makers will mainly depend on the type of data available. Computing the level of the uniform payment scheme requires data on both the highest and the lowest production potentials across farm lands. In the case of the constant payment, only data on the highest production potentials is necessary (areas were the peat layer is the thickest). The fact that uniform payments perform well is encouraging because they tend to be easier to implement both in terms of information requirements and in terms of political feasibility. Uniform payment design may also be more easily accepted by farmers than payments for which the design either points at inequalities among farmers. However, uniform payments set as an average of farmer’s conservation costs would necessitate contracts or side-payment agreements between low-costs and high-costs of farmers. This will generate transaction costs. For this reason, a constant payment scheme may be a more promising option for such urgent matter. While it creates some bargaining issues, just like for the uniform payment scheme, payoff redistribution among farmers of a same area is not necessary to their cooperation. Furthermore, it is important to highlight that the acceptance of a certain payment design depends very much on the context and the population to whom that payment is addressed. In some cases, the introduction of an agri-environmental payment may offend those farmers who already contribute, without any payment, to the provision of public positive externalities. In some instances, it has been shown, that this can lead to perverse outcomes as the intrinsic motivation for conservation is “crowded- out”. In the case investigated in this thesis, payments are tested on farm apprentices. Therefore, evidence for crowding-out of incentivized conservation behavior could not be tested. Finally, the total cost of such policy for Switzerland needs to be addressed. Assuming constant payments calibrated according to initial opportunity costs of adopting a sustainable use, the cost of preserving the

96 organic soils in Seeland using such payment scheme would be equal to, at least, 16,687,500 CHF/year.1 This highlights the need to identify the soil locations for which the preservation would be the most/least beneficial to society (i.e. significant thickness of the peat), as well as the spatial distribution of soil production potentials. This also highlights the need to evaluate the potential of other policy options, such as laws and regulations that could be more environmentally and cost effective than payments to sustain organic soils. Given that Swiss farmers already receive subsidies (cf. chapter 1), an elimination of subsidies of farm products produced on organic soils and drainage technology needs to be considered. Generally, this issue needs to be addressed through a combination of both climate and agricultural policy frameworks. This means on the one hand a framework permitting the offset of emissions from domestic cultivated peat areas, and on the other hand a framework facilitating a change in the use of these organic soils by promoting change at farm-management level and establishment of new economic activities on rewetted soils. The former could include the investment by nature conservation organizations in the preservation of these areas and the establishment of a voluntary carbon markets enabling land managers to buy carbon credits. While the latter did not appear as promising in the Swiss context, it would likely be appealing with regard to cost effectiveness; more research is needed in this direction.

1.3. Sustainable development of organic soils

Is it better to preserve the remaining peat layer than to maintain vegetable production on these soils that are in any case threatened in the future? This question implies a comparison of the short-term societal benefits generated by agricultural production on these soils to the long-term societal benefits generated by the preservation of these very same soils and their carbon store. A few studies have in fact estimated the long-term benefits from the preservation of organic soils to be higher than the costs of their conservation, including opportunity costs (see also chapter 1). For the Seeland region, it is therefore necessary to compare these two scenarios with a long-term time horizon. Due to the continuous reduction of the peat layer, certain locations are no longer drained efficiently. Furthermore, one major complexity highlighted in chapter 1 relates to the substantial lack of data on the distribution of soil types, soil properties, and thickness of the peat layer. This poses high uncertainties in the possibility for maintaining current agricultural practices and in the effects associated with possible action plans on these soils. In addition, a situation where organic soils would be degraded and only farmers with good underlying mineral soils would be able to produce vegetables would also be problematic as the surface level of the soil would still be low, and would necessitate drainage. This raises important concerns with regard to the viability of pursuing vegetable farming on organic soils and with respect to financing the maintenance and replacement of drainage systems, which is extremely costly. Furthermore, if agri-environmental payments are considered as the most suitable policy options, their duration or permanence would need to be discussed. Such a policy program with extraordinarily high agri-environmental payments may rather serve as a bridge to enable land users to modify the use of their organic soils. Less costly long-term measures would need to be explored as more profitable production activities on rewetted soils are also being developed. As depicted in chapter 1, the preservation of organic soils refers mainly to a political and societal prioritization of trade-offs between current and

1 This cost is calculated as follows. The total acreage of cultivated soils under vegetable farming in Seeland represents about 2500 ha (chapter 1). Assuming that 50% of these soils consist of organic soils, 1250 ha is the acreage that could be enrolled into the conservation program. The opportunity costs for adopting an alternative activity on rewetted organic soils are estimated at 14,000 – 650 = 13,350 CHF/ha/year (chapter 1). The total opportunity cost of preserving the remaining peat in Seeland is therefore equal to 13,350*1250 = 16,687,500 CHF/year. 97 future losses. Increasing the awareness of the public on this issue and its diverse implications will definitely facilitate such a decision. This thesis provides some inputs to support this process.

2. Contributions to behavioral and experimental economics

2.1. An innovative tool of analysis

A major outcome of this thesis resides in the design of an innovative experiment that is highly contextualized by the issue of cultivated organic soils. A unique experiment is developed based on an investigation of the context and decision situation of vegetable producers on organic soils, an understanding of the implications of current management practices on organic soils, and an evaluation of sustainable management practices on organic soils. Through an initial detailed understanding of the issue at hand, this experiment incorporates key features that are likely to affect farmers’ decisions in the use of the soils. As a result, it combines several complexities that have not been addressed in the broader literature on the design of payments for environmental services. While the game structure resembles in parts the commonly applied coordination game structure used in previous analyses of agglomeration payments, it also differs in important ways. First, it studies a two-stage decision problem consisting of a collective choice and an individual land use decision. Second, it incorporates player heterogeneity in opportunity costs. Third, it analyzes the potential of side payments to deal with the asymmetry in payoffs among players. Fourth, it considers the impact of a dynamic evolution of opportunity costs as a result of environmental degradation. Each of these aspects adds to the literature and only their combination allows to capture key features of the problem examined. The experiment permits to make an informed prediction of the potential behavior of farmers under different policy options that promote sustainable use of organic soils. In contrast to a modelling approach, the experimental approach allows the analysis of alternative policy designs without making strong assumptions on human behavior. The method further allows the linking of the outcomes of the experiment to social preferences of the individuals and other personal socio-economic characteristics that were collected using separate surveys. While developing this type of highly contextualized experiment is costly and very demanding (e.g. in terms of programming, videos for the experimental instructions), contextualization has two main advantages. First, it enables more guidance, more control, and less room for the expression of players’ own interpretations. Indeed, in the case of neutral instructions for instance, which subjects also interpret, it is difficult to identify how players bring their exogenous, heterogeneous background in the experiment. Second, contextualization has the benefit of being more convincing to policy makers. It shows a higher level of understanding of the complexity of the issue, which is crucial to the development of suitable policies. This is even more the case if the experiment is conducted with the population concerned by the issue. Thus, findings may have higher chances to be taken on board for the design of policies.

2.2. Understanding cooperation among players

With respect to the literature on game theory and behavioral economics, it is important to reflect on the proportion of players who adopt a sustainable use of organic soils in the presence of an agri- environmental payment. Assuming profit-maximizing players, this proportion is lower than one could predict. A significant portion of players (farm apprentices and university students) do not adopt sustainable land use, e.g. more than 30% in the constant uniform payment (chapter 3), despite the fact

98 that there are gains from trade from joint cooperation (i.e., joint payoffs are higher under the sustainable land use). There are several potential explanations for this finding. First, it is largely recognized in behavioral economics that players do not act purely on payoff-maximization. It is commonly observed in experimental economic studies that some players behave following non-economic motivations. This study finds also strong evidence for social preferences, particularly inequality aversion, influencing individual behavior and resulting cooperation. Coordination failure - on one of the equilibria or on the payoff dominant equilibrium specifically - is acknowledged as a common phenomenon in the laboratory, for which the major determinants are trust and risk attitudes (Devetag and Ortmann, 2007). The rate of successful coordination can be increased through repeated interactions that reduce strategical uncertainty (Banerjee, 2011), communication among players (ibid), and possibility for players to disclose their actions to others (Masiliunas, 2016). A second explanation to the low rate of adoption of the sustainable land use is the very high opportunity costs that characterize the case examined in this thesis. Those may necessitate strong incentives for players to change activity, especially under agglomeration payment treatments that involve a risk of coordination failure. In this regard, it would be interesting to conduct the experiment with a higher incentive than the one tested in this study. Third (among the farm apprentices mainly), a portion of players seems to have a higher preference for conventional land use than for the alternative land use, despite the economic incentive. This could be linked to farmers’ identity as producers, which needs to be accounted for in the design of payment schemes. The stronger the farming identity, the higher the required incentive for farmers to cooperate may need to be. Fourth, highlighting inequalities among farmers leads to bargaining issues. In the agglomeration payments in particular, a number of farmers tend to abuse the difficult situation of other farmers with short-term production potential by conditioning their cooperation on a side payment from their partner (chapter 3). Fifth, coordination failure may be related to one of the players wanting to defect in stage 2. This player may want to either “punish” the other group member for not complying with the promised side-payment offer, or diminish payoffs inequalities without losing much gain. Finally, group members build their equilibrium over time. It usually takes a few rounds for them to reach their optimal equilibrium, with regard to land use and side payment level. Each group has its own dynamic, which depends on players’ individual preferences. An additional aspect not accounted for in the experiments, but which is also likely to influence farmers’ behavior are time preferences. These are linked to how farmers envision the future of their farm. For instance, farmers who do not intend to pass their farm onto their children may place little importance on the depletion of the peat layer, even knowing that the underneath mineral soil layer does not suit current practices. In this thesis, experiments were conducted with (young) farm apprentices. There was therefore no control for the succession of the farms to future generations. It is also well known that time preferences do not play out as importantly in short lab experiments as they do in real world decision making involving long time spans. It should be noted that economic theory could suggest that low soil production potential is reflected in land prices. If this is expected, farmers should not make a difference between passing on farms to their children or selling the farm at the end of their lifetime. Yet, in our study period, competition for land in other non-agricultural uses may confound the impact of soil degradation on land value.

2.3. Importance of the choice of subject pool in experimental economics

From the perspective of policy implications, this thesis also investigates the extent to which the behavioral outcomes of university students and farm apprentices are comparable (cf. chapter 4). It contributes to the understanding of the extrapolation of the behavior of students, who are standard

99 subject pool in economic experimental studies, to the field population. The results of this thesis are mixed in this regard. While the behavior of the two subject pools is comparable in terms of direction of effects and the ranking of policy options, the magnitude of the results and the impact of socio-economic characteristics and preferences on behavior varies significantly between the subject pools. First, university students reveal more homogeneous patterns of decision-making than farm apprentices. Second, the proportion of players who adopt the sustainable management of organic soils is lower among farm apprentices than among university students, which affects the environmental and cost-effectiveness of the policy. Third, farm apprentices make a higher use of bargaining for side payments than university students, which probably explains why they delay their cooperation in the experiment. Moreover, social preferences and other personal characteristics of players have a significant influence on the decisions of players and this influence differs across subject pools. Indeed, apprentices were characterized by a higher level of prosociality and were more influenced in their decisions by aspects related to farming than by aspects related to environmental concern, risks, and social preferences. Yet, apprentices showed a higher concern for inequalities in payoffs as compared to students. Due to experience, background, history, and perceptions, these various effects may be even stronger, and the results more divergent, with a pool of experienced farmers. As observed with the treatments conducted in a static setting, the type of subject does not affect the ranking of payment schemes in their performance. This is good news, as the identification of most versus least performant policy options is a fundamental output for policy recommendations. Based on this, one could therefore conclude that conducting experiments with university students is a reliable step to assess the behavior of a given population within a certain policy context. However, this study did not enable to analyze the effect of subject pool on the ranking of payment scheme performance using a more complex experiment resembling the real-world situation. This thesis also contributes to the discussion of the external validity of results generated by abstract experiments. The introduction of a specific agricultural framing does not significantly affect the behavior of university students for the treatments tested.

2.4. The role of social preferences and other socio-demographic characteristics

This thesis reveals that social preferences are strong predictors of cooperation among players (in both farm apprentices and university students), and eventually, of the effectiveness of the agri-environmental payment schemes. Players characterized as prosocial are generally more likely to successfully coordinate with their group members and to agree on rewetting the soils. As an example, in the absence of policy, players with short-term production potential on organic soils face a challenging situation. Some of the players with long-term production potential are willing to incur additional expenses and to share part of their resources in order to enable their group member to switch to the sustainable land use. Moreover, inequality-averse are frequently willing to redistribute payoffs in order to diminish inequalities in income among players. Moreover, under these two payment treatments while farmers have a joint incentive to adopt the alternative land use, side payments are taking place from the long- term production potential to those having short-term production potential farmers. Such a behavior is not consistent with payoff-maximizing behavior. Through data collected via the SVO test, this thesis contributes to an understanding of social preferences as drivers of the different types of behavior. The difference in behavior between farm apprentices and university students is also partly linked to social preferences. The prevalence of social preferences and their effect on decisions differ between the two subject pools. Moreover, the impact of social preferences and some other personal characteristics on behavior differ between framed and unframed versions of the experiment (cf. chapter 4). This indicates that framing may well affect the behavior of farm apprentices. 100

Other personal characteristics of participants appear to be strong predictors of players’ decisions, namely willingness to take risks, belief in cooperative approaches in agricultural management, care for the environment, and personal reputation. Their consideration may therefore be important to the development of policy instruments for the sustainable management of organic soils. Furthermore, as stressed in chapter 1, factors of pride and identity also need to be considered in the design of policy interventions. They are potential constraints to the adoption of sustainable practices, amplified by the lack of knowledge/information about the timeline over which soils are being degraded.

3. Limitations of the thesis and future research needs

This thesis has several limitations. These are important to highlight and they imply potential directions for future research. This section first stresses the relevance to explore the effect of other types of policy approaches to address the management of organic soils. Second, it highlights the importance to predict the conditions of acceptance for sustainable management of organic soils among other stakeholders than farmers. Third, it discusses one particular aspect of the experimental design. It also reinforces the importance to explore some of the players’ personal characteristics in more details. Finally, it discusses the overall transferability of the experimental results.

3.1. Exploration of the potential of other types of policy instruments

The thesis examines the potential of agri-environmental payments to induce a change of management practices on organic soils. Those appeared as the most relevant options, especially considering the need to secure cooperation among land users. However, future research is required to explore the effect of other types of policy approaches to solve this management issue. This includes other types of market- based incentives, such as taxes or product certification for vegetables produced outside the peat areas. The latter option may not be able to compensate farmers for their lost profits from raising the soil water table. The former option, following a polluter pays principle, could be a more efficient approach to preserve the remaining peat if the preservation of organic soils is prioritized at the national level. An environmental tax could be levied on draining organic soils. The tax could only concern the zones that present a high remaining store of carbon (i.e. thick peat layer) and therefore a high potential for societal benefits in a rewetted state. Depending on the interest of society and policy makers, the tax scheme could also be established gradually. In a first phase, the tax level could be - above a certain threshold – a function of the average intensity of the soil drainage along the year. Few farmers already implement a more moderated regulation of the water table, which means that they allow a higher water table than they used to in periods that do not necessitate accessing the field. Thus, they aim at reducing the rate of peat degradation and increasing the longevity of conventional land use on organic soils. While such a tax is likely to face strong opposition from farmers, it has also advantages: it does not require public funding and therefore does not lead to possible inefficiencies in allocating this funding (e.g. lack of additionality), and it creates revenue for society. A second type of policy approach that needs to be investigated and compared to the potential of agri-environmental payments to solve this management issue is a command-and-control approach. This could consist of standards on the level of the water table and zoning of areas where drainage would be prohibited. Nevertheless, such a policy instrument would also require detailed data on the properties and level of degradation of the soils in order to identify the most promising areas for peat preservation and for societal benefits. Such cartographic representations are not available at the moment. Furthermore, such regulatory approaches are generally recognized to be less cost effective than market-based incentives.

101

A further option to be considered consists of payments conditional on performance, i.e. outcome-based payments instead of activity-based payments. Those may be more stimulating to farmers. The idea would be that farmers are not paid on the basis of measures but are paid for proving the quality of their soil. Such an approach has the advantage of not requiring the evaluation of each farm’s production potential, which means lower costs for the environmental and governmental agencies. Yet, it implies to control for the rate of soil degradation, which could be complex and costly as well.

Economic incentives and command-and-control approaches would need to be complemented by information-based instruments. Some of the farmers state that there is a lack of open discussion on the future of these soils and the future of their professions. Most of these farmers aim at passing their farm onto their children. More transparency on the diverse impacts associated with each management option is needed in communication with the diverse types of stakeholders concerned by the issue, and to society at a broader level.

3.2. Assessing the possibility of a sustainable use of organic soils among other actors

Chapter 1 highlights the necessity of a decision at a societal level on whether to preserve the organic soils in the study area. The experiments conducted in this thesis test the potential behavior of farmers under an agri-environmental payment promoting a sustainable use of organic soils. Farmers are the main managers of these soils; hence, they are key actors in the decision-making process. However, farmers have little opportunities to develop long-term planning that could lead to the reconsideration of management practices on these soils and the preservation of the soil production potential. This is mainly due to the fact that their livelihood depends on these soils. In addition to identifying policy schemes that would encourage farmers to adopt sustainable use of organic soils, it would be valuable to assess the level of acceptance of scenario 2 described in chapter 1 (i.e. whereby organic soils are rewetted) among other regional actors, such as policy makers, retailers and the general population. This would enable a more complete analysis of the issue. Society has an important role to play in launching a concrete discussion on the future management of these soils. This includes the elaboration of sustainable management practices and the development of appropriate solutions to transfer the volume of production generated by these soils. Economic valuation methods, such as contingent valuation or choice experiments, could be used to assess how the Swiss population values scenario 2 in comparison with the current situation. This would imply to present the different costs and benefits associated to each one of these scenarios, e.g. impacts on climate, the regional economy, soil productivity, and national imports. Moreover, this would allow analyzing societal preferences for the different types of ecosystem services provided by peatlands.

