ECOSYSTEM SERVICES, HUMAN WELL-BE ING, AND POLICIES IN COUPLED HUMAN AND NATURAL SYSTEMS By Wu Yang A DISSERTATION Submitted to Michigan State University In partial fulfillment of the requirements for the degree of Fisheries and Wildlife Œ Doctor of Philosophy 2013 ABSTRACT ECOSYSTEM SERVICES, HUMAN WELL-BE ING, AND POLICIES IN COUPLED HUMAN AND NATURAL SYSTEMS By Wu Yang Over the past decades, human activities have led to unprecedented biod iversity losses and socioeconomic costs. Unless effective changes in policies, institutions, and practices are made, the deterioration is predicted to be even graver in the future. The fundamental challenge to reverse the situation for achieving both environm ental and socioeconomic sustainability lies in improving the understanding and management of human-nature interactions. To address such challenge, this dissert ation focuses on improving the understanding of linkages between ecosystem services (ES) and human well-being (HWB), and examining complex policy effects on both ES and HWB. Specific objectives are to: (1) develop an integrated approach to understand the linkages between ES and HWB; (2) understand the effects and underlying mechanisms of i ndirect and direct drivers, including group size, on collective action and ES management outcomes; (3) test th e interaction effects of different policies on HWB; (4) understand the effects of payments for ecosystem services (PES) programs on both ES and HWB; and (5) examine the effects of the pos t-disaster reconstruction policy on both ES and HWB. To achieve my objectives, I chose Wolong Nature Reserve and the adjacent Sanjiang Township in Sichuan Province, southwestern China as my study areas. Combining long-term data from household surveys, field plots, and re motely sensed images as well as extensive local knowledge, I used various methods (e.g., ordinary least-squared regression, Tobit models, instrumental variable analysis, confirmatory f actor analysis, structural equation models, and spatial autoregressive models) to test hypotheses and answer research questions. Major findings from this dissertation include: (1) the constr uction of quantitative indicators for ES and HWB as well as integrated models is a viable appr oach in forwarding the understanding of linkages between ES and HWB. Such integrated approach also generated some important findings. For example, those who are mo re vulnerable to disasters are disadvantaged households with lower access to multiple forms of capital, more property damages, or larger revenue reductions. Diversifying human dependence on ES helps to alleviate disaster impacts on HWB; (2) group size has nonlinear effects on both collective action and resource outcomes, with groups of intermediate size contributing the most effort and leading to the best outcomes; (3) there are synergistic and antagonistic effects among conservation and/or development policies, which can even lead to unanticipated consequences; (4) the Natural Fore st Conservation Program (NFCP) had an overall positive effect on ES, a nd mixed effects on local livelihood. To enhance the performance of PES programs, it is importa nt to adapt to local conditions and integrate mechanisms in policy design and implementation; (5) the effects of post-disaster reconstruction efforts can differ from one scale to another. Th erefore, capacity buildi ng and recovery require integrated planning and implementation targeting each form of capital at multiple scales. Advances in methodology and scientific know ledge from this dissertation may also be applied to study and manage other coupled human and natural systems (CHANS). Hopefully the accumulated knowledge from a set of literature will lead to coherent theories (e.g., human-nature feedbacks theory, vulnerability/resilience theo ry of CHANS) to guide the management of human-nature interactions. iv It is easy to make an argument but much more difficult to make a difference. To those who strive to make a difference for both nature and people. v ACKNOWLEDGEMENTS Without the help and support of many peopl e, this dissertation would not have been completed successfully and rapidly. I would like to first thank those local people at Wolong Nature Reserve and Sanjiang Township for their participation in our surveys across years, and willingness to share with me their local knowledge and life experiences. I am grateful for the support and guidance from all my committee members. I thank Dr. Jianguo (Jack) Liu, my major advisor, for hi s continuous support and encouragement to implement my ideas. I also thank him for his high expectation and for always encouraging me to aim higher and achieve greater. I thank Dr. Th omas Dietz for his technical guidance and insightful suggestions. I also appreciate him for sparing precious time and efforts to provide me with constructive, meticulous, and rigorous commen ts and edits on our collaborative chapters. I thank Dr. Andrés Viña for his endless advice and passion in discussing with me in all aspects as a friend. I also thank him for his pursuit of excellence in our collaborative chapters. I thank Dr. Daniel Boyd Kramer for his insi ghtful suggestions in designing the human well-being instrument. I also thank him for his timely response and cons tructive comments on our collaborative chapters. I would also like to thank the logistic help and support from Wolong Administrative Bureau and the Research Center for Eco-Envi ronmental Sciences at Chinese Academy of Sciences. Special acknowledgements go to Hemi n Zhang, Zhiyun Ouyang, Weihua Xu, Yi Xiao, Yan Yan, Xiaogang Shi, Meng Ming, Dong Chen, Senlong Jin, Jian Yang, Biyou Yang, Bihong Yang, Lun Wang, Bo Wang, Maomi ng Wang, Jifu Wang, Datian He, Jianping Liu, Mingchong Liu, Weihong Tan, Changyou Yang, Liang Zhang, Kaiq iang Zhang, Zhijun Zh ang and Sha Zhou. I would also thank Lin Wang for inputting part of the data. vi I also appreciate the inspiring discussion and emotional support fr om many current and previous members in Center for Systems Integrat ion and Sustainability, especially Neil Carter, Xiaodong Chen, Ken Frank, Guangming He, Vane ssa Hull, Shuxin Li, Yu Li, Junyan Luo, William McConnell, Sue Nichols, William Ta ylor, Mao-Ning Tuanmu, and Jindong Zhang. I would also like to thank my family and fr iends. I thank all my family members for their unconditional support, apprehension, and love. I would also thank one of my best friends Qing Zhang who shares with me both positive and negativ e emotions, always stands by me, trusts me, encourages me, and feels proud of me. Any achievement that I made from the completion of this dissertation is also theirs. Finally, I gratefully acknowledge financial supports from the National Science Foundation, and National Aeronautics and Space Administration of United States, and the Environmental Science and Policy Program, College of Agriculture and Natural Resources, and Graduate Office at Michigan State University. vii TABLE OF CONTENTS LIST OF TABLES ................................................................................................................ .......... x LIST OF FIGURES ............................................................................................................... ...... xiii CHAPTER 1 INTRODUCTION .................................................................................................................. ........ 