3.3. Conducting the experiment with a higher number of players per group

This thesis tested the potential behavior of farmers under alternative agri-environmental payment treatments using groups of two players. It would be of interest to conduct the experiment with larger group sizes, e.g. n H types and n L types within a same group with n > 2. However, this has several implications that need to be reflected: L-type players and H-type players, among each other, may have different social preferences. Before making any decision, they would therefore need to first agree among each other on the land use type and the side-payment offer to make to the other player-type. While such design would be more realistic as more representative of the community of farmers depending on a same drainage system, it would also make the interpretation of the results more difficult, with regard to disentangling the effects etc. A further option could be to model such a situation using the experimental results with two players per group, and different players’ profiles. 102

3.4. Non-binding side payments in the experimental setup

The experimental setup developed in this thesis allows for binding side-payment offers; the transfer of the side payment is conditional on the potential beneficiary adopting the sustainable land use. This implies that group members need to agree on rewetting the peat area for the transfer to potentially take place. Future research needs to test an alternative version of this experimental set up, which would be such that side-payments are conditional on the player’s vote for rewetting, and not on the land use choice. Thus, because rewetting is what L needs to adopt a sustainable land use, L could pay H for rewetting and reverting to activity B. This could be particularly useful in the baseline phase where L would “only” have to cover the costs of H’s drainage system. Yet, profit maximizers may still maximize their payoffs from vegetable production before undertaking this strategy, and thus deplete the peat. Moreover, depending on the level of the side payment made by L, such design may be problematic in presence of an agglomeration payment. H players may be tempted not to coordinate with L in the sustainable land use and still receive the side payment.

3.5. Further investigation of framing and subject-pool effects

With the treatments conducted in a static setting, the type of subject does not affect the ranking of payment schemes in their performance. However, the case examined in this thesis is not sufficient to draw a firm conclusion on this aspect. First, this study compared only three payment designs. A larger range of design options with stronger differences in payoff structure would be needed to test the impact of subject pool on policy ranking. Second, the study did not allow to compare the ranking of policy options in the dynamic experiment, which captures the real-world context more closely and exhibited stronger differences in behavior between subject pools. It would be particularly relevant to test the impact of subject pool on the ranking of alternative agri-environmental payments in their performance using the contextualized dynamic setting. In particular, it would be interesting to compare the effect of the variable agglomeration payment on the decision of farm apprentices to the ones of university students. For the purpose of this study, the subject-pool effect was investigated using the framed design, as it is the most relevant framing with respect to potential policy implications for the issue. While the introduction of a specific framing does not significantly affect the behavior of university students for the treatments tested, it would be highly interesting to test the framing effect on farm apprentices. It is likely that farm apprentices for whom the personal background is in line with the subject matter investigated would behave differently in an unframed experiment that does not contain any farming context. Furthermore, besides the relative importance to conduct economic experiments with the potentially concerned population, there is always a part of uncertainty with regard to how the sample population participating to the study resembles the real-world population. Hence, the importance of the sample size and the collection of socio-demographic characteristics.

3.6. Further analysis of behavioral factors

The effect of additional behavioral variables and personal characteristics on players’ decisions would need to be investigated. It would be relevant to test among university students the effect of the academic background on decisions. Economics students may behave more as payoff-maximizers than art students, for instance. Furthermore, some of the players’ characteristics collected in the survey and considered in the thesis would require a deeper understanding on how they could either favor or hinder a certain decision. This includes individual time preferences and inequality aversion. The former is very relevant 103 in the context investigated here, where current management practices affect the future payoffs of L players. Depending on their time preference, players may attach more or less value to future payoffs, which therefore affects their incentive to adopt the sustainable land use. Time preferences may also influence players for whom the future payoff from conventional land use is not negatively affected. In the presence of an agri-environmental payment maximizing their payoffs, players characterized by a high discount rate or present bias would value rapid coordination with their group member more than players who discount the future less. The format of the experiment investigated in this thesis (1-2 hours) makes it, however, difficult to capture the effect of time preferences and the fact that real time periods considered are years or decades. While payoffs to participants were made at the end of the experimental session, making payouts only after 1-2 weeks after the experiment can enable a higher level of appreciation of players’ time preferences. The second variable that necessitates further investigation is players’ aversion to payoff inequality. In the presence of a differentiated or variable payment, players’ inequality aversion appears to be an important explanatory factor for delays in cooperation and the bargaining process among group members. It would therefore be useful to further understand players’ level of inequality aversion and thus complement the SVO measure. One option is to estimate advantageous vs. disadvantageous inequality aversion among players (Blanco et al., 2011). This means to deeper investigate how players react to situations in which an inequitable distribution of resources favors their group member versus to situations in which an inequitable distribution favors themselves. Besides those general personal characteristics, the participant’s mood on the day of the experiment may also influence his/her decisions by affecting emotions (see e.g., Kirchsteiger et al., 2006; Mislin et al., 2015). This is especially important for the type of experiments investigated in this thesis given the asymmetry among players and the opportunity they are given to communicate and cooperate with each other. Despite the difficulty to assess this variable, asking participants to rate their mood using a Likert scale from “good” to “bad” before starting the experiment may be useful. In particular, H players with a rather bad mood in the dynamic setting may have a low incentive to cooperate with their group member despite the decline in payoffs of the latter in the absence of coordination. L players with a “bad” mood may show lack of interest for the experiment being conducted and may not care about the depletion of their soils and reduction of their farm profits. They may thus not put a lot of efforts in bargaining with their group members towards rewetting the soil. Such Likert scale measure could enable a better understanding of players’ behavior that do not match with behavioral predictions. Finally, based on an interview with three farmers working in the Seeland region, the effect of an agri- environmental program may be strongly dependent on farmers’ own personality, farm vision, and level of identification with being a producer. Part of the farmers may indeed be receptive to such a payment incentive while others may not accept it under any kind of circumstances.

3.7. Transferability of the experimental results

One of the main criticisms related to the use of economic experiments concerns the generalizability or external validity of the results and their representativeness to the subjects concerned by the issue (Harrison and List, 2004), here vegetable producers working on organic soils. According to Harrison and List (2004), the main factors that influence the external validity of experiments are the type of subject pool, the nature of the task and context that the subjects operate in, and the nature of the information and experience that the subjects bring to the task (e.g. players’ considerations). Apart from players’ background and the general strategy adopted in the experiment, we did not survey players’ considerations in the experiment. We discuss below the external validity of the results obtained in this thesis based on the first two factors and with regard to the actual policy implementation.

104

 Participants Contextualized experiments were conducted with farm apprentices whom the majority aims at becoming farmers and who show high opportunity costs of taking part to the experiment. While these apprentices do not possess the background of experienced farmers in the profession, their vision is interesting and represents future potential changes for the management of organic soils. A policy addressing this issue will indeed take a few years before being in place. In that sense, the experiment has also an education purpose on the question, and enables to raise awareness on the management of these soils. The specificity of the problem is such that the experiment would ideally need to be conducted with farmers operating on organic soils. This would considerably increase the external validity of the results. Farmers who are not familiar with the different implications of farming organic soils would, like farm apprentices, need explanations on the context prior to the experiment. On the other hand, most producers do not know their production potentials on farm lands. Like apprentices, they would have to accept the hypothetic profile (H, L farmer) without associating it to their own situation. Moreover, due to a high level of specialization, these “experienced” farmers may have difficulties fitting into the framework offered by the experiment. They could question the design and structure. Instead, farm apprentices may be more flexible in accepting the experimental situation, and willing to abstract from the “imperfections” of the experiment (e.g. non-representation of certain elements). In those respects, the observation of farm apprentices’ decisions seems valuable. Nevertheless, it would be important to test the various hypotheses tested in this thesis with experienced farmers. One difficulty for this concerns the involvement of a sufficient number of farmers.  Experimental design The experimental setting tested in this thesis attempts to integrate some of the core components of farmers’ decisions related to a change of management practice on the organic soils and therefore represents a relatively higher level of “realism” compared to most of the literature on economic experiments (see Reutemann et al., 2016 for an exception). Yet, the experiment is still a highly simplified version of the real-world decision situation. It offers a limited set of actions, it is conducted in a very short time (statements over long-time horizons are derived from a 2-hours interaction), and it abstracts from a number of factors present in the farmers’ actual environment. These factors include uncertainty and risk, e.g., regarding market prices, weather and extreme weather events, but also heterogeneities in both ecological features (e.g. thickness of peat layer, other ecosystem services linked to peatlands) and economic aspects among land users (e.g. farm sizes). Furthermore, while players face binomial decisions in the experiment (e.g. yes/no), real-world land owners’ decisions may rather be in a form of a continuum. Yet, the specificity of the structure and parameters used in the experiment increase the validity of the results for the case study itself and for situations with similar characteristics.  Payment schemes Coordination failure in a situation where players have a joint gain from coordinating is one of the major findings of this thesis. Several reasons were discussed in the previous sections to explain this finding in the context of the experiment. In reality, coordination failure could as well occur in several instances. Farmers may change their mind between the time of rewetting and the time of implementing the sustainable land use. Potential reasons include the discovery of cheaper or more innovative techniques to produce vegetables with lower rates of degradation, and the influence of a neighbor who did not sign up for rewetting. The implementation of an agglomeration payment may therefore require a procedure favoring commitment among farmers in their decisions. Differentiated payments revealed to be either less effective or as effective as uniform payment rates. Differentiated payments may, however, be relevant where policy makers want to highlight inequalities

105 among land users, e.g. at economic and environmental levels. A differentiated payment has the advantage to increase consciousness among producers on future production potentials, and to create awareness with regard to the farm land situation. This may be effective for short-term change of land management practices. Yet, this requires some care with regard to social norms: not all land users would like their situation to be emphasized. Uniform (average) payments are a relevant policy option where equality among farmers in benefitting from the agri-environmental payment is prioritized by policy makers. This may especially be the case in situations where economic inequalities among farmers are middle- or long-term based, like in this thesis. Indeed, at a specific time and for any positive amount of peat remaining, farmers face the same conservation costs. This is true with the assumption that farmers degrade the peat at the same rate. However, such uniform payment implies side payments among farmers, which stimulate farmers to equalize incentives. Side payments have the advantage to give a specific group of farmers room to influence the specific payment amounts resulting for each farmer, and to provide flexibility in the way to achieve the social optimal. Thus, each group, depending on its own dynamics and fairness preferences can find its own way to achieve the equilibrium. This aspect is important considering the complexity of the real context. With regard to the actual implementation of side payments, the current lack of public information on production potentials could make the contract negotiations between high- and low- production potential farmers difficult. Due to information asymmetries, farmers may indeed not reveal their actual production potential in order to benefit from a higher side payment from their neighbor. This issue could be overcome through intermediaries or mediators, but at the expense of transaction costs. The payment schemes tested in this thesis account for opportunity cost of farmers for changing land use. Conservation costs also include the technical cost of implementation of the sustainable land use, transaction costs involved in switching activities and enrolling into the PES program, and costs of transferring the volume of production. These additional costs need to be evaluated for an actual implementation of the program.  Social preferences This thesis reveals that social preferences are strong predictors of cooperation among players, and eventually, of the effectiveness of the agri-environmental payment schemes. However, the social preferences that were identified among the groups of participants may not be valid externally. Understanding social preference-types among the population of potential participants to the conservation program will be necessary in order to foresee the effectiveness of a particular payment scheme. Instead of an experiment, proxies of individual preferences could be collected using survey. This includes surveying, for instance, the drivers of farmers’ decisions (based on economic versus non- economic motivations) and the involvement of farmers in organizations (i.e. whether farmers use part of their time “for free” outside the farm management). To overcome the cost of information collection, a more realistic alternative consists of gathering this information among leaders of farming organizations.

Yet, the economic experiment described in this thesis enables a better understanding of the type of land use and level of side-payments that can be expected for a particular social-preference type. Collecting players’ social preferences and other individual characteristics enable to increase the validity of the results.

Experiments are recognized to be on average relevant tools to understand some of the underlying mechanisms of stakeholders’ behavior and to gain some insights from a particular situation (Herbst & Mas, 2015; Levitt & List, 2007). Given the complexity of the issue addressed in this thesis, for which no solution is being foreseen yet, the development of such an experiment represents one step towards the evaluation of the potential of specific policy options to address the issue. For an extension of this 106 work, an experimental setup including uncertainty on the farmers’ production potentials is planned by the author.

107

108

REFERENCES

A Porta, C. (2016): “Guten Boden effizient verbessern” (translation: “To improve the efficiency of good soils”), BauernZeitung, Nordwestschweiz Publiziert: 10.12.2016, http://www.bauernzeitung.ch/news-archiv/2016/guten-boden-effizient-verbessern/ accessed 16.01.2017. Abel, S. (2016): “Edible and medical plants from paludiculture”, In Wichtmann, W., Schröder, C., Joosten, H. (eds.): “Paludiculture – cultivation of wet peatlands. Climate protection, biodiversity, regional economic beneﬁts”, Schweizerbart Science Publishers, Stuttgart, 37-39. Abel, S., Couwenberg, J., Dahms, T., Joosten, H. (2013): “The database of Potential Paludicultural Plants (DPPP) and results for Western Pomerania”, Plant Diversity and Evolution 130: 219-229. AGRIDEA (2014): “PER – Romandie 2014 Prestations Ecologiques requises: Règles techniques Exploitations avec grandes cultures, production fourragère et cultures maraîchères” (translation: “Romandie 2014 Proof of Ecological Performance: Technical rules of farms with field crops, fodder production and market garden crops”), Groupe romand de coordination PER grandes cultures, fourragères, maraîchères et PIOCH 21.06.13/MAM/cfr/PER - ReglesTechniques-2014 https://www.agridea.ch/fileadmin/thematic/PER/2013-2014/site-per-2-01-maj130919-per- reglestechniques-2014-300dpl.pdf accessed 12.01.2017. Alix-Garcia, J., De Janvry, A., Sadoulet, E. (2008): “The role of deforestation risk and calibrated compensation in designing payments for environmental services”, Environmental and Development Economics 13: 375-394. Anderson, J., Burks, S. V., Carpenter, J., Götte, L., Maurer, M., Nosenzo, D., Potter, R., Rocha, R., Rustichini, A. (2013): “Self-selection and variations in the laboratory measurement of other- regarding preferences across subject pools: evidence from one college student and two adult samples”, Experimental Economics 16: 170-189. Armsworth, P. R., Szvetlana, A., Dallimer, M., Gaston, K. J., Hanley, N., Wilson, P. (2012): “The cost of policy simplification in conservation incentive programs”, Ecology Letters 15: 406-414. Bader, C., Müller, M., Schulin, R., Leifeld, J. (2017): “Amount and stability of recent and aged plant residues in degrading peatland soils”, Soil Biology and Biochemistry 109: 167-175. Bain, C., Joosten, H., Smith, P., Reed, M., Evans, E., Thompson, A., Coupar, A., Coath, M., (2012): “Kyoto Protocol and National Accounting for Peatlands”, IUCN UK Peatland Program Briefing January 2012 (revised). http://www.iucn-uk-peatlandprogramme.org/sites/www.iucn-uk- peatlandprogramme.org/files/Kyoto%20Protocol%20&%20National%20Accounting%20for%20Pe atlands%20revised%202012.pdf accessed May 2017. Bamière, L., David, M., Vermont, B. (2013): “Agri-environmental policies for biodiversity when the spatial pattern of the reserve matters”, Ecological Economics 85: 97-104. Banerjee, C. (2014): “Up, Up and Away! The Economics of Vertical Farming”, Journal of Agricultural Studies 2 (1): 40-60. Banerjee, S., Cason, T. N., de Vries, F. P., Hanley, N. (2014): “Transaction Costs and Spatial Coordination in Conservation Incentive Schemes: An Experimental Study”, submission BIOECON Conference, http://www.bioecon-network.org/pages/16th_2014/Cason.pdf accessed December 2016. Banerjee, S., Cason, T. N., De Vries, F. P., Hanley, N. (2015): “Coordination in Agglomeration Bonus Schemes with Transaction Costs and Communication: An Experiment Study”, University of St. Andrews, Discussion papers in Environmental Economics; Paper 2015-10.

109

Banerjee, S., De Vries, F. P, Hanley, N., Van Soest, D. (2013): “The Impact of Information Provision on Agglomeration Bonus Performance: An Experimental Study on Local Networks”, Stirling Economics Discussion Paper, 2013-09. Banerjee, S., De Vries, F. P., Hanley, N., van Soest, D. P. (2014): “The Impact of Information Provision on Agglomeration Bonus Performance: An Experimental Study on Local Networks”, American Journal of Agricultural Economics 96 (4): 1009-1029. Banerjee, S., Kwasnica, A. M., Shortle, J. S. (2011): “Agglomeration Bonus in Local Networks: A laboratory examination of spatial coordination failure”, Stirling Economics Discussion Paper 2011- 18 https://www.stir.ac.uk/media/schools/management/documents/workingpapers/SEDP-2011-18- Banerjee-Kwasnica-Shortle.pdf accessed November 2016. Banerjee, S., Kwasnica, A. M., Shortle, J. S. (2012): “Agglomeration bonus in small and large local networks: A laboratory examination of spatial coordination”, Ecological Economics 84: 142-152. Barrett, S. and Dannenberg, A. (2012): “Climate negotiations under scientiﬁc uncertainty”, PNAS 109 (43): 17372-17376. Baylis, K., Peplow, S., Rausser, G., Simon, L. (2008): “Agri-environmental policies in the EU and United States: A comparison”, Ecological Economics 65 (4): 753-764. Bell, A., Parkhurst, G., Droppelmann, K., Benton, T. G. (2016): “Scaling up pro-environmental agricultural practice using agglomeration payments: Proof of concept from an agent-based model”, Ecological Economics 126: 32-41. Bellemare, C., and Kröger, S. (2007): “On representative social capital”, European Economic Review 5 (1): 183-202. Belot, M., Duch, R., Miller, R. (2010): “Who Should be Called to the Lab? A Comprehensive Comparison of Students and Non-Students in Classic Experimental Games”, Nuffield center for experimental social sciences, Discussion paper series. Bernold, E., Gsottbauer, E., Ackermann, K. A., Murphy, R. O. (2015): “Social framing and cooperation: the roles and interaction of preferences and beliefs”, Available at SSRN: https://ssrn.com/abstract=2557927 or http://dx.doi.org/10.2139/ssrn.2557927 Beytes, C. (2011): “Whatever Floats Your Flowers”, Growertalks, http://www.ballpublishing.com/growertalks/ViewArticle.aspx?articleid=18299 accessed 08.03.2017. Biancalani, R. and Avagyan, A. (Eds.) (2014): “Towards climate-responsible peatlands management”, Mitigation of Climate Change in Agriculture, Series 9. FAO, Rome. Birrer, S., R. Bircher, V. Chevillat, R. Graf, D. Hagist, R. Home, M. Jenny, L. Pfiffner, S. Stöckli, B. Steinemann, G. Weidmann, J. Zellweger-Fischer (2015): “Mit Vielfalt punkten – Bauern beleben die Natur – Jahresbericht 2015” (translation: "With points for diversity - farmers revive nature - annual report 2015”), Schweizerische Vogelwarte, Sempach und Forschungsinstitut für biologischen Landbau (FiBL), Frick. Bonn, A., Allott, T., Evans, M., Joosten, H., Stoneman, R. (2016): “Peatland restoration and ecosystem Services - science, policy and practice”, Cambridge University Press. Bonn, A., Reed, M. S., Evans, C.D., Joosten, H. Bain, C., Farmer, J., Emmer, I., Couwenberg, J., Moxey, A., Artz, R., Tanneberger, F., von Unger, M., Smyth, M-A., Birnie, D. (2014): “Investing in nature: Developing ecosystem service markets for peatland restoration”, Ecosystem Services 9: 54-65. Bonn, A. (2012): “Peatland Code for the UK”, IUCN UK Peatland Programme, Launch of the global ‘Organic Soils and Peatlands Climate Change Mitigation Initiative’, 17 May 2012, Bonn Climate Change Conference, side event, Ministry of Environment. Burton, R. J. F., Kuczera, C., Schwarz, G. (2008): “Exploring Farmers’ Cultural Resistance to Voluntary Agri-environmental Schemes”, European Society for Rural Sociology 48 (1): 16-37.

110

Butorin, A. (2015a): “Das Grosse Moos verschwindet” (translation: “The Seeland disappears”), Berner Zeitung, http://www.bernerzeitung.ch/region/bern/Das-Grosse-Moos-verschwindet/story/30339490 accessed 16.01.2017. Butorin, A. (2015b): “Man kann das Seeland nur in Zentimetern retten” (translation: “the Seeland can only be saved in terms of centimeters”), Brennpunkt, Bieler Tagblatt Mittwoch, 23.09.2015, http://www.proagricultura.ch/sites/default/files/brennpunkt_grosses_moos_20150915.pdf accessed 16.01.2017. Cardenas, J-C., Janssen, M., Bousquet, F. (2013): “Dynamics of Rules and Resources: Three New Field Experiments on Water, Forests and Fisheries”, In: “Handbook on Experimental Economics and the Environment”: 319-345, Edward Elgar Publishing Ltd. Carpenter, J. and Seki, E. (2011): “Do social preferences increase productivity? Field experimental evidence from fishermen in Toyama Bay”, Economic Inquiry 49 (2): 612-630. CCM, Central Suisse de la Culture Maraîchère et des Cultures Spéciales (translation : Swiss center for vegetable production and special cultures) (2016): “Données sur la culture maraîchère suisse Aperçu du marché suisse des légumes en 2014” (translation: “Data on Swiss market gardening productions - Overview of the Swiss vegetable market in 2014”, http://www.gemuese.ch/Ressourcen-fr/PDF-fr/Politik-fr/CHGemuesebau_BROSCHURE_LID- fr.PDF. Chapman, S., Buttler, A., Francez, A-J., Laggoun-Défarge, F., Vasander, H., Schloter, M., Combe, J., Grosvernier, P., Harms, H., Epron, D., Gilbert, D., Mitchell, E. (2003): “Exploitation of northern peatlands and biodiversity maintenance: a conflict between economy and ecology”, Frontiers in Ecology and the Environment 1 (10): 525-532. Chevalley, M. (2012): “The fruit and vegetable market in Switzerland, Overview of the market and access information for international trading companies”, SWISSCOFEL. Cinyabuguma, M., Page, T., Putterman, L. (2005): “Cooperation under the threat of expulsion in a public goods experiment”, Journal of Public Economics 89: 1421-1435. Clymo, R. S. (1984): “The limit to peat growth”, Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 303 (1117): 605-654. Confédération Suisse (2016): “Food and Agriculture, Pocket statistics 2016”, Federal Statistics Office. 1112. 2600. Confédération-Suisse (2013): “Rapport Agricole 2013” (translation: “Agricultural Report 2013”), Confédération Suisse. Confédération-Suisse (2014): “Paiements directs aux exploitations suisses à l’année” (translation: “Annual direct payments to swiss farms”), Berne, Confédération Suisse. Confédération-Suisse (2015): “National Report on the implementation of the Ramsar Convention on Wetlands”, National Reports to be submitted to the 12 the Meeting of the Conference of the Contracting Parties, Uruguay, 2015 http://www.sib.admin.ch/fileadmin/chm- dateien/Nationalberichte/RAMSAR_CH_cop12_nrform_e_140804.pdf. Couwenberg, J., Thiele, A., Tanneberger, F., Augustin, J., Bärisch, S., Dubovik, D., Liashchynskaya, N., Michaelis, D., Minke, M., Skuratovich, A., Joosten, H. (2011): “Assessing greenhouse gas emissions from peatlands using vegetation as a proxy”, Hydrobiologia 674 (1): 67-89. Crisafulli, M. A., Von Holle, A. Bulik, C. M. (2008): “Attitudes Towards Anorexia Nervosa: The Impact of Framing on Blame and Stigma”, International Journal of Eating Disorders 41 (4): 333-339. Cronk, L. and Wasielewski, H. (2008): “An unfamiliar social norm rapidly produces framing effects in an economic game”, Journal of Evolutionary Psychology 6443 (4): 1789-2082. Cronk, L. (2007): “The influence of cultural framing on play in the trust game: a Maasai example”, Evolution and Human Behavior 28 (5): 352-358.

111

Dahms, T. and Schäfer, A. (2016): “Certiﬁcation of biomass from paludiculture”, In Wichtmann, W., Schröder, C., Joosten, H. (eds.): “Paludiculture - management of wet peatlands. Climate protection - biodiversity - regional added value”, 119-131. Danielson, A. J. and Holm, H. J. (2004): “Do you trust your brethren? Eliciting trust attitudes and trust behavior in a Tanzanian”, Journal of Economic Behavior and Organization 62 (2). Devetag, G. and Ortmann, A. (2007): “When and Why? A Critical Survey on Coordination Failure in the Laboratory”, Experimental Economics10 (3): 331-344. Drechsler, M. Wätzold, F., Johst, K., Shogren, J. F. (2007): “An agglomeration payment for cost- effective biodiversity conservation in spatially structured landscapes”, UFZ-Diskussionspapiere, No. 4/2007. Drechsler, M., Wätzold, F., Karin, J., Shogren, J. F. (2010): “An agglomeration payment for cost- effective biodiversity conservation in spatially structured landscapes”, Resource and Energy Economics 32 (2): 261-275. Drechsler, M., Smith, H. G. Sturm, A., Wätzold, F. (2016): “Cost-effectiveness of conservation payment schemes for species with different range sizes”, Research Note. Conservation Biology 30 (4): 894- 899. Druckman, J. N. (2001): “Evaluating framing effects”, Journal of Economic Psychology 22: 91-101. Dunn, C. and Freeman, F. (2011): “Peatlands: our greatest source of carbon credits?”, Carbon Management 2 (3): 289-301. Durlauf, L. E. and Blume, S. N. (2009): “Behavioural and Experimental Economics”, ISBN 978-0-230- 23868-8. Ellingsen, T., Johannesson, M., Mollerstrom, J., Munkhammar, S. (2011): “Social framing effects: Preferences or beliefs?”, Games and Economic Behavior 76 (1): 117-130. Elliott, C. S. and Hayward, D. M. (1998a): “The Expanding Definition of Framing and Its Particular Impact on Economic Experimentation”, Journal of Socio-Economics 27 (2): 229-243. Elliott, C. S., Hayward, D. M., Canon, S. (1998b): “Institutional framing: Some experimental evidence”, Journal of Economic Behavior and Organization 35 (4): 455-464. Engel, S. (2016): “The Devil in the Detail: A Practical Guide on Designing Payments for Environmental Services”, International Review of Environmental and Resource Economics 9: 131-177. Engel, S., Pagiola, S., Wunder, S. (2008): “Designing payments for environmental services in theory and practice: An overview of the issues”, Ecological Economics 65 (4): 663-674. Everett, K. R. (1983): “Histosols”, In Wilding, L. P., Smeck, N. E., Hall G. F. (eds.): “Pedogenesis and Soil Taxonomy, II”, The Soil Orders, 1-53. Exadaktylos, F., Espín, A. M., Brañas-Garza, P. (2013): “Experimental subjects are not different”, Scientific Reports 3 (1213). Ezzine de Blas, D., Wunder, S., Ruiz-Pérez, M., Moreno-Sanchez, RdP. (2016): “Global Patterns in the Implementation of Payments for Environmental Services”, PLoS ONE 11 (3). Falk, A., Meier, S., Zehnder, C. (2011): “Do Lab Experiments Misrepresent Social Preferences? The case of self-selected student samples”, Journal of the European Economic Association, https://www.cens.uni-bonn.de/team/board/armin-falk/do-lab-experiments-misrepresent-social- preferences-the-case-of-self-selected-student-samples-forthcoming-in-journal-of-the-european- economic-association-with-stephan-meier-and-christian-zehnder..pdf accessed May 2017. FAO, Food and Agricultural Organization (1988): “Nature and Management of tropical peat soils”, In FAO Soils Bulletin 59. Rome, by Andriesse J.P., Soil Resources, Management and Conservation Service, FAO Land and Water Development Division, http://www.fao.org/docrep/x5872e/x5872e00.htm accessed 16.01.2017. FAO, Food and Agricultural Organization (2005): “Materials for subsurface land drainage systems”, Irrigation and Drainage paper 60 (1).

112

FAO, Food and Agricultural Organization (2013): “FAO statistical yearbook 2013 – World Food and Agriculture”, http://www.fao.org/docrep/018/i3107e/i3107e.PDF accessed 16.01.2017. FAO, Food and Agricultural Organization (2014): “Towards climate-responsible peatlands management”, edited by Biancalani, R and Avagyan, A., http://www.fao.org/3/a-i4029e.pdf accessed 16.01.2017. FAO, Food and Agricultural Organization (2016): “Agricultural management”, http://www.fao.org/docrep/x5872e/x5872e0a.htm accessed 24.11.2016. Fehr, E. and Fischbacher, U. (2002): “Why social preferences matter – the Impact of non-selfish motives on competition, cooperation and incentives”, The Economic Journal 112: C1-C33. Fehr, E., Fischbacher, U., Gächter, S. (2002): “Strong reciprocity, human cooperation, and the enforcement of social norms”, Human Nature 13 (1): 1-25. Fehr, E. and List, J. A. (2004): “The hidden costs and returns of incentives – trust and trustworthiness among CEOs”, Journal of the European Economics Association 2 (5): 743-771. Ferraro, P. J. (2008): “Asymmetric information and contract design for payments for environmental services”, Ecological Economics 65: 810-821. FOAG, Federal Office for Agriculture (2004): “Swiss agricultural policy - Objectives, tools, prospect”, https://www.cbd.int/financial/pes/swiss-pesagriculturalpolicy.pdf, accessed 24.02.2017. FOAG, Federal Office for Agriculture (2017): “Prestations écologiques requises” (translation: “Proof of Ecological performance”), https://www.blw.admin.ch/blw/fr/home/instrumente/direktzahlungen/oekologischer -leistungsnachweis.html accessed 12.01.2017.; http://www.bafu.admin.ch/bodenschutz/15206/index.html?lang=fr accessed 12.01.2017. FOAG, Federal Office for Agriculture FOAG (2013): “Issues related to Agriculture”, September 2013. FOAG, Federal Office for Agriculture FOAG (2015): “Biodiversity for food and agriculture in Switzerland”, Abridged version and main findings of Switzerland’s Country Report on the State of Biodiversity for Food and Agriculture, Federal Office for Agriculture, Bern.. FOEN and FOAG (2013): “Protection des sols dans l’agriculture - Un module de l’aide à l’exécution pour la protection de l’environnement dans l’agriculture” (translation: “Soil Protection in Agriculture - A Module for help to the implementation of Environmental Protection in Agriculture”), http://www.bafu.admin.ch/publikationen/publikation/01721/index.html?lang=fr. accessed 12.01.2017. FOEN, Federal Office for the Environment (2002): “Les marais et leur protection en Suisse" (translation : “The wetlands and their protection in Switzerland”), Numéro DIV-8815-F, http://www.bafu.admin.ch/publikationen/publikation/00893/index.html?lang=fr, accessed 12.01.2017. FOEN, Federal Office for the Environment (2016a): “NIR, Swiss National Inventory Report (2016) - Switzerland’s Greenhouse Gas Inventory 1990-2014”, http://www.bafu.admin.ch/klima/13879/13880/15473/index.html?lang=en accessed 12.01.2016. FOEN, Federal Office for the Environment (2016b): “La future politique climatique de la Suisse en consultation” (translation: “Discussion around the future Swiss climate policy”, http://www.bafu.admin.ch/klima/13805/16344/16714/index.html?lang=fr accessed 17.01.2017. FOEN, Federal Office for the Environnent (2007): “État et évolution des marais en Suisse - Résultats du suivi de la protection des marais. État: Juin 2007”, (translation: “State and evolvement of peatlands in Switzerland. Results from the monitoring of peatland protection”). FOEN, Federal Office for the Environment (2014): “Régulation des lacs du pied du Jura” (translation: “Regulation of the lakes at the foot of the Jura”). FOEN, Federal Office for the Environnent (2015): “Sols et constructions - Etat de la technique et des pratiques” (translation: “Soil and constructions - State of the technique and practices”),

113

http://www.vd.ch/fileadmin/user_upload/themes/environnement/sol/fichiers_pdf/GEODE_SOLS_s ols_et_constructions.pdf accessed 12.01.2017. Fréchette, G. R. (2014): “Experimental Economics Across Subject Populations”, New York University. Fréchette, G.R. (2009): “Laboratory Experiments: Professionals versus Students”, New York University Frisch, D. (1993): “Reasons for framing effects”, Organizational Behavior and Human Decisional Processes 53: 399-429. FSO, Federal Statistical Office (2013): “Conditions d’exploitation, Landwirtschaftsbetriebe nach ihrer betriebswirtschaftlichen Ausrichtung, 1.5 Exploitations agricoles selon l’orientation technico- économique”, http://www.swissinfo.ch/fre/economie/Les_cooperatives,_un_capital_de_democratie.html?cid=318 90502, by Stefania Summermatter, 09 Janvier 2012 - 11:00. FSO, Federal Statistical Office (2004): “Inventaire fédéral des hauts-marais et des marais de transition d'importance nationa” (translation: “Federal Inventory of high peatlands and transition wetlands of national importance”), http://files.be.ch/bve/agi/geoportal/geo/lpi/HOCHMOOR_2003_01_LANG_FR.PDF accessed 12.01.2017 or https://www.admin.ch/opc/fr/classified-compilation/19910009/index.html. Gächter, S., Orzen, H., Renner, H, Starmer, C. (2009): “Are experimental economists prone to framing effects? A natural field experiment”, Journal of Economic Behavior and Organization 70: 443-446. Gaudig, G., Fengler, F., Krebs, M., Prager, A., Schulz, J., Wichmann, S., Joosten, H. (2013): “Sphagnum farming in Germany – a review of progress”, Mires and Peat 13 (08): 1-11. Glenk, K., Schaafsma, M., Moxey, A., Martin-Ortega, J., Hanley, N. (2014): “A framework for valuing spatially targeted peatland restoration”, Ecosystem Services 9: 20-33. Göttlich, Kh. and Kuntze, H. (1980): “Moorkultivierung, Nutzung und Verwendung in Land- und Forstwirtschaft” (translation: “The use of cultivated peatlands and their utilization in agriculture and forestry”), In Göttlich, Kh. (ed): “Moor- und Torfkunde”, Schweizerbar, Stuttgar, 231-248. Graves, A. R. and Morris, J. (2013): “Restoration of fen peatland under climate change”, Report to the Adaptation Sub-Committee of the Committee on Climate Change, School of Applies Sciences, Cranfield University, https://www.theccc.org.uk/wp-content/uploads/2013/07/Report-for-ASC- project_FINAL-9-July.pdf accessed 04.02.2017. Grob, N. (2012): “Zukunft der Bodennutzung im Berner Seeland aus der Sicht betroffener Landwirte” (translation: “The future of land use in the Bernese Seeland from the perspective of farmers”), Master of Science in Life Sciences / Applied Agricultural and Forestry Sciences, HAFL, Berner Fachhochschule. Gsottbauer, E. and Engel, S. (2016): “Coordination and Inequalities in Agglomeration Payments: An Experimental Analysis”, working paper. Günter, M., Schläpfer, F., Walter, T., Herzog, F. (2002): “Direct payments for biodiversity provided by Swiss farmers: an economic interpretation of direct democratic decision, case study Switzerland”, Environment Directorate, Environment Policy Committee, OECD, (ENV/EPOC/GEEI/BIO/(2001)9/FINAL). Harrison, G. W. and List, J. A. (2004): “Field Experiments”, Journal of Economic Literature 42 (4): 1009-1055. Hatala, J. A., Detto, M. Sonnentag, O., Deverel, S. J., Verfaillie, J., Baldocchi, D. D. (2012): “Greenhouse gas (CO2, CH4, H2O) ﬂuxes from drained and ﬂooded agricultural peatlands in the Sacramento–San Joaquin Delta”, Agriculture, Ecosystems & Environment 150: 1-18. Henrich, J., Heine, S. J., Norenzayan, A. (2010): “The weirdest people in the world?”, Behavioral and Brain Sciences 33: 61-135. Herbst, D. and Mas, A. (2015): “Peer effects on worker output in the laboratory generalize to the field”, Science (350, 6260): 545-549.