1 1.1 Background ................................................................................................................ ............2 1.2 Research objectives ....................................................................................................... .........6 CHAPTER 2 GOING BEYOND THE MILLENNIUM ECOSYSTEM ASSESSMENT: AN INDEX SYSTEM OF HUMAN DEPENDENCE ON ECOSYSTEM SERVICES ..................................................... 9 Abstract .................................................................................................................................10 2.1 Introduction .............................................................................................................. ............11 2.2 Methods for developing an index system of human dependence on ecosystem services ....15 2.2.1 Conceptualization of the index system ......................................................................... 15 2.2.2 Methods for constructing the index system .................................................................. 17 2.3 Empirical demonstration of co nstructing the index system .................................................19 2.3.1 Description of the demonstration area .......................................................................... 19 2.3.2 Data and methods ........................................................................................................ . 21 2.3.3 Results of the index system .......................................................................................... 24 2.3.4 Advantages and applications of the index system ........................................................ 25 2.4 Discussion and conclusions ................................................................................................ ..27 CHAPTER 3 GOING BEYOND THE MILLENNIUM ECOSYSTEM ASSESSMENT: AN INDEX SYSTEM OF HUMAN WELL-BEING ........................................................................................................ 36 Abstract..................................................................................................................................37 3.1 Introduction .............................................................................................................. ............38 3.2 Methods ................................................................................................................... .............42 3.2.1 Ethics statement ........................................................................................................ .... 42 3.2.2 Development of the human well- being index (HWBI) system .................................... 42 3.2.3 Description of the demonstration sites ......................................................................... 44 3.2.4 Household surveys ....................................................................................................... . 46 3.3 Results ................................................................................................................... ...............47 3.3.1 Internal consistency of the HWBI system .................................................................... 47 3.3.2 CFA results of the HWBI system ................................................................................. 47 3.3.3 Impacts of the earthquake on HWBI ............................................................................ 48 3.4 Discussion ................................................................................................................ ............49 3.5 Conclusions and implications .............................................................................................. 51 viii CHAPTER 4 AN INTEGRATED APPROACH TO UN DERSTAND THE LINKAGES BETWEEN ECOSYSTEM SERVICES AND HUMAN WELL-BEING ....................................................... 58 Abstract..................................................................................................................................59 4.1 Introduction .............................................................................................................. ............60 4.2 Materials and Methods ..................................................................................................... ....64 4.3 Results ................................................................................................................... ...............66 4.3.1 Simultaneous impacts of the ea rthquake on IDES and HWBI. .................................... 66 4.3.2 Effects of IDES on change of HWBI. .......................................................................... 67 4.3.3 Effects of other indirect and direct drivers on change of HWBI. ................................. 68 4.4 Discussion ................................................................................................................ ............69 CHAPTER 5 NONLINEAR EFFECTS OF GROUP SIZE ON COLLECTIVE ACTION AND RESOURCE OUTCOMES................................................................................................................................. 