114

Hossain, T., List, J. A. (2012): “The Behavioralist Visits the Factory: Increasing Productivity Using Simple Framing Manipulations”, Management Science 58 (12): 2151-2167. Howley, P., Buckley, C., Donoghue, C. O., Ryan, M. (2015): “Explaining the economic ‘irrationality’ of farmers' land use behaviour: The role of productivist attitudes and non-pecuniary benefits”, Ecological Economics 109: 186-193. http://www.soilcarbon.org.uk/volume71.html http://www.spavetti.ch/spavetti_fr/seeland.html.accessed 16.01.2017. IPCC, Intergovernmental Panel on Climate Change (2014): “Climate Change 2014: Mitigation of Climate Change”, Contribution of Working Group III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Edenhofer, O., R. Pichs-Madruga, Y. Sokona, E. Farahani, S. Kadner, K. Seyboth, A. Adler, I. Baum, S. Brunner, P. Eickemeier, B. Kriemann, J. Savolainen, S. Schlömer, C. von Stechow, T. Zwickel and J.C. Minx (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, Chapter 11: Agriculture, Forestry and Other Land Use (AFOLU). IPCC, Intergovernmental Panel on Climate Change (2014): “Supplement to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories: Wetlands, Hiraishi, T., Krug, T., Tanabe, K., Srivastava, N., Baasansuren, J., Fukuda, M. and Troxler, T.G. (eds). Published: IPCC, Switzerland. IUCN (2015): “Peatland Code – UK Peatland code 1.0”, http://www.iucn-uk- peatlandprogramme.org/sites/www.iucn-uk- peatlandprogramme.org/files/Peatland%20Code%20v1.0.pdf accessed 17.03.2017. IUCN Peatland ecosystems: https://www.iucn.org/commissions/commission-ecosystem- management/our-work/cems-thematic-groups/peatland-ecosystems accessed 24.11.2016. Jahrl, I., Rudmann, C., Pfiffner, L., Balmer, O. (2012): “Motivations pour la réalisation de mesures de compensation écologique” (translation: “Motivations for the implementation of ecological compensation measures”), Recherche Agronomique Suisse 3 (4): 208-215. Janssen, M. A., Holahan, R., Lee, A., Ostrom, E. (2010): “Lab Experiments for the Study of Social- Ecological Systems”, Science: 328. Janssen, M. A., Lee, A., Waring, T. M. (2014): “Experimental platforms for behavioral experiments on social-ecological systems”, Ecology and Society 19 (4): 20. Jaton, J-F. (2005): “Gestion intégrée de la plaine de l’Orbe (GESORBE)” (translation: “Integrated management of the plain of Orbe”), Geomatique Suisse 10/2005: 557-558. Jauhiainen, J., Page, S. E., Vasander, H., (2016): “Greenhouse gas dynamics in degraded and restored tropical Peatlands”, Mires and Peat 17 (6): 1-12. Joerin, R. (2007): “The Swiss direct payment system and the concept of cross compliance”, ETH Zurich, Institute for Environmental Decisions (IED), paper presented at the University of Bergen, Department of Economics, September 7, 2007, https://www.ethz.ch/content/dam/ethz/special- interest/usys/ied/agricultural-economics- dam/documents/Robert%20J/Presentations/2007_Bergen_Swiss%20direct%20payments_Universit y%20of%20Bergen.pdf accessed 17.01.2017. Joosten, H. and Clark, D. (2002): “Wise use of mires and Peatlands – Background and principles including a framework for decision-making”, International Mire Conservation Group and International Peat Society, http://www.imcg.net/media/download_gallery/books/wump_wise_use_of_mires_and_peatlands_bo ok.pdf accessed 13.05.2017. Joosten, H. and Couwenberg, J. (2009): “Are Emission Reductions from Peatlands MRV-Able?”, Wetlands International, Ede. http://www.imcg.net/media/download_gallery/climate/joosten_couwenberg_2009.pdf accessed 20.03.2017.

115

Joosten, H. (2007): “The International Peat Society: fossil or renewable? An analysis of the IPS stand towards peat renewability and climate change”, International Mire Conservation Group (IMCG) newsletter issue 2-2007, http://www.imcg.net/media/newsletter/nl0702.pdf. Joosten, H. (2009): “The Global Peatland CO2 Picture. Peatland Status and Emissions in All Countries of the World”, Wetlands International, Ede, https://unfccc.int/files/kyoto_protocol/application/pdf/draftpeatlandco2report.pdf accesed 13.05.2017. Joosten, H. (2010): “The Global Peatland CO2 Picture, Peatland status and drainage associated emissions in all countries of the world”, Wetlands International, Ede, the Netherlands. Joosten, H., Brust, K., Couwenberg, J., Gerner, A., Holsten, B., Permien, T., Schäfer, A., Tanneberger, F., Trepel, M., Wahren, A. (2015): “MoorFutures® Integration of additional ecosystem services (including biodiversity) into carbon credits – standard, methodology and transferability to other regions”, BfN-Skripten 407, Federal Agency for Nature Conservation, http://www.moorfutures.de/app/download/21996565/Skript_407_barrierefrei.pdf accessed 04.02.2017. Joosten, H., Gaudig, G., Krawczynski, R., Tanneberger, F., Wichmann, S., Wichtmann, W. (2016): “Paludicultures: Sustainable productive use of wet and rewetted Peatlands”, In Bonn, A., Allott, T., Evans, M., Joosten, H., Stoneman, R. (eds.) (2016): “Peatland restoration and ecosystem services: Science, policy, and practice”, Cambridge University Press, Cambridge, 339-357. Joosten, H., Gaudig, G., Krawczynski, R., Tanneberger, F., Wichmann, S., Wichtmann, W. (2015): “Managing Soil Carbon in Europe: Paludicultures as a New Perspective for Peatlands”, In Banwart, S. A, Noellemeyer, E., Milne, E. (eds.): “Soil Carbon. Science, Management and Policy for Multiple Benefits”, SCOPE series 71 CAB International, Wallingford, ISBN: 9781780645322. Kasimir-Klemedtsson, A., Klemedtsson, L., Berglund, K., Martikainen, P., Silvola, J., Oenema, O. (1997): “Geenhouse gas emission from farmed organic soils – a review”, Soil Use and Management 13: 245-250. Kersten, U., Lindner, H., Melzer, R., Rehberg, U., Staak, R., Werner, W. (1999): “Ergebnisse des Projektes Regeneration und alternative Nutzung von Niedermoorﬂächen im Landkreis Ostvorpommern”, Kurzfassung. Stiftung Odermündung, Anklam. Kimbrough, E. O. and Sheremeta, R. M. (2013): “Side-Payments and the Costs of Conflict”, International Journal of Industrial Organization 31 (3): 278-286. Klaus, G. (réd.) (2007): “État et évolution des marais en Suisse. Résultats du suivi de la protection des marais” (translation : “State and evolvement of peatlands in Switzerland. Results from the monitoring of peatland protection”), État de l’environnement nº 0730. Office fédéral de l’environnement, Bern. Klaus, G., Künzle, I., Larcher, P., Pauli, D. (2007): “Biodiversité dans les zones humides, Biodiversité: dialogue entre recherche et pratiques, informations du forum biodiversité Suisse” (translation: “Biodiversity in wetlands, Biodiversity: dialogue between research and practices, information from the Swiss biodiversity forum”), HOTSPOT, Forum Biodiversité Suisse. Knowler, D. and Bradshaw, B. (2007): “Farmers’ adoption of conservation agriculture: A review and synthesis of recent research”, Food Policy 32: 25-48. Koneberg, C., Yaish, M., Stocké, V. (2010): “Norms and rationality in electoral participation and in the rescue of Jews in WWII: An application of the model of frame selection”, Rationality and Society 22: 3-36. Kotowski, W., Ackerman, M., Grootjans, A. P., Wheeler, B. (2016): “Restoration of temperate fens: matching strategies with site potential”, In “Peatland Restoration and Ecosystem Services: Science, Policy and Practice”, Chapter 10, Publisher: Cambridge University Press, Editors: Bonn, A., Allott, T., Evans, M., Joosten, H., Stoneman, R., 170-191.

116

Leifeld, J. (2013): “Prologue paper: Soil carbon losses from land-use change and the global agricultural greenhouse gas budget”, Science of the Total Environment 465: 3-6. Leifeld, J., Müller, M., Fuhrer, J., (2011): “Peatland subsidence and carbon loss from drained temperate fens”, Soil Use and Management 27: 170-176. Leifeld, J., von Lützow, M., (2014): “Chemical and microbial activation energies of soil organic matter decomposition”, Biology and Fertility of Soils 50: 147-153. Levin, I. P, Schneider, S. L., Gaeth, G. J. (1998): “All Frames Are Not Created Equal: A Typology and Critical Analysis of Framing Effects”, Organizational Behavior and Human Decision Processes 76 (2): 149-188. Levitt, S. D., and List, J. A. (2007): “What Do Laboratory Experiments Measuring Social Preferences Reveal about the Real World?”, The Journal of Economic Perspectives 21 (2): 153-174. Loisel, J., Yu, Z., Beilman, D. W., Camill, P., Alm, J, Amesbury, M. J., Anderson, D., Andersson, S., Bochicchio, C., Barber, K., et al. (2014): “A database and synthesis of northern peatland soil properties and Holocene carbon and nitrogen accumulation”, The Holocene 24 (9):1028-1042. Lunt, P., Allott, T., Anderson, P., Buckler, M., Coupar, A., Jones, P., Labadz, J., Worrall, P. (2010): “Impact of Peatland Restoration”, Scientific Review December 2010, IUCN peatland program. Lüscher, A. and Chappuis, P. L. (1984): “Le drainage du Plateau de Diesse 1918-1923 et sa reconstruction 1977-1984” (translation: “The drainage of the Plateau de Diesse 1918-1923 and its reconstruction 1977-1984”), Mensuration, Photogrammetrie, Génie rural 9 (84). Mann, S. (2005): “Farm Size Growth and Participation in Agri-environmental Schemes: A Configural Frequency Analysis of the Swiss Case”, Journal of Agricultural Economics 56 (3): 373-384. Mann, S., A. Breukers, J. Schweiger, G. Mack (2011): “Greenhouse vegetable production in the Netherlands and Switzerland – a grounded look at sector competitiveness”, Competitiveness Review 21 (4): 339-351. Masiliunas, A. (2016): “Overcoming Coordination Failure in a Critical Mass Game: Strategic Motives and Action Disclosure”, AMSE Working Papers, WP 2016 (9), https://www.amse- aixmarseille.fr/sites/default/files/_dt/2012/wp_2016_-_nr_09.pdf accessed 14.07.2017. MEA, Millennium Ecosystem Assessment (2005): “Ecosystems and human wellbeing: wetland and water”, http://www.millenniumassessment.org/documents/document.358.aspx.pdf accessed 16.01.2017. Meier, S. (2006): “Does framing matter for conditional cooperation? Evidence from a natural field experiment”, Contributions to Economic Analysis & Policy 5 (2). Meyers-Levy, J. and Maheswaran, D. (2009): “Exploring Message Framing Outcomes When Systematic, Heuristic, or Both Types of Processing Occur”, Journal of Consumer Psychology 14 (1 and 2): 159-167. Miéville-Ott, V., Praz, P. (2004): “Logiques identitaires et stratégies des agriculteurs dans le processus d’aménagement de la plaine de l’Orbe” (translation: “Logics of identity and strategies of the farmers in the process of development of the plain of the Orbe”), Economie et Sociologie Rurales 2 (4): 135-152. Milinski, M., Semmann, D., Krambeck, H-J., Marotzke, J. (2006): “Stabilizing the Earth’s climate is not a losing game: Supporting evidence from public goods experiments”, PNAS 103 (11): 3994- 3998. Milinski, M., Sommerfeld, R. D., Krambeck, H-J., Reed, F. A., Marotzke, J. (2008): “The collective- risk social dilemma and the prevention of simulated dangerous climate change”, PNAS 105 (7): 2291-2294. Möhring, A., Mack, G., Willersinn, C. (2012): “Gemüseanbau – Modellierung der Heterogenität und Intensität”, Agrarforschung Schweiz 3 (7–8): 382–389.

117

Muller, A., Bader, C., Ferré, M., Engel, S., Gattinger, A., Holzkämper, A., Huber, R., Müller, M., Six, J. (2016): “Can soil-less crop production be a sustainable option for soil conservation and future agriculture?”, working paper, submitted to Land Use Policy. Murphy, R. O., Ackermann, K. A., Handgraaf, M. J. J. (2011): “Measuring Social Value Orientation”, Judgment and Decision Making 6 (8): 771-781. Musarika, S., Atherton, C. E., Gomersall, T., Wells, M. J., Kaduk, J., Cumming, A. M. J., Page, S. E., Oechel, W. C., Zona, D. (2017): “Effect of water table management and elevated CO2 on radish productivity and on CH4 and CO2 fluxes from peatlands converted to agriculture”, The Science of the Total Environment (584): 665-672. OECD, Organization for Economic Co-operation and Development (2010): “Environmental Cross Compliance in Agriculture”, http://www.oecd.org/tad/sustainable-agriculture/44737935.pdf accessed 17.01.2017. OED, Office des Eaux et des Déchets (English translation : Office of Water and Waste) (2013): “Barrage de régulation de Port - L’ouvrage majeur de la correction des eaux du Jura" (translation : “The port dam regulation - The major structure of the Jura water correction”, Direction des travaux publics, https://www.bve.be.ch/bve/fr/index/wasser/wasser/wasserregulierung/RegulierwehrPort.assetref/da m/documents/BVE/AWA/fr/GWR_JGK/wwa_barrage_regulation_de_port.pdf, accessed 24.02.2017. Oleszczuk, R., Regina, K., Szajdak, L., Höper, H., Maryganova, V. (2008): “Impacts of agricultural utilization of peat soils on the greenhouse gas balance”, In Strack, M. (ed): “Peatlands and Climate change”, Chapter 3, IPS, International Peat Society, http://www.peatsociety.org/sites/default/files/files/PeatlandsandClimateChangeBookIPS2008.pdf accessed 16.01.2017. Osterberg, B., (1999). “Agri-environment programmes in Germany - implementation, acceptance and aspects of their evaluation”, 1st Workshop on the Management and Monitoring of Agri- Environment Schemes Joint Research Centre (JRC) Ispra. Pagiola, S., Arcenas, A., Platais, G. (2005): “Can Payments for Environmental Services Help Reduce Poverty? An Exploration of the Issues and the Evidence to Date from Latin America”, World Development 33 (2): 237-253. Parish, F., Sirin, A., Charman, D., Joosten, H., Minayeva, T., Silvius, M. and Stringer, L. (Eds.) (2008): “Assessment on Peatlands, Biodiversity and Climate Change: Main Report”, Global Environment Centre, Kuala Lumpur and Wetlands International, Wageningen. Parkhurst, G. M. and Shogren, J. F. (2007): “Spatial incentives to coordinate contiguous habitat”, Ecological Economics 64 (2): 344-355. Parkhurst, G. M. and Shogren, J. F. (2008): “Smart subsidies for conservation”, American Journal of Agricultural Economics 9: 1192-1200. Parkhurst, G. M., Shogren, J. F., Bastian, C., Kivi, P., Donner, J., Smith, R. B. W. (2002): “Agglomeration bonus: an incentive mechanism to reunite fragmented habitat for biodiversity conservation”, Ecological Economics 41: 305-328. Pascual, U., Muradian, R., Rodríguez, L. C., Duraiappah, A. (2010): “Exploring the links between equity and efficiency in payments for environmental services: A conceptual approach”, Ecological Economics 69: 1237-1244. Perman, R., Ma, Y., McGilvray, J., Common, M. (2003): “Natural Resource and Environmental Economics”, Third Edition, Chapter 7. Pick-up (2003), magazine, http://nomadsystems2.net/Documents/AgriInfo/pick_up_magazine_11.pdf accessed 16.01.2017.

118

Pillutla, M. M., Chen, X. (1999): “Social norms and cooperation in social dilemmas: the effects of context and feedback”, Organizational Behavior and Human Decision Processes 78 (2): 81-103. Prasuhn, V., Liniger, H., Gisler, S., Herweg, K., Candinas, A., Clément, J-P. (2013):“A high- resolution soil erosion risk map of Switzerland as strategic policy support system”, Land Use Policy 32: 281-291. Reed, M. Buckmaster, S., Moxey, A., Keenleyside, C., Fazey, I., Scott, A., Thomson, K., Simon Thorp, S., Anderson, R., Bateman, I., Bryce, B., Christie, M., Glass, J., Hubacek, K., Quinn, C., Maffey, G., Midgely, A., Robinson, G., Stringer, L., Lowe, P., Slee, B. (2010): “Policy Options for Sustainable Management of UK Peatlands”, draft scientific review.http://www.iucn-uk- peatlandprogramme.org/sites/www.iucn-uk- peatlandprogramme.org/files/Review%208%20Policy%20Options_0.pdf accessed 17.11.2016. Rege, M. and Telle, K. (2004): “The impact of social approval and framing on cooperation in public good situations”, Journal of Public Economics 88: 1625-1644. Regina, K., Budiman, A., Greve, M. H., Grønlund, A., Kasimir, A., Lehtonen, H., Petersen, S. O., Smith, P., Wösten, H. (2016): “GHG mitigation of agricultural peatlands requires coherent policies”, Climate Policy 16 (4): 522-541. Reutemann, T. and Engel, S. (2014): “Return per hectare or per dollar? Opportunity costs of conservation under capital constraints and intensification potential”, Working paper, Professorship of Environmental Policy and Economics, ETH Zurich. Reutemann, T., Engel, S., Pareja, E. (2016): “How (not) to Pay - Field Experimental Evidence on the Design of REDD+ Payments”, Ecological Economics 129: 220-229. Röder, N. and Osterburg, B. (2012): “Reducing GHG emissions by abandoning agricultural land use on organic soils - A cost assessment”, Selected Paper prepared for presentation at the International Association of Agricultural Economists (IAAE) Triennial Conference, Foz do Iguaçu, Brazil, August, 2012. Roulet, N. T. (2000): “Peatlands, carbon storage, greenhouse gases, and the Kyoto Protocol: Prospects and significance for Canada”, Wetlands 20 (4): 605-615. RTS (2011): “L'agriculture suisse continue à perdre des exploitations". http://www.rts.ch/info/suisse/4180766-l-agriculture-suisse-continue-a-perdre-des-exploitations- .html. Rucker, D. D., Petty, R. E., Briñol, P. (2008): “What's in a frame anyway?: A meta-cognitive analysis of the impact of one versus two sided message framing on attitude certainty”, Journal of Consumer Psychology 18: 137-149. Ryan, R. L., Erickson, D. L., De Young, R. (2003): “Farmers' Motivations for Adopting Conservation Practices along Riparian Zones in a Mid-western Agricultural Watershed”, Journal of Environmental Planning and Management 46 (1): 19-37. Rydin, H. and Jeglum, J. K. (2013): “The Biology of Peatlands”, 2e, Book, Oxford University Press. Schader, C., Sanders, J., Offermann, F., Lampkin, N., Stolze, M. (2008): “Cost-Effectiveness Evaluation of Swiss Agri-Environmental Measures on Sector Level”, Paper prepared for presentation at the 107th EAAE Seminar “Modelling of Agricultural and Rural Development Policies”, Sevilla, Spain, 2008. Schäfer, A. and Joosten, H. (eds.) (2005): “Erlenaufforstung auf wieder vernässten Niedermooren - ALNUS-Leitfaden” (translation: “Afforestation with alder on rewetted fens - ALNUS guide”), DUENE e.V., Greifswald. Schaller, L. (2014): “Landwirtschaftliche Nutzung von Moorflächen in Deutschland – Sozioökonomische Aspekte einer klimaschonenden Bewitschaftung” (translation: “Agricultural use of peatlands in Germany – socio-economic aspects of climate-friendly farming”), PhD Dissertation,

119

Fakultät Wissenschaftszentrum, Weihenstephan für Ernährung, Landnutzung, und Umwelt der Technischen Universität München. Scharlemann, J. P. W., Tanner, E. V. J., Hiederer, R., Kapos, V. (2014): “Global soil carbon: understanding and managing the largest terrestrial carbon pool”, Carbon Management 5 (1): 81-91. Schrier-Uijl, A. P., Kroon, P. S., Hendriks, D. M. D., Hensen, A., Van Huissteden, J., Berendse, F., Veenendaal, E. M. (2014): “Agricultural peatlands: towards a greenhouse gas sink – a synthesis of a Dutch landscape study”, Biogeosciences 11: 4559-4576. Schumann, M. and Joosten, H. (2008): “Global Peatland Restoration Manual. Institute of Botany and Landscape Ecology”, Greifswald University. Seiler, D. (2014): “Protection of Agricultural Peatlands in Switzerland – a comparative exploration of potential agricultural peatland protection tools for Switzerland with regard to relevant case studies”, Internship report, Institute for Environmental Decisions, ETH Zurich. Serrano, R. and Feldman, A. M. (vc 2010, 2011): Lesson 14 Game Theory. Soil Science Society of Switzerland (SSSS) (2015): “Le sol marécageux” (translation: “The marshy soil”), http://www.boden-des- jahres.ch/fileadmin/Bilder/Downloads/BGS_Flyer_Le_sol_marecageux.pdf accessed 12.01.2017. Spavetti (2011): http://www.spavetti.ch/spavetti_fr/seeland.html accessed 12.01.2017. Sprecher, S. (2015): “Gemüsebauern existenziell gefährdet” (translation: “The existence of vegetable farmers in danger”), Freiburger Nachrichten, http://www.freiburger-nachrichten.ch/nachrichten- see/gemuesebauern-existenziell-gefaehrdet accessed 16.01.2016. Sweers, W., Horn, S., Grenzdörffer, G., Müller, J. (2013): “Regulation of reed (Phragmites australis) by water buffalo grazing: use in coastal conservation”, Mires and Peat 13 (3). Sweers, W., Möhring, T., Müller, J. (2014): “The economics of water buffalo (Bubalis bubalis) breeding, rearing and direct marketing”, Archiv für Tierzucht 57: 1-11. Swisscofel (2014): Fédérations et organisations de la branche maraîchère, FSO, (translation: “Federations and organizations of the market garden sector”), http://www.swisscofel.ch/wAssets/docs/Wissen/Le_marche_Suisse_des_legumes_2014_Poster.pdf accessed 18.11.2016. Talbot, D. (2007): “Saving Holland”, Sustainable Energy, MIT technology view, https://www.technologyreview.com/s/408182/saving-holland/ accessed 08.03.2017. Tavoni, A., Dannenberg, A., Kallis, G., Löschel, A. (2011): “Inequality, communication, and the avoidance of disastrous climate change in a public goods game”, PNAS 108 (29): 11825-11829. Tversky, A. and Kahneman, D. (1981): “The framing of decisions and the psychology of choice”, Science 211: 453-458. UMS, Union Maraîchère Suisse (translation : Swiss Union of Vegetable Producers) (2017): “Fournisseurs et acheteurs”, http://www.gemuese.ch/Dossiers-fr/Marche-suisse-des- legumes/Fournisseurs-acheteurs accessed 06.03.20017. USP, Union Suisse des Paysans (in English: Swiss farmers’ union) (2015): “Rapport de Situation 2014” (translation: “Progress report”), http://www.sbv- usp.ch/fileadmin/sbvuspch/05_Publikationen/Situationsberichte/150106_USP_Rapport_de_situatio n.pdf accessed 12.01.2017. Van Beukering, P., Schaafsma, M., Davies, O., Oskolokaite, I. (2008): “The economic value of peatland resources within the Central Kalimantan Peatland Project in Indonesia – Perception of local communities”, Report E-08/05, Wetlands International and the University Palangkaraya under the Central Kalimantan Peatlands Project (CKPP). Van Soest, D., Dijk, J., Arguedas, C. (2013): “Preference heterogeneity, asymmetric information, and the inevitability of informational rents in environmental subsidy programs”,

120

http://www.lse.ac.uk/GranthamInstitute/wp- content/uploads/2014/10/grantham_vansoest_2015_01_07.pdf accessed 08.03.2016. Verein Seeland.biel/bienne (2016): “Finanzierungsbeitrag Bodenkartierung Grosses Moos” (translation: Contribution to the cartography of the Seeland), Mitgliederversammlung vom 8. Juni 2016, Traktandum, http://www.seeland-biel- bienne.ch/images/Pdf/1500_23_bodenkartierung_grosses_moos_160608.pdf accessed 16.01.2017. Verheijen, F. G. A., Jones, R. J. A., Rickson, R. J., Smith, C. J. (2009): “Tolerable versus actual soil erosion rates in Europe”, Earth Science Reviews 94 (1): 23-38 Verhoeven, J. T. A. and Setter, T. L. (2010): “Agricultural use of wetlands: opportunities and limitations”, Annals of Botany 105: 155-163. Wätzold and Dreshler (2014): “Agglomeration payment, agglomeration bonus or homogeneous payment?”, Resource and Energy Economics 37: 85-101. WEC, World Energy Council (2013): “World Energy Resources: Peat”, https://www.worldenergy.org/wp-content/uploads/2013/09/Complete_WER_2013_Survey.pdf. Wichmann, S. and Köbbing, J. F. (2015): “Common reed for thatching—A first review of the European market”, Industrial Crops and Products 77 (23): 1063-1073. Wichmann, S. (2016): “Commercial viability of paludiculture: A comparison of harvesting reeds for biogas production, direct combustion, and thatching”, Ecological Engineering, In Press. Wichmann, S., Brander, L., Schäfer, A., Schaafsma, M., van Beukering, P., Tinch, D., Bonn, A. (2016): “Valuing peatland ecosystem services”, In Bonn, A., Allott, T., Evans, M., Joosten, H., Stoneman, R. (eds.): “Peatland restoration and ecosystem services: Science, policy, and practice”, Cambridge University Press, Cambridge, 314-338. Wichtmann, W. and Schäfer, A. (2007): “Alternative management options for degraded fens – Utilisation of biomass from rewetted peatlands, In Okruszko, T., Maltby, E., Szatylowicz, J. Swiatek, D., W. Kotowski (2007): “Wetlands: Monitoring, Modeling and Management”, Taylor and Francis, Leiden, the Netherlands, 273-279. Wichtmann, W. (2011): “Land use options for rewetted peatlands - Biomass use for food and fodder”, In Tanneberger F, Wichtmann W (eds.): (2011): “Carbon credits from peatland rewetting”, Schweizerbart Science Publishers, Stuttgart, Germany, 110-113. Wichtmann, W., Schröder, C., Joosten, H. (2016): “Paludiculture – cultivation of wet peatlands. Climate protection, biodiversity, regional economic beneﬁts”, Schweizerbart Science Publishers. Wilson, L., Wilson, J., Holden, J., Johnstone, I., Armstrong, A., Morris, M. (2011): “Ditch blocking, water chemistry and organic carbon flux: Evidence that blanket bog restoration reduces erosion and fluvial carbon loss”, Science of the Total Environment 409 (11): 2010-2018. World Energy Resources: Peat World Energy Council 2013, https://www.worldenergy.org/wp- content/uploads/2013/10/WER_2013_6_Peat.pdf accessed 20.03.2017 Wunder, S. (2005): “Payments for Environmental Services: Some Nuts and Bolts”, Occasional Paper No. 42, CIFOR, Bogor. Wunder, S. (2007): “The Efficiency of Payments for Environmental Services in Tropical Conservation”, Conservation Biology 21 (1): 48-58. Wunder, S., Engel, S., Pagiola, S. (2008): “Taking stock: A comparative analysis of payments for environmental services programs in developed and developing countries”, Ecological Economics 65: 834-852. Wünscher, T., Engel, S, Wunder, S. (2008): “A tool for boosting conservation benefits”, Ecological Economics 65: 822-833. Wüst-Galley, C. and Leifeld, J. (2017): “Entwicklung der Kohlenstoffvorräte und Treibhausgasemissionen der Moorböden seit 1850”, In Stuber, M., Müller, M., Bürgi, M. (Eds.):

121

“Vom “eroberten Land” zum Renaturierungsprojekt – Geschichte der Feuchtgebiete in der Schweiz seit 1700”, Paul Haupt Verlag in Bern, accepted. Wüst-Galley, C., Grünig, A., Leifeld, J. (2015): “Locating Organic Soils for the Swiss Greenhouse Gas Inventory”, Environment Agroscope Science (26). Xintu, L. (2009): “Conditions of peat formation”, In “Coal, oil shale, natural bitumen, heavy oil and peat”, http://www.eolss.net/ebooklib/ebookcontents/E3-04-ThemeContents.pdf accessed 13.05.2017. Yu, Z., Loisel, J., Brosseau, D. P, Beilman, D. W., Hunt, S. J. (2010): “Global peatland dynamics since the Last Glacial Maximum”, Geophysical Research Letters 37 (L13402). Zeitz, J. (2016): “Sanddeckkultur (sand cover cultivation)”, In Wichtmann, W., Schröder, C., Joosten, H. (eds.): “Paludiculture – cultivation of wet peatlands. Climate protection, biodiversity, regional economic beneﬁts”, Schweizerbart Science Publishers, 5-6. Zollinger (2006): “Drainagen: Unterhalt bis in alle Ewigkeit?”, Améliorations structurelles/Génie rural, Géomatique Suisse 12, http://www.geomatik.ch/fileadmin/redaktion/download/2006/Fach/FA_12_2006_6.pdf accessed 12.01.2017.

122

APPENDICES

1. Appendix A – Appendix to chapter 1

The regional experts who were interviewed and contributed to this chapter are: - Nathalie Grob As part of a master project “future of land use in the Bernese Seeland from the perspective of farmers” (translated from german: “Zukunft der Bodennutzung im Berner Seeland aus der Sicht betroffener Landwirte”), Nathalie Grob conducted a study in the Müntschemier region and surveyed a sample of 14 farmers working on organic soils (12 vegetable producers and 1 dairy producer). - Elena Havlicek Federal office for the Environment (FOEN), Soil and Biotechnology Division, Responsibilities: soil protection for projects subject to an EIA, soil biology, soil biodiversity, Bern. - Jean-François Jaton Institute of Rural Engineering, Hydrology and Development (translated from french: Institut de génie rural, hydrologie et aménagements), EPFL Lausanne. Head of the GESORBE project “Integrated management of the Orbe plain” (translated from French: “Gestion intégrée de la plaine de l’Orbe”), Chief of the water, soils, and sanitation service (translated from French: Service des eaux, sols et assainissement, SESA), in Vaud Canton, formerly head of the land improvement service (Service des améliorations foncières, SAF). - Martin Lichtenhann Terraviva ag/sa – organic farmer company, Innovation and Development. Formerly at FiBL, Frick (Research institute of organic Agriculture). - Albert Lüscher President of the “Network and Biotope foundation of the Seeland”; (translated from the French “Fondation Réseau des biotopes du Grand Marais”). Formerly: Lüscher and Aeschlimann AG, engineering and survey (translated from the French “Bureau d’ingénieurs et de géomètres officiels”), including the implementation of drainage systems and soil improvements in the Seeland region. - Stefan Mann Agroscope, Socioeconomics Research Group, Tänikon, Switzerland. Involved in the design and evaluation of agricultural policies in Switzerland. - Moritz Müller, Soil scientist, Professor at HAFL (School of Agricultural, Forest, and Food sciences), Zollikofen, Switzerland. - Peter Trachsel Former head of the Witzwill farm domain (700ha) that is located in Seeland on organic soils. Currently: INFORAMA in Bern, Zollikofen, domain of soil protection. - Three farmers from the plain of Orbe: Jaques-Yves Deriaz, now working at the land melioration service (Service des Améliorations Foncières, SAF); Willy Stoll (300 ha including 250 ha on organic soils); Mr. Egger (240ha of cultivated lands including 108 ha on organic soils).

123

2. Appendix B – Appendix to chapter 2

B1. Screen shot of the experiment

Figure B1: screen shot of the experiment

B2. Experimental instructions – Baseline

This experiment consists of Part 2 and Part 3, which are independent. Each part comprises 10 rounds of a decision situation. This document explains Part 2. You are randomly assigned to a group with one other participant. This participant can be any other participant present here today. Any interactions in this study are anonymous: you will never know the identity of this other participant. You will be assigned to another participant in Part 3. At the end of the experiment, one round will randomly be picked from this part and your earning in points from this round will be exchanged into CHF at a rate of 100 points = 0.40 CHF. Before starting Part 2, there will be a learning round, where the decisions you will make will not be recorded and will therefore not determine your end payment. Note that there are no right or wrong answers in your choices and actions; this is only a matter of preferences.