77 Abstract..................................................................................................................................78 5.1 Introduction .............................................................................................................. ............79 5.2 Materials and Methods ..................................................................................................... ....83 5.3 Results ................................................................................................................... ...............85 5.4 Discussion ................................................................................................................ ............88 CHAPTER 6 INTERACTION EFFECTS OF CONSERVATION AND DEVELOPMENT POLICIES ON RURAL HOUSEHOLD INCOME AND INCOME STRUCTURE ............................................ 97 Abstract..................................................................................................................................98 6.1 Introduction .............................................................................................................. ............99 6.2 Methods ................................................................................................................... ...........101 6.2.2 Data collection and analyses....................................................................................... 104 6.2.3 Econometric Models ................................................................................................... 10 5 6.3 Results ................................................................................................................... .............106 6.4 Discussion ................................................................................................................ ..........108 CHAPTER 7 PERFORMANCE AND PROSPECTS OF PA YMENTS FOR ECOSYSTEM SERVICES PROGRAMS: EVIDENCE FROM CHINA .............................................................................. 120 Abstract...............................................................................................................................121 7.1. Introduction ............................................................................................................. ..........122 7.2. Materials and methods .................................................................................................... ..124 7.2.1. Study area ............................................................................................................. ..... 124 7.2.2. Forest cover dynamics ............................................................................................... 12 7 7.2.3. Focus group, individual and household interviews ................................................... 128 7.2.4. Local adaptation and implementation of the NFCP .................................................. 129 7.2.5. Baseline for environmental benefits .......................................................................... 131 7.3. Results .................................................................................................................. .............133 7.3.1. Environmental effects of the NFCP implementation ................................................ 133 7.3.2. Socioeconomic effects of NFCP implementation ..................................................... 135 7.4. Discussion ............................................................................................................... ..........136 ix 7.4.1. Lessons learned through NFCP implementation ....................................................... 138 7.4.2. Challenges and Opportunities .................................................................................... 141 7.5. Conclusions .............................................................................................................. .........145 CHAPTER 8 DYNAMICS OF HOUSEHOLD CAPITAL IN PROTECTED AREAS DURING POST- DISASTER RECONSTRUCTION ............................................................................................ 154 Abstract................................................................................................................................155 8.1 Introduction .............................................................................................................. ..........156 8.2 Materials and methods ..................................................................................................... ..159 8.2.1 Study area .............................................................................................................. ..... 159 8.2.2 Data collection and analyses....................................................................................... 160 8.3 Results ................................................................................................................... .............161 8.4 Discussion ................................................................................................................ ..........163 8.4.1 Causes and effects for observed changes in household capital .................................. 164 8.4.2 Policy recommendations............................................................................................. 166 8.5 Conclusion ................................................................................................................ ..........169 CHAPTER 9 CONCLUSIONS........................................................................................................................ 181 APPENDICES .................................................................................................................... ........ 186 APPENDIX A SUPPORTING INFORMATION FOR CHAPTER 2 ...................................187 APPENDIX B SUPPORTING INFORMATION FOR CHAPTER 3 ....................................189 APPENDIX C SUPPORTING INFORMATION FOR CHAPTER 4 ....................................194 APPENDIX D SUPPORTING INFORMATION FOR CHAPTER 5 ...................................198 APPENDIX E SUPPORTING INFORMATION FOR CHAPTER 6 ....................................239 REFERENCES .................................................................................................................... ....... 247 x LIST OF TABLES Table 2.1. Net benefits, overall IDES and sub-indices in 1998 and 2007. ................................... 32 Table 2.2. Comparison of overall IDES and agricultura l income share for thei r associations with gross household income. ....................................................................................................... ........ 33Table 2.3. Regression of sources of variation on overall IDES. ................................................... 34 Table 3.1. Summary of model fit information fo r the confirmatory factor analysis. ................... 