Background:

Peat soils (Torf / organic soils) consist of a first layer of peat and an underneath mineral soil layer (Figure B2.1). The peat is very good for vegetable production. The peat layer is very rich in organic matter (and thus in carbon) which accumulated in high water level conditions. Since a few decades, the water table (or groundwater) is artificially maintained at a low level by a drainage system in order to enable agricultural activities on peat soils. This drainage leads to the reduction of the peat soil layer, in particular, because of carbon emissions. Depending on the location, the quality of this underneath mineral soil layer varies.

124

Peat soil layer

Underneath mineral soil layer variable: either suitable to vegetable production, or not.

Figure B2.1: Peat soil profile One of the most effective ways to avoid the loss of peat soils is to rewet them, so to stop the drainage. The drainage systems are regulated at a community level. Therefore stopping the drainage system needs to be decided unanimously. The water table will therefore only be increased if all farmers agree on it (unanimous decision).

Overview of the experiment:

You will take the role of a farmer: you and your group member are vegetable producers. The group you are assigned to will constitute of one Yellow-farmer and one Blue-farmer. You will be randomly selected to be one or the other and you will keep this profile in all rounds, and in Part 3. Yellow and Blue differ in the type of the underneath mineral soil layer of their peat soils. - Yellow has an underneath mineral soil layer which is not suitable for vegetable production; - Blue has an underneath mineral soil layer which is suitable for vegetable production. Blue and Yellow therefore have different farm profits. Each round follows the same pattern: You first need to decide whether to rewet the peat soils of your land or not to do so. They will only be rewetted if both of you vote for it.  If at least one of you rejects to stop the drainage system and thus to increase the water table, you keep producing vegetables. You will be directed to next round where you will be asked the same question.

 If both of you vote for increasing the water table, peat soils of both members are rewetted. You and your group member need then to decide between two types of land use on these rewetted peat soils: - A) Adopting another land use (e.g. crops or livestock adapted to high water level conditions), your farm profit will then be lower than with vegetables, or - B) Pursuing vegetable production by implementing a personal local drainage system on peat soils at a cost to you. You will then be directed to next round where you will be asked the same questions. However, before these choices are made, Yellow may offer a binding amount to be transferred by Yellow to Blue if and only if Blue chooses A. Blue (at the same time) may offer a binding transfer- payment to be transferred by Blue to Yellow if and only if Yellow chooses A. Note that transfer-payment offers have the effect of changing the total payoffs. Whatever amount you state will be transferred to the other farmer if he/she plays A; this money will be transferred regardless of your own choice of land use.

Procedure:

The general procedure of one round and the effect of your actions on your profit are as follows: 125

1. Before making a choice in a round, you can first test how a transfer-payment offer to your group member would change your payoffs (about 3 minutes). 2. Each of you can make an offer for a transfer-payment that you would pay to the other player in case he/she chooses option A. Whatever offer you make, it is binding. That is, the offered amount will be transferred from you to the other player automatically if the other player chooses option A. 3. Both members Yellow and Blue are asked to vote on increasing the water table for their group, or not. There are 2 possible outcomes: o If at least one of you rejects to increase the water table, the water table is not increased and peat soils are not rewetted. Both members continue with vegetable farming on peat soils. Yellow-farmer’s and Blue-farmer’s farm profits are respectively 2500 and 4500, as in the table below. Yellow-farmer’s farm profit is lower than Blue-farmer’s because of the unsuitability of his/her underneath mineral soils for vegetable production. No transfer- payments are made in this case because both of you automatically choose vegetable production.

Blue Farmer Does not increase the water table: continues farming vegetables on the peat soils Does not increase the water table: continues farming 2500 Yellow Farmer vegetables on the peat soils 4500 o If both of you vote for increasing the water table, peat soils of both members are rewetted. In that case, you need to choose between: - A) Dealing with the high water table and establishing a different land use (your farm profit will then be 250) and - B) Installing a personal drainage system on your peat soils at your own cost (the cost is 220) and continue producing vegetables. If no transfers are made, the actions’ payoffs are as below.

Please note that the yellow numbers represent what the Yellow Farmer gets and the blue numbers represent what the Blue farmer gets. These profits apply if the farmers have taken the corresponding actions, which are read on the left side of the box for the Yellow farmer and on the upper side for the Blue farmer. For instance, if both of you choose to rewet and then choose A, that means both of you earn 250. As another example, if both of you choose to rewet and then Yellow Farmer chooses A and Blue farmer chooses B, Yellow Farmer earns 250 and Blue farmer earns 4280.

Blue Farmer Adopt another land use on Pursue vegetable production (B) rewetted peat soils (A) Adopt another land use on 250 250 Yellow rewetted peat soils (A) 250 4500 – 220 = 4280 Farmer 2500 – 220 = 2280 2500 – 220 = 2280 Pursue vegetable production (B) 250 4500 – 220 = 4280

After all have made their choices, your screen will display your choice and profit for this round, as well as the ones of your other group member. This round will then be repeated 10 times.

Examples of transfer-payments:

Suppose for example, that Yellow offers to pay 25 to Blue if Blue plays A. Then the action’s payoffs become (transfer-payments and their effects are put in bold in the table):

126

Blue Farmer Adopt another land use on Pursue vegetable production (B) rewetted peat soil (A) Adopt another land use on 250 - 25 = 225 250 Yellow rewetted peat soils (A) 250 + 25 = 275 4500 - 220 = 4280 Farmer 2500 – 220 - 25 = 2255 2500 – 220 = 2280 Pursue vegetable production (B) 250 + 25 = 275 4500 - 220 = 4280 As another example, suppose that Blue offers to pay 25 to Yellow if Yellow plays A. Then the action’s payoffs become:

Blue Farmer Adopt another land use on Pursue vegetable production (B) rewetted peat soil (A) Adopt another land use on 250 + 25 = 275 250 + 25 = 275 Yellow rewetted peat soils (A) 250 – 25 = 225 4500 – 220 - 25 = 4255 Farmer 2500 – 220 = 2280 2500 – 220 = 2280 Pursue vegetable production (B) 250 4500 - 220 = 4280

Throughout the experiment, you will be able to consult your profits obtained in the previous rounds.

B3. Players’ payoffs if organic soils are rewetted, for each payment design tested

Player H Activity A: sustainable use Activity B: vegetable farming 250 + 2300 – SL + SH 250 + SH Activity A: sustainable use 250 + 4300 + SL - SH 4500 - 220 - SH Player L 2500 - 220 - SL 2500 – 220 Activity B: vegetable farming 250 + SL 4500 – 220 Figure B3.1: Player’s payoffs when soils are rewetted - treatment DA

Player H Activity A: sustainable use Activity B: vegetable farming 250 + 3300 – SL + SH 250 + SH Activity A: sustainable use 250 + 3300 + SL - SH 4500 - 220 - SH Player L 2500 - 220 - SL 2500 – 220 Activity B: vegetable farming 250 + SL 4500 – 220 Figure B3.2: Player’s payoffs when soils are rewetted - treatment UA

Player H Activity A: sustainable use Activity B: vegetable farming 250 + 3300 – SL + SH 250 + 3300 + SH Activity A: sustainable use 250 + 3300 + SL - SH 4500 - 220 - SH Player L 2500 - 220 - SL 2500 – 220 Activity B: vegetable farming 250 + 3300 + SL 4500 – 220 Figure B3.3: Player’s payoffs when soils are rewetted - treatment UI

127

B4. Derivation of behavioral predictions

We derive the subgame perfect equilibria for profit-maximizing players for the baseline and each payment treatment. Let (aL, aH) denote the combination of land use strategies of players L and H when rewetting takes place (ai ϵ {A, B}, and let (πL, πH) denote the resulting payoffs in Figure 6. The derivation steps for the predictions are as follows. First, we derive the Nash equilibria for the payoff matrix in Figure 6 in the absence of side payments to understand the basic payoff structure when rewetting takes place and no side payments are made. Side payments could in principle change this payoff structure. Note that side payments only make sense for payoff-maximizing players if there are gains from trade. That is, total payoffs under the equilibrium achieved via side payments have to be larger than total payoffs under the status quo (i.e., πL + πH > 2500 + 4500). Moreover, note that there is only one outcome under rewetting, namely (A, A), that could ever induce both players to vote for rewetting. A payoff- maximizing player votes for rewetting if his/her own expected payoffs from that are at least as high as his/her status quo payoffs. For (B, B), both players payoffs are always higher under the status quo (2500 – 220 > 2500; 4500 – 220 > 4500), so both players would vote against rewetting if (B, B) was the expected outcome. For (A, B), payoffs of player H are always higher under the status quo (4500 > 4500 – 220 – SH). Similarly, for (B, A), payoffs of player L are always higher under the status quo (2500 > 2500 – 220 – SL). Thus, there would be no unanimous vote for rewetting when (A, B) or (B, A) were the expected outcome. Therefore, our second to fourth derivation steps are as follows. The second step is to analyze if there are gains from trade for the case (A, A), i.e. if πL + πH > 2500 + 4500, where πL and

πH are given by the upper left corner of the payoff matrix in Figure 6. If this is the case, the third step is to analyze whether there is a combination of non-negative side payments that induces (A, A) to be a Nash equilibrium in Figure 6. If so, the fourth step is to analyze under which condition both players would vote for rewetting, i.e., for which side payment πL ≥ 2500 and πH ≥ 4500, where again πL and πH are given by the upper left corner of the payoff matrix in Figure 6. For the baseline we also analyze the expected behavior under social preferences. For this, if there are several potential equilibria, we assume that payoff-maximization is the secondary objective.

B4.1. Baseline (UI=APi = 0) The payoff matrix in the baseline scenario and without side payments (UI=APi=SL=SH=0 in Figure 6) has the structure of a prisoner’s dilemma. The dominant strategy for both players under payoff- maximization is to choose activity B. Thus, (B, B) is the unique Nash equilibrium (NE). There are no gains from trade in the baseline. The sum of payoffs under (A, A) is 250 + 250 = 500, which is clearly less than the sum of status quo payoffs (2500 + 4500 = 7000). Therefore, there are no side payments consistent with payoff-maximizing players that would induce (A, A) to become a NE. Also, payoffs for each individual player are higher under the status quo than under (B, B). Therefore, we expect both players to vote against rewetting. Thus, the subgame-perfect equilibrium for payoff-maximizing players is for both players to make no side payment, vote against rewetting and pursue activity B (by default). Players’ social preferences may generate different outcomes. We develop behavioral predictions for “joint profit maximizers” and “inequality averse” players. Joint profits are highest in the status quo (2500 + 4500). Thus, we predict that joint profit maximizers will also vote against rewetting and pursue activity B. If joint profit maximization is the only motive or payoff maximization is the secondary objective, there is no reason for side payments.

Equal payoffs between group members can only be realized if rewetting takes place (i.e. in the payoff matrix in Figure 6). Specifically, there are three options for equalizing payoffs. First, under (A, A) both players receive a payoff of 250. Second, under (A, B) payoffs can be equated via a side payment from H to L, namely SH = ((4500 – 220) - 250)/2 = 2015. In that case, both players’ payoffs are equal to 2265. 128

Third, under (B, A) payoffs can be equated via a side payment from L to H, namely SL = ((2500 – 220) - 250)/2 = 1015. In that case, both players’ payoffs are equal to 1265. Among these three payoff- equalizing options, the second option yields the highest payoff for each of the two players. Therefore, assuming that payoff-maximization is a secondary objective, we expect the following outcome for inequality-averse players: H makes a side payment to L of 2015; both players vote for rewetting; L chooses activity A and H reverts back to activity B.

B4.2. Differentiated agglomeration payment (APH = DAH = 4300 and APL = DAL = 2300, UI = 0) In this case, the payoff structure in Figure 6 without side payments (SL=SH=0) represents a coordination game. The equilibrium (A, A) is also the social optimum; the equilibrium that the policy instrument aims at achieving. If L chooses A, H also chooses A. If L chooses B, H also chooses B, and vice versa. Thus, there are two NE, namely (A, A) and (B, B). No side payment is needed to induce (A, A) to be an equilibrium. Moreover, both players payoffs are higher under (A, A) than under the status quo (2550 > 2500, and 4550 > 4500). Therefore, the subgame-perfect equilibrium for payoff-maximizing and risk- neutral players is as follows: no side payments are offered, both players vote for rewetting, both players choose activity A. Yet, the coordination-game structure of the payoffs under rewetting implies a risk of coordination failure. Moreover, payoffs under (A, A) exceed status quo payoffs only by 50 per player. Thus, risk averse players may prefer to not rewet and obtain the certain payoffs of the status quo.

B4.3. Uniform agglomeration payment (APH = APL = UA = 3300, UI = 0) The payoff structure under UA and without side payments is that of a social dilemma. For player H the dominant strategy is to choose activity B. This is because the uniform payment is too low to cover the opportunity cost of player H. Given the structure of the agglomeration payment, if H chooses B, it is optimal for player L to also choose B. Thus, in the absence of side payments, (B, B) is the unique NE under rewetting. But contrary to the baseline case, under UA there are gains from trade that could motivate side payments. The total payoffs under (A, A) are higher than under the status quo (250 + 3300 + 250 + 3300 = 7100 > 7000). Thus, our next step is to analyze if there is a side payment that could transform (A, A) into a NE in Figure 6. If L chooses A, then H also chooses A if 250 + 3300 + SL - SH ≥ 4500 - 220 - SH, i.e. if SL ≥ 730. If H chooses A, then L also chooses A if 250 + 3300 - SL + SH ≥ 2500 – 220 - SL, which always holds for non-negative side payments. The maximum side payment L could make to induce H to choose A is such that L’s payoff from (A, A) are higher than payoffs from (B, B), i.e. 250 + 3300 – SL > 2500 – 220, which is equivalent to SL ≤ 1270. Thus, a side payment from L to H that satisfies 730 ≤ SL ≤ 1270 induces (A, A) to be a NE under rewetting. Therefore, in the presence of such a side-payment offer from L to H, Figure 6 is turned into a coordination game as in DA with two NEs that are (A, A) and (B, B). Next we examine under which condition both players would vote for rewetting to get to (A, A) NE. Player L votes for rewetting if his/her payoffs from (A, A) are at least as high as under the status quo, i.e., if 250 + 3300 – SL ≥ 2500. This requires that SL ≥ 950. Similarly, player H votes for rewetting if 250 + 3300 + SL ≥ 4500, requiring that SL ≤ 1050. This implies that players would prefer rewetting if and only if 950 ≤ SL ≤ 1050. This condition on side payments is stronger than the condition for (A, A) to be a NE under rewetting (730 ≤ SL ≤ 1270). Thus, the subgame-perfect equilibrium for payoff- maximizing players is as follows. Player L offers a side payment that satisfies 950 ≤ SL ≤ 1050. Both players then vote for rewetting and choose activity A. Note, however, that for 950 ≤ SL ≤ 1050, the payoff matrix in Figure 6 has the structure of a coordination game, similar to the DA treatment. Thus risk averse players may still prefer to vote against rewetting and obtain certain payoffs from the status quo. However, in this treatment the side payment may serve as a signal from the L-player that (s)he aims at the equilibrium. 129

B4.4. Uniform individual payment (UI = 3300, APi = 0) The payoff structure in Figure 6 under the UI treatment and without side payments is as follows. Regardless of the land use choice of H, the L player is better off choosing activity A. Thus, A is the dominant strategy of player L. Similarly, the H player’s dominant strategy is to choose activity B. Intuitively, this happens because the UI payment is made regardless of the activity choice of the other players. Thus, players choose activity A as long as the payment covers their opportunity cost. Under the uniform payment, this is the case for L, but not for H. Thus, in the absence of side payments, there is a unique NE under rewetting, namely (A, B). The next steps are exactly identical for UI as for UA. There are gains from trade, i.e. the sum of payoffs under (A, A) exceeds that under the status quo. A side payment 730 ≤ SL ≤ 1270 transforms (A, A) into a NE in Figure 6. Both players vote for rewetting to achieve this outcome if 950 ≤ SL ≤ 1050. Thus, the subgame-perfect equilibrium for payoff-maximizing players is as for UA: Player L offers a side payment that satisfies 950 ≤ SL ≤ 1050. Both then vote for rewetting and choose activity A. There are, however, two differences between UI and UA that should be noted. First, for UI, the game structure under 950 ≤ SL ≤ 1050 is that of a non-dilemma. With the optimal side payment, activity A becomes the dominant strategy for both players. Therefore, there is no risk of coordination failure, once the optimal side payment is offered. Second, note that joint payoffs are actually highest under (A, B), the NE under rewetting without side payments. This equilibrium makes player H better off and player L equally well off as (A, A). However, the Pareto-efficient equilibrium (A, B) is not a subgame-perfect equilibrium because it yields lower payoffs for H than under the status quo (4500 – 220 < 4500). This cannot be offset by side payments since these are received only at the condition of choosing activity A.

B5. Representation of social preferences-types

Table B5: Representation of social preferences-types across treatments, in percentage

Mean SVO angle Prosocial Proself N DA 24.4° (12.6) 55.4% (IA: 36.5%; JM: 6.8%) 44.60% 74 UA 22.2° (12.7) 50% (IA: 31.1%; JM: 6.8%) 50% 74 UI 24.5° (13.3) 59.4% (IA: 31.1%; JM: 13.5%) 40.50% 74 Note: The table also indicates the percentage of players identified as either Inequality averse (IA) or joint-profit maximizer (JM).

B6. Representation of players’ characteristics across treatments

Table B6: Other players’ characteristics across treatments

Treatments UI DA UA Age 22.3 (3.1) 22.2 (3.1) 21.9 (3.5) Feedback instructions; 0=very clear, 3= very difficult 0.5 (0.5) 0.5 (0.6) 0.5 (0.6) Opinion on peat degradation among the players who are knowledgeable about it; 1= not a problem at all, 4= a very 1.5 (0.6) 1.6 (0.6) 1.7 (0.6) serious problem Index Altruism (in %) 47.6 (10.5) 46.9 (11.8) 48.3 (11.3) Time preference (money amount) 858 (917) 852 (964) 713 (707) (outliers excluded) Environmental scheme preferences. Preference for collective participation ; collective participation = 0, individual 73.0 75.7 82.4 participation= 1 Note: Standard deviation in brackets. The representation of the variables is not significantly different across treatments (proportion tests and t-test at 95% confidence level), indicating successful randomization across treatments.

130

3. Appendix C – Appendix to chapter 3

C1. SVO test description

The analysis of the pattern of choice of the individuals across a 15-items slider of decision distribution enables to compute a score (an angle ranging between -16.26° and 61.39°). This angle represents the extent to which the player gives importance to his own income in relation to the outcome of other players (other participants of the experiment), as their choice in this test will affect their own outcome as well as the outcome of another person. The higher the angle, the higher the prosocial inclination of the player. The lower the angle, the more individualistic the person is (i.e. the more the person gives importance to his/her own outcome) (cf. Murphy et al., 2011). The instructions for the SVO test were provided as follows: Procedure: In this task you will make a series of 15 decisions about allocations of points for you and for some other participant. The other person is someone you do not know. He or she does not know you either. You will both remain mutually anonymous and all of your choices are confidential. Your decisions will yield points for both yourself and for the other person. In the example below, a person has chosen to distribute the points so that he or she receives 85 points, while the other person receives 75 points. The combinations regarding the distribution of points are thus fixed and you will have to choose which combination you prefer.

As you can see, your choice will influence both the number of points you receive as well as the number of points someone else receives. There are no right or wrong answers for this task. This is only a matter of your preference. For each of the 15 questions, please indicate the distribution you prefer most by clicking on it. Once you have taken your decisions, the resulting money distribution will appear in the space on the right hand side of the question. Earnings: Each decision you make will potentially be chosen to determine your earnings of this part. At the end of the whole experiment, a number between 1 and 15 will be randomly drawn by your computer. The corresponding decision (recall that you have to make 15 decisions) will be used to calculate your earnings for this part. For instance, if the number 12 is drawn, decision n° 12 will be used to calculate your earnings for this part. A randomly selected person will receive the amount of points you allocated to the “other person” in this decision. You will also receive the points that a different randomly selected person allocated to the “other person” in this decision. Your final earnings will be the sum of both the amount you allocated to yourself and the amount a randomly selected other person allocated to the “other person” in the relevant decision. The conversion rate for this part 10 points = 1.40 CHF.

C2. Experimental design

Organic soils consist of a first layer of peat (rich in organic matter) and an underneath mineral soil layer. Each group consists of two extremes cases of players (farmers) i who differ by the type of the underlying soil layer of their organic soils: i= {L, H}. H player has high quality mineral soils and therefore long- 131 term production potential in the conventional land use (Blue farmer in the experiment). L player has low quality mineral soils and therefore short-term production potential (Yellow farmer in the experiment).

The dynamics of cultivated organic soils are represented by 푑푡, where 푑푡≡ drainage at time t, defined as: 0, organic soils are drained (current situation) 푑 = { } , 푡 = {1, . . . ,10} 푡 1, organic soils are rewetted by stopping drainage

푖 The thickness of the peat layer degrades and reduces with drainage. 푇푡 is defined as the thickness of the peat layer for player i at time t.

푖 푖 0: 푇푡+1 = 훼푇푡 , 푤𝑖푡ℎ 0 < 훼 < 1, 훼 is a constant which is independant of 𝑖 퐼푓 푑푡 = { 푖 푖 } 1: 푇푡+1 = 푇푡 As the peat layer reduces with vegetable farming, the underlying mineral soil layer gets closer to the surface. L has an underlying mineral soil layer that is not suitable for vegetable production while H has one which is suitable for vegetable production. This translates into different profit functions for L and H (Figure C2.1). The degradation of the peat does not affect H’s farm profit: regardless of the number of time periods H produces vegetables, his farm profit remains constant for this activity. The degradation of the peat negatively affects L’s farm profit which depends on the number of time periods used previously for farming vegetables.

Figure C2.1: Profit functions of H and L players over time

푖 휋푛,푡 is defined as the agricultural productivity for vegetable farming of the organic soil for player i at time t, given a number n of previous time periods under vegetable farming (n={0,...,9}). It is defined as: 퐼푓 𝑖 = 퐻, 휋퐻 = 휋퐻 푛+1,푡+1 푛,푡 (the productivity of H for vegetable farming is constant) 퐼푓 푛 = 0, 1, 2, 푡ℎ푒푛 휋퐿 = 휋퐻 = 800, 푡 = {1, … ,10} 푛,푡 푛,푡 퐼푓 푑푡 = 0: 퐿 퐼푓 푛 = 3, 푡ℎ푒푛 휋푛,푡 = 550, 푡 = {4, … ,10} 퐼푓 𝑖 = 퐿: 퐿 퐼푓 푛 = 4, 푡ℎ푒푛 휋푛,푡 = 160, 푡 = {5, … ,10} 퐿 { { 퐼푓 푛 = 5, 6, 7, 8, 9, 푡ℎ푒푛 휋푛,푡 = 0, 푡 = {6, … ,10} }} 푖 푖 퐼푓 푑푡 = 1: 휋푛,푡+1 = 휋푛,푡 The procedure of a time period is as follows. Group members can first anonymously communicate through chat messages. Each member is invited to vote on rewetting organic soils for the group, or not, i.e. to vote for 푑푡 = 0 or 1. If at least one of the members rejects to rewet organic soils, players continue producing vegetables (activity B). If both players vote for rewetting, activity B is no longer possible. Players are then asked to make a choice between two types of agricultural activities: activity A) implementing a sustainable land use adapted to high water table (profit equal to 40) or activity B)

132 pursuing activity B by draining back organic soils with the installation of a personal drainage system (cost equal to 25).

푖 The set of actions 푎푡 of the player i at time t is therefore as follows: 퐴: 𝑖 adopts a sustainable land use on rewetted organic soils

푖 퐼푓 푑푡 = 1: {퐵: 푓 pursues vegetable farming using a personal drainage system } 푎푡 = (cost = 25) { 퐼푓 푑푡 = 0: 퐵: 𝑖 automatically pursues vegetable farming on organic soils }

In the baseline phase, player’s payoffs 푢푡 are as follows: 퐻 40 𝑖푓 (푑푡; 푎푡 ) = (1; 퐴) 퐻 퐻 퐻 퐻 For H players: 푢푡 (푑푡; 푎푡 ) = { 휋푛,푡 − 25 = 775 𝑖푓 (푑푡; 푎푡 ) = (1; 퐵)} ; 푡 = 1, … , 10 퐻 퐻 휋푛,푡 = 800 𝑖푓 (푑푡; 푎푡 ) = (0; 퐵) 퐿 40 𝑖푓 (푑푡; 푎푡 ) = (1; 퐴) 퐿 퐿 퐿 퐿 For L players: 푢푡 (푑푡; 푎푡 ) = { 휋푛,푡 − 25 𝑖푓 (푑푡; 푎푡 ) = (1; 퐵)} ; 푡 = 1, … , 10; 푛 = 0, … ,9 퐿 퐿 휋푛,푡 𝑖푓 (푑푡; 푎푡 ) = (0; 퐵) In the treatment phase, we test one type of agglomeration payment. Players receive the payment (AP) if both group members vote for adopting sustainable use of organic soils (activity A). Thus:

퐿 퐻 퐿 퐿 퐻 퐻 If [(푑푡; 푎푡 ) = (1; 퐴)] ∩ [(푑푡; 푎푡 ) = (1; 퐴)], then 푢푡 (푑푡; 푎푡 ) = 푢푡 (푑푡; 푎푡 ) = 40 + 퐴푃 Before proceeding to the vote, each player can make a binding side-payment offer to the other member. The side payment is only transferred if both members vote for rewetting the soils and on the condition that the potential beneficiary adopts activity A, regardless of the choice of the player making the offer (see Figure C2.3). The payoff matrixes at time t are as in Figures C2.2 and C2.3 below. Payoff matrixes are displayed on the player’s screen all along the experiment and their content is updated at each time period based on players’ actions.

H player Activity B: vegetable farming 휋퐿 L player Activity B: vegetable farming 푛,푡 훑퐇 Figure C2.2: Players’ payoffs at time t if organic soils are not rewetted.

H player Activity A: Sustainable use of organic soils Activity B: vegetable farming 퐿 L H H Activity A: Sustainable 40 + 퐴푃푛,푡 – S + S 40 + S use of organic soils 40 + 퐀퐏퐇 + SL - SH 훑퐇 - 25 - SH L player 퐿 L 퐿 Activity B: vegetable 휋푛,푡 - 25 - S 휋푛,푡– 25 farming 40 + SL 훑퐇 – 25 Figure C2.3: Players’ payoffs at time t if organic soils are rewetted.

L H L H Note: APn,t and AP = agglomeration payment. S : side-payment offer made by L to H, S : side-payment offer made by H to L.

133

The two agglomeration payment schemes tested are defined as: APC: constant agri-environmental agglomeration payment, equal between L and H players regardless of the time period (equal to 770), and

APV: variable agri-environmental agglomeration payment which mirrors players’ opportunity costs of adopting a sustainable land use, and can therefore vary. It is constant for H. For L, it evolves according to the number of time periods used for vegetable farming: when L’s farm profit decreases, the payment that (s)he could receive in exchange of adopting activity A decreases by the same proportion; it is defined 퐿 as 퐴푃푛,푡: payment that L receives at t, given a number n of previous time periods under vegetable farming.

퐻 퐻 퐼푓 𝑖 = 퐻, 퐴푃푛+1,푡+1 = 퐴푃푛,푡 = 770 (payment constant in time); 퐼푓 푛 = 0, 1, 2, 푡ℎ푒푛 퐴푃퐿 = 퐴푃퐻 = 770, 푡 = {1, … ,10} 푛,푡 푛,푡 퐿 퐼푓 푛 = 3, 푡ℎ푒푛 퐴푃푛,푡 = 520, 푡 = {4, … ,10} 퐼푓 𝑖 = 퐿: 퐿 퐼푓 푛 = 4, 푡ℎ푒푛 퐴푃푛,푡 = 130, 푡 = {5, … ,10} 퐿 { 퐼푓 푛 = 5, 6, 7, 8, 9, 푡ℎ푒푛 퐴푃푛,푡 = 10, 푡 = {6, … ,10} }

C3. Instructions – Baseline scenario

Part 2 comprises 10 time periods of a decision process. You are randomly assigned to a group with one other participant. This participant can be any other participant present here today. Any interactions in this study are anonymous. At the end of the experiment, one time period will randomly be picked from this part and your earning in points from this round will be exchanged into francs (CHF) at a rate of 10 points = 0.25 CHF. Before starting Part 2, there will be a learning time period during which your decisions will not be recorded and will therefore not determine your end payment. Note that there are no right or wrong answers in your choices and actions; this is only a matter of preferences.

Background:

Peat soils (Torf / organic soils) consist of a first layer of peat and an underneath mineral soil layer (Figure C3.1). The peat is very good for vegetable production. The peat layer is very rich in organic matter (and thus in carbon) which accumulated in high water level conditions. Since a few decades, the water table (or groundwater) is artificially maintained at a low level by a drainage system in order to enable agricultural activities on peat soils. Drainage leads to the reduction of the peat soil layer over time (see Figure C3.2), in particular, because of carbon emissions. The peat soil layer disappears at a rate of about 1cm/year. Once the peat soil layer will be gone, farmers will have to produce on the underneath mineral soil layer. Depending on the location, the quality of this underneath mineral soil layer varies.

134

Peat soil layer

Underneath mineral soil layer variable: either suitable to vegetable production, or not.

Figure C3.1: Peat soil profile Figure C3.2: Evolvement of the peat soil profile over time

One of the most effective ways to avoid the loss of peat soils is to rewet them, so to stop the drainage. The drainage systems are regulated at a community level. Therefore stopping the drainage system needs to be decided unanimously. The water table will therefore only be increased if all farmers agree on it (unanimous decision).

Overview of the experiment:

- Blue-farmer has an underneath mineral soil layer which is suitable for vegetable production. This means that regardless of the number of time periods he/she uses his/her peat soils for vegetable production, the loss of the peat soil layer will not affect his/her farm profit, which will stay constant. Each time period follows the same pattern: You first need to decide whether to rewet the peat soils of your land or not to do so. They will only be rewetted if both of you vote for it.  If at least one of you rejects to stop the drainage system and thus to increase the water table, you keep producing vegetables. You will be directed to next time period where you will be asked the same question.

- A) Adopting another land use (e.g. crops or livestock adapted to high water level conditions); your farm profit will then be lower than with vegetables or

- B) Pursuing vegetable production by implementing a personal local drainage system on peat soils at a cost to you. You will then be directed to next time period where you will be asked the same questions.

135

However, before these choices are made, Yellow may offer a binding amount to be transferred by Yellow to Blue if and only if Blue chooses A. Blue (at the same time) may offer a binding transfer payment to be transferred by Blue to Yellow if and only if Yellow chooses A. Note that transfer-payment offers have the effect of changing the total payoffs. Whatever amount you state will be transferred to the other farmer if he/she plays A; this money will be transferred regardless of your own choice of land use on rewetted peat soils.

Procedure:

The general procedure of one time period and the effect of your actions on your profit for the first time period are as follows: o Before making a choice in a time period, you can discuss with your group member via a “chat box window”. As you discuss, you can test how a transfer-payment offer to your group member would change your payoffs. During this discussion time, you are free to discuss the experiment or other matters (you have about 3 minutes each time period). As you exchange messages, please be civic and respectful to one another.. o After the discussion, each of you can make an offer for a transfer o payment that you would pay to the other player in case he/she chooses option A. Whatever offer you make, it is binding. That is, the offered amount will be transferred from you to the other player automatically if the other player chooses option A. o Both members Yellow and Blue are asked to vote on increasing the water table for their group, or not. There are 2 possible outcomes: - If at least one of you rejects to increase the water table, the water table is not increased and peat soils are not rewetted. Both members continue with vegetable farming on peat soils. Yellow-farmer’s and Blue-farmer’s farm profits for this time period are respectively 800 and 800. No transfer

- payments are made in this case because both of you automatically choose vegetable production.

Blue Farmer Does not increase the water table: continues farming vegetables on the peat soils Yellow Does not increase the water table: continues 800 Farmer farming vegetables on the peat soils 800

o If both of you vote for increasing the water table, peat soils of both members are rewetted. In that case, you need to choose between:

- A) dealing with the high water table and establishing a different land use (your farm profit will then be 40) and

- B) installing a personal drainage system on your peat soils at your own cost (cost of 25) and continue producing vegetables. If no transfers are made, the actions’ payoffs are as below.

136

Blue Farmer Adopt another land use on Pursue vegetable production (B) rewetted peat soils (A) Adopt another land use on rewetted 40 40 Yellow peat soils (A) 40 800 – 25 = 775 Farmer 800 – 25 = 775 800 – 25 = 775 Pursue vegetable production (B) 40 800 – 25 = 775

Please note that the yellow numbers represents what the Yellow Farmer gets and the blue numbers represents what the Blue farmer gets. These profits apply if the farmers have taken the corresponding actions, which are read on the left side of the box for the Yellow farmer and on the upper side for the Blue farmer. For instance, if both of you choose to rewet and then choose A, that means both of you earn 40. As another example, if both of you choose to rewet and then Yellow Farmer chooses A and Blue farmer chooses B, Yellow Farmer earns 40 and Blue farmer earns 775.

After all have made their choices, your screen will display your choice and profit for this time period, as well as the ones of your other group member. The game then continues to next time period.