53 Table 3.2. Descriptive statistics of sub-indices and overall HWBI both inside and outside the reserve before and after the earthquake. ...................................................................................... . 54Table 3.3. Impacts of the earthquake on sub-indices and overall HWBI inside and outside the reserve. ...................................................................................................................... .................... 55Table 4.1. Factors associated with the change of HWBI before and after the earthquake. .......... 72 Table 4.2. Example hypotheses and questions for fu ture research on human-nature interactions........................................................................................................................................................ 73Table 5.1. Coefficients of the Tobit model for the nonlinear effect of group size on collective action. ....................................................................................................................... ..................... 91Table 5.2. Coefficients of the spatial autoregressive error model for the nonlinear effect of group size on resource outcomes. .................................................................................................... ....... 92Table 5.3. Path analysis of the two opposing forc es through which group size affects collective action. ....................................................................................................................... ..................... 93Table 6.1. General information of the four c onservation and development policies in Wolong Nature Reserve. ............................................................................................................... ............ 111Table 6.2. Effects of conservation and devel opment policies on changes in total household income. ....................................................................................................................... ................. 113Table 6.3. Effects of conservation and developmen t policies on changes in agricultural income percentage. .................................................................................................................................. 115Table 7.1. General information about the PES programs in Wolong Nature Reserve. .............. 147 Table 7.2. Paired t-test for indicators of energy use before and after the NFCP implementation. ..................................................................................................................................................... 148Table 8.1. The definition and selected indi cators of four forms of capital. ................................ 171 xi Table 8.2. Descriptive statistics of cropland per household inside and outside the Reserve. ..... 172 Table S2.1. Detailed classification of household net income and a voided costs by type of related ecosystem services............................................................................................................. ..........186 Table S3.1. The index system for assessing human well-being based on the Millennium Ecosystem Assessment conceptual framework...........................................................................188 Table S3.2. Standardized coefficients of the confirmatory factor analys is for Human Well-Being Index (HWBI)..............................................................................................................................191 Table S4.1. Descriptive statistics of variables used in the model................................................193 Table S4.2. Impacts of the earthquake on net be nefits from socioeconomic activities and ecosystem services............................................................................................................. ..........194 Table S4.3. Associations between HWBI and affluence before and after the earthquake..........195 Table S4.4. Supplementary regressions for factors associated with the change of HWBI before and after the earthquake...............................................................................................................196 Table S5.1. Descriptive statistics of variables for 156 randomly sampled monitoring households....................................................................................................................................224 Table S5.2. Correlation between group size and other biophysical, demographic, and socioeconomic variables........................................................................................................ ......224 Table S5.3. Descriptive statistics of variables for 151 household-monitored parcels.................225 Table S5.4. Descriptive statistics of characteristics of 113 randomly sampled monitoring households and their assigned groups..........................................................................................225 Table S5.5. Path analysis of the two opposing forces through which group size affects collective action............................................................................................................................................226 Table S5.6. Characteristics of households with strong and weak social ties to local leaders......227 Table S5.7. Logit estimation of factors associat ed with social ties to local leaders....................228 Table S5.8. Tobit models for hypothesi zed instrumental variables.............................................229 Table S5.9. Coefficients of the Tobit model for the nonlinear effect of group size on collective action............................................................................................................................................230 Table S5.10. First-stage regression results of the two-step Tobit model.....................................231 Table S5.11. Second-stage regression results of the two-step Tobit model................................232 xii Table S5.12. Variance inflation factors for variables used in the Tobit model examining the nonlinear group-size effect (Table 5.1 or Table S5.9).................................................................232 Table S5.13. Different combinations of Tobit models for the nonlinear effect of group size on collective action...........................................................................................................................233 Table S5.14. Coefficients of the spatial