Evolvement of the farm-profits over the time periods:

When farming vegetables, the peat soil layer is used and then irreversibly reduces. - This does not affect the profit of Blue-farmer: regardless of the number of time periods Blue- farmer produces vegetables, his/her farm profit remains constant for this activity (800 or 775 (800-25)). - But peat soil degradation negatively affects the farm profit of Yellow-farmer. If Yellow-farmer produces vegetables for 3 time periods (successively or not), at the 4th time period he/she produces vegetables, Yellow-farmer’s farm profit will decrease to 550. If he/she produces vegetables for 1 time period more, his/her farm profit will decrease to 160. If he/she produces vegetables for 1 time period more (i.e. at the 6th time period that he/she produces vegetables during the experiment), his/her farm profit will then be equal to 0. This is due to the unsuitability of Yellow-farmer’s underneath mineral soil layer for vegetable production. Concretely, the numbers circled below will decline:

Blue-Farmer Does not increase the water table: continues farming vegetables on the peat soils Yellow- Does not increase the water table: continues 800 Farmer farming vegetables on the peat soils 800 And if you both agree on increasing the water table, the Yellow-farmer’s profits from installing a local drainage system (option B), circled below, will be affected too:

Blue-Farmer Adopt another land use on Installs a local drainage system on rewetted peat soils (A) rewetted peat soils (B) Adopt another land use on rewetted 40 40 Yellow- peat soils (A) 40 800 – 25 = 775 Farmer Installs a local drainage system on 800 – 25 = 775 800 – 25 = 775 rewetted peat soils (B) 40 800 – 25 = 775

More precisely, if Yellow-farmer produces vegetables on peat soils during 10 time periods successively, his/her respective farm profit will vary as follows:

137

Farm profit of Yellow-farmer when producing vegetables during - 10 time periods successively 1000 800 800 800 Yellow 800 550

600 farmer 400 160 200 Farm profit profit Farm of 0 0 0 0 0 0 1 2 3 4 5 6 7 8 9 10 Time periods under vegetable production

Figure C3.3: Evolvement farm profit of Yellow farmer

When installing a personal drainage system on the rewetted peat soils (Option B), the peat soil is used and thus degrades. Therefore the Yellow-farmer’s farm profit will decrease accordingly (the amount indicated in Figure C3.3 minus the constant cost of 25). The Blue-farmer’s farm profit will stay constant (equal to 775) because the use of the peat soil layer for vegetables does not affect Blue’s farm profit. Any time Yellow-farmer or Blue-farmer decide to increase the water table and adopt another land use (Option A), their profit will be equal to 40. For Blue-farmer, if he/she then starts farming vegetable again on his/her peat soil, his/her farm profit will then be 800 (or 775 (800-25)). For Yellow-farmer, if he/she then starts farming vegetable again on his/her peat soil, his/her farm profit will depend on the number of time periods he/she has been farming vegetables previously. This includes the times where the water table was not increased, and the time where Yellow chooses option B. NOTE: Rewetting the peat soils by stopping the drainage is a way to preserve the remaining peat layer; in no way, it allows to recover it. Therefore, it does not allow to recover Yellow’s farm profit neither.

Examples of transfer payments:

Suppose for example, that Yellow offers to pay 10 to Blue if Blue plays A. Then the action’s payoffs become (transfer payments are put in bold in the table):

Blue Farmer Adopt another land use on Pursue vegetable production (B) rewetted peat soils (A) Adopt another land use on rewetted 40 – 10 = 30 40 Yellow peat soils (A) 40 + 10 = 50 800 – 25 = 775 Farmer 800 – 25 - 10 = 765 800 – 25 = 775 Pursue vegetable production (B) 40 + 10 = 50 800 – 25 = 775 As another example, suppose that Blue offers to pay 10 to Yellow if Yellow plays A. Then the action’s payoffs become:

Blue Farmer Adopt another land use on Pursue vegetable production (B) rewetted peat soils (A) Adopt another land use on rewetted 40 + 10 = 50 40 + 10 = 50 Yellow peat soils (A) 40 – 10 = 30 800 – 25 – 10 = 765 Farmer 800 – 25 = 775 800 – 25 = 775 Pursue vegetable production (B) 40 800 – 25 = 775

138

Throughout the experiment, you will be able to consult the payoffs obtained in the previous time periods.

C4. Percentage of H and L players who adopt activity A

100 90 80 70 60 50 40 30 20 10 0 Percentage of L players who adopt activity A A activity adopt who L players of Percentage 1 2 3 4 5 6 7 8 9 10 Time-periods Baseline (39 L players) Constant agglomeration payment (44 L players) Variable agglomeration payment (44 L players) Figure C4.1: Percentage of L players who adopt activity A

100 90 80 70 60 50 40 30 20 10 0 Percentage of H players who adopt activity A A activity adopt who players H of Percentage 1 2 3 4 5 6 7 8 9 10 Time-periods Baseline (39 H players) Constant agglomeration payment (44 H players) Variable agglomeration payment (44 H players) Figure C4.2: Percentage of H players who adopt activity A

C5. Social preference types

Table C5: Representation of social preference types in each scenario, in percentage

Baseline (78 players) 퐀퐏퐅 (88 players) 퐀퐏퐕 (88 players) Proself Individualistic 39.7 31.8 38.6 Inequality averse 37.2 42.0 23.9 Prosocial Joint profit maximizer 6.4 6.8 3.4

Note: the rest of the players was either classified as inconsistent according to the rule of Murphy et al., (2011) or did not reveal any trend towards joint maximizers or inequality averse profiles. The representation of players’ types does not vary significantly across treatment scenarios (proportion test at 95% confidence level). 139

C6. Representation of players’ characteristics across scenarios

Table C6: Players’ characteristics per player type across treatments

Players’ characteristics (on average) Baseline 퐀퐏퐅 퐀퐏퐕 Player type H L H L H L 18.8 19.2 20.6 19.9 20.8 20.2 Age (2.5) (1.4) (5.0) (2.8) (3.7) (3.1) Feedback instructions 1.3 (0.8) 1.5 (1.5) 1.2 (0.8) 1.2 (0.7) 1.4 (0.8) 1.5 (0.9) 0=very clear, 3= very difficult % of players who have a farm at home 30.8 30.8 68 70 56.8 63.6 Number of players who have organic soils in the 2 2 11 10 9 10 farm % of participants who plan to be farmer 54 79.5 98 97.7 97.7 97.7 Level of players’ knowledge about degradation of organic soils in CH; 1.3(0.6) 1.2 (0.8) 1.1(0.8) 1.2 (0.5) 1.0 (0.8) 1.1 (0.5) 0 = great deal, 3 = nothing Opinion on degradation of organic soils among the players who have knowledge about the issue; 2.1 (0.6) 1.9 (0.9) 2.0 (0.8) 2.2 (0.6) 2.1 (0.6) 1.9 (0.9) 1= not a problem at all, 4= a very serious problem 46.3 49.5 48.2 47.5 45.1 46.9 Index Altruism (in %) (9.8) (14.8) (12.5) (11.2) (11.3) (12.9) 54.6 54.4 58.1 57.1 58.5 55.7 Index Care for the environment (in %) (14.7) (15.5) (14.4) (16.7) (14.7) (15.4) Financial risk 0 =avoid financial risk, 10 =willing to take 3.9 (2.3) 4.4 (2.5) 4.6 (2.1) 4.5 (2.7) 5.0 (2.8) 5.0 (2.2) financial risk Risk 5.9 (2.0) 6.3 (2.9) 6.6. (2.0) 6.3 (2.3) 7.1 (2.1) 6.8 (2.0) 0 = avoid risk, 10=willing to take risk Time preference (money amount) 1189 1146 948 775.7 1196 778 (outliers excluded) (1036) (1135) (1074) (841) (1372) (913) Impatience 5.06 5.1 (3.0) 4.2 (3.4) 5.0 (2.6) 5.5. (2.7) 4.8 (2.3) 0 = very impatient, 10 = very patient (2.4) Care for reputation 1.34 1.3 (0.9) 1.1 (0.9) 1.2 (0.6) 1.4 (0.7) 1.4 (0.7) 1= care a lot, 4 = don’t care at all (0.8) Environmental scheme preferences Preference for collective participation as 79.5*% 43.6*% 56.8 % 62.8% 70.4% 63.6% compared to individual participation Opinion of cooperative approaches in Ag. 0.95 management; 0=more successful than other 0.9 (0.6) 0.9 (0.6) 1.1 (0.7) 1.0 (0.7) 1.1 (0.8) (0.6) approach, 3=not at all Note: standard deviation in brackets; (*): significant difference between H and L players at 95% confidence level; Underlined: For a same player type (H or L), the value underlined is significantly different from the same value in the other treatments, at 95% confidence level.

140

4. Appendix D – Appendix to chapter 4

D1. Background visualization

Below are screen shots of the interfaces of the players in the framed versus the unframed experiment.

Figure D1.1: Visual background of the framed experiment.

Note: the background represents two vegetable farming systems (i.e. group members) separated by a joint drainage canal

Figure D1.2: Visual background of the unframed experiment: context-free

141

D2. Side payments

Table D2: Analysis of side payments in the experimental setups

Mean % of groups who Mean net H side-payment offer make an offer Experimental Set up Among groups Among all Among groups Among all groups who cooperate groups who cooperate Baseline R4 to R10: 173.7 (55.0) - 25 - Dynamic UA -40 (23) -71.3 47.5 23 Apprentices Baseline - - 78 - framed Static UI - -188.4 (198.4) 85 15 DA 136.1 (152.9) 129.2 (105.0) 78.6 37 Baseline R4 to R10: 171.3 (63.9) - 40 - Dynamic UA -250 (328) -250 (328) 10.2 10.2 Students Baseline 77.8 (74.0) - 65 - framed UA -427 (108.4) -337.9 (126.7) 77 41 Static UI -182.8 (33.7) -184.0 (284.4) 76 18 DA 110.5 (64.2) 139.1 (81.7) 74 35 Baseline TP4 to R10: 203.6 (25.9) - 37 - Dynamic Students UA -147.0 (46.1) -147.0 (46.1) 10.5 10.5 unframed Baseline - - 59 - Static UA -307.2 (104.2) -477.59 (116.7) 80 36.4 Note: R = round. “-“: the mean is not significantly different from 0 at 90% conf. level (t-test). Groups cooperate = both members adopt Activity 2.

D3. Framing and treatment cost effectiveness

Table D3: Impact of the framing on cost effectiveness (students)

Peat preserved as Payment made as % of the Cost- N % of the total total payment possible effectiveness Dynamic UA 93.75 97.5 1,601.6 80 Framed Static UA 51.2 49.7 3,204.2 74 Dynamic UA 87.9 93.9 1,646.05 76 Unframed Static UA 41.3 39.5 3,156.5 78

D4. Framing and income inequalities

Table D4: Impact of the framing effect on inequalities (students)

Gini coefficient (std.) N Baseline 0.27 (0.16) 80 Dynamic UA 0.03 (0.01) 80 Framed Baseline 0.16 (0.01) 222 Static UA 0.14 (0.01) 74 Baseline 0.31 (0.18) 76 Dynamic UA 0.02 (0.01) 76 Unframed Baseline 0.15 (0.01) 78 Static UA 0.14 (0.01) 78

142

D5. Characteristics of players

Table D5.1: Characteristics of players in student (S) versus apprentice (A), across treatments and settings

Dynamic setting Static setting Characteristics of players UA UA UI UI DA DA S A S A S A 21.3 20.3 22.3 22.3 22.1 21.6 Age (2.3) (4.1) (3.1) (2.3) (3.0) (1.6) Feedback instructions; 0.8 (0.6) 1.2 (0.8) 0.5 (0.5) 0.9 (0.7) 0.5 (0.6) 0.9 (0.7) (0=very clear, 3= very difficult) Level of knowledge about degradation of peat 1.15 2.2 (0.9) 2.3 (0.8) 1.3 (0.9) 2.3 (0.9) 1.2 (0.8) soils in CH; (0 = great deal, 3 = nothing) (0.7) Opinion of peat soil degradation among players who have knowledge about it; (1= not a 1.9 (0.5) 2.1 (0.7) 1.5 (0.6) 2.0 (0.6) 1.5 (0.6) 2.1 (0.8) problem at all, 4= a very serious problem) 46.1 47.9 47.6 44.0 46.9 46.9 Index of Altruism (11.4) (11.8) (10.5) (8.7) (11.8) (10.4) Index of Care for the environment 62.6 57.6 64.0 68.8 63.1 62.0 (13.2) (15.5) (14.1) (13.8) (11.8) (10.1) Financial risks; (0 =avoid, 10 =willing to take 3.4 (2.3) 4.6 (2.4) 3.4 (2.6) 3.4 (2.4) 3.2 (2.2) 4.1 (2.4) financial risks) Risks; (0 = avoid, 10=willing to take risks) 4.9 (2.3) 6.4 (2.2) 4.6 (2.4) 5.8 (2.1) 4.6 (2.3) 5.3 (2.4) Time preference measure (money amount) 597 860 857.8 723 853 504 (outliers excluded) (472) (961) (917.5) (730) (964) (287) Impatience; (0 = very impatient, 10 = very 5.7 (2.3) 5.0 (2.5) 5.4 (2.5) 4.9 (2.6) 5.8 (2.3) 5.5 (2.7) patient) Care for reputation; (1= care a lot, 4 = don’t 1.5 (0.7) 1.2 (0.7) 1.5 (0.7) 1.7 (0.7) 1.5 (0.7) 1.4 (0.6) care at all) % players who prefer individual as compared to collective participation in environmental 16.25% 40.2% 27.0% 33.3% 24.3% 39.3% scheme Opinion of cooperative approaches in 1.06 agricultural management; (0=more successful 0.6 (0.7) 1.0 (0.6) 0.9 (0.6) 1.0 (0.8) 1.1 (0.7) (0.7) than other approaches, 3=not at all) Gender (% females) 58.8 20.7 63.5 53.3 54.0 57.1 SVO angle 22.3 25.1 24.6 25.0 24.4 24.1 (13.4) (15.7) (13.3) (12.0) (12.6) (15.1) Note: Within apprentices, in UA-dynamic, 69% of the players have a farm at home and 98% plan to become farmers. In UI- static, 60% have a farm at home and 75% want to become farmers, and in DA-static, 57% have a farm at home and 75% want to become farmers.

143

Table D5.2 Characteristics of players in student (S) versus apprentice (A)

Farm Characteristics of players Students apprentices Age 21.6 (2.9) 22.0 (2.01) Feedback instructions; 0.6 (0.6) 0.97 (0.8) (0=very clear, 3= very difficult) Level of knowledge about degradation of peat soils in CH; (0 = great deal, 3 2.3 (0.9) 2.1 (0.75) = nothing) Opinion of peat soil degradation among players who have knowledge about it; 1.7 (0.6) 2.09 (0.7) (1= not a problem at all, 4= a very serious problem) Index of Altruism 46.8 (11.2) 46.9 (11.0) Index of Care for the environment 62.0 (13.1) 60.8 (14.9) Financial risks; (0 =avoid, 10 =willing to take financial risks) 3.32 (2.3) 4.23 (2.4) Risks; (0 = avoid, 10=willing to take risks) 4.7 (2.3) 6.08 (2.2) Time preference measure (money amount) 722 (681) 733 (707) (outliers excluded) Impatience; (0 = very impatient, 10 = very patient) 5.6 (2.3) 5.1 (2.5) Care for reputation; (1= care a lot, 4 = don’t care at all) 1.5 (0.7) 1.6 (0.7) % players who prefer individual as compared to collective participation in 22 38.6 environmental scheme Opinion of cooperative approaches in agricultural management; (0=more 2.1 (0.75) 2.2 (0.65) successful than other approaches, 3=not at all) Gender (% females) 58.5 42 SVO angle 23.7 (13.1) 24.9 (14.8)

Table D5.3: Comparative analysis of the characteristics of student players according to framing type

Dynamic setting Static setting UA UA UA UA Characteristics of players Unframed framed Unframed framed Previous participation experiment 57.9 47.5 47.4 ND Number experiments participated in 2.02 (1.9) 2.1 (2.4) 2.0 (1.9) ND Age 22.1 (3.8) 21.3 (2.3) 21.6 (3.5) 21.9 (3.5) Feedback instructions 0.84 (0.85) 0.8 (0.6) 0.6 (0.6) 0.5 (0.6) (0=very clear, 3= very difficult) Level of knowledge about degradation of peat soils in CH; (0 = 2.4 (0.7) 2.2 (0.9) 2.6 (0.6) 2.5 (0.7) great deal, 3 = nothing) Opinion of peat soil degradation among the players who have knowledge of the issue; (1= not a problem at all, 4= a very 1.8 (0.5) 1.9 (0.5) 1.8 (0.7) 1.7 (0.6) serious problem) Index of Altruism (in %) 46.5 (9.9) 46.1 (11.4) 46.5 (11.0) 48.3 (11.3) Index of Care for the environment (in %) 62.7 (13.3) 62.6 (13.2) 61.7 (12.4) 62.8 (11.7) Financial risks; (0 =avoid, 10 =willing to take financial risks) 3.3 (2.2) 3.4 (2.3) 3.3 (2.0) 3.5 (2.6) Risks; (0 = avoid, 10=willing to take risks) 4.7 (2.2) 4.9 (2.3) 4.8 (2.1) 5.0 (2.3) Measure of time preference (money amount) 706.5 597 (472) 858 (938) 714 (707) (outliers excluded) (694.3) Impatience; (0 = very impatient, 10 = very patient) 5.8 (2.2) 5.7 (2.3) 5.8 (2.5) 5.8 (2.6) Care for reputation; (1= care a lot, 4 = don’t care at all) 1.3 (0.7) 1.5 (0.7) 1.6 (0.7) 1.4 (0.8) Environmental scheme preferences: % of players who prefer 10.5 16.25 15.4 17.8 individual participation as compared to collective participation Opinion of cooperative approaches in Agricultural management ( 0.8 (0.6) 0.6 (0.7) 0.8 (0.5) 0.8 (0.7) 0=more successful than other approach, 3=not at all) Gender (% females) 48.7 58.8 60.3 46.0 SVO angle 22.4 (13.3) 22.3 (13.4) 24.3 (13.7) 22.2 (12.7)

144