autoregressive error model for the linear effect of group size...............................................................................................................................................234 Table S5.15. Coefficients of the multiple linea r regression for the nonlinear effect of group size...............................................................................................................................................235 Table S5.16. Coefficients of the spatial autoregressive error model for the nonlinear effect of group size on resource outcomes................................................................................................ .236 Table S5.17. Variance inflation factors (VIFs) for variables used in the spatial simultaneous autoregressive error model (Table 5.1 or Table S5.16)...............................................................237 Table S6.1. Descriptive statistics of variables used in modeling the effects of conservation and development policies on changes in total household income and income structure...................240 Table S6.2. Complementary results of the effect s of conservation and development policies on changes in total household income.............................................................................................. 242 Table S6.3. Complementary results of the effect s of conservation and development policies on changes in the percentage of agricultural income.......................................................................244 xiii LIST OF FIGURES Figure 2.1. Wolong Nature Reserve in Sichuan Provi nce, southwestern China. For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this dissertation. ................................................................................................. .......... 35Figure 3.1. Wolong Nature Reserve and adjacent Sanjiang Township in Wenchuan County, Sichuan Province, southwestern China. ........................................................................................ 56Figure 3.2. Path diagram of the confirmatory factor anal ysis model for HWBI. ......................... 57 Figure 4.1. Wolong Nature Reserve in Wenchuan County, Sichuan Province, southwestern China. ....................................................................................................................................................... 74Figure 4.2. Impacts of the earthquake on huma n dependence on ecosystem services. ................ 75 Figure 4.3. Impacts of the earthquake on human well-being. ....................................................... 76 Figure 5.1. Hypothetical effects of free riding, within-group enforcement, and group size on collective action and resource outcomes. ...................................................................................... 94 Figure 5.2. Map of the location, main road, fo rest cover in 2007, and household monitoring parcels of Wolong Nature Reserve in Sichuan Province, China. ................................................. 95 Figure 5.3. The nonlinear group-size effects on collective action and forest outcomes. .............. 96 Figure 6.1. Hypothetical outcomes of interacti on effects among diffe rent policies. .................. 117 Figure 6.2. Wolong Nature Reserve in Sichuan Province, southwestern China. ........................ 119 Figure 7.1. Map of the Wolong Nature Reserve. ........................................................................ 149 Figure 7.2. Natural Forest Conservation Program (NFCP) implementation in the Wolong Nature Reserve (WNR). ................................................................................................................ .......... 150Figure 7.3. Forest cover area before and after the Natural Forest Conservation Program (NFCP) implementation in 2001. ....................................................................................................... ...... 151Figure 7.4. Perceptions of local households toward the Natural Forest Conservation Program (NFCP). ....................................................................................................................................... 152Figure 8.1. Map of Wolong Nature Reserve and the adjacent Sanjiang Township. ................... 173 Figure 8.2. Box plots of household expenditures on food (a), education (b), and medical care (c) before and after the earthquake at Wolong Nature Reserve. ...................................................... 174 xiv Figure 8.3. Dynamics of the house type inside (a ) and outside the Reserve (b) from 1970s to 2010s. .......................................................................................................................................... 176Figure 8.4. Dynamics of the total constructed area per household (a) and loans per household for post-disaster housing repair or reconstruction (b). ..................................................................... 177 Figure 8.5. Total amount of social support and social support per lender before and after the earthquake both inside and outside the Reserve. ........................................................................ 178 Figure 8.6. Proportions of external and internal sources of social support for households before and after the earthquake both inside and outside the Reserve. ................................................... 179 Figure 8.7. Dynamics of the gross household inco me inside and outside the Reserve from 1970s to 2010s. ..................................................................................................................... ................. 180 1 CHAPTER 1 INTRODUCTION 2 1.1 Background Over the past decades, the Earth™s ecosystems have experienced unprecedented degradation due to human activitie s, which has resulted in a substantial and largely irreversible loss of biodiversity (MA, 2005). The dramatic changes to ecosystems have substantially contributed to human we ll-being (HWB) and economic development at the costs of degradation of many ecosystem services (ES), increased risk s of nonlinear changes, and the exacerbation of poverty of some groups of people (Carpenter et al., 2006; MA, 2005). It is predicted that the deteriorating of ecosystems could become a grave problem during the first half of this century (MA, 2005). Unless effective changes in policies, institutions, and practices can be made, this degradation would be a barrier to achieve sustainability identified by the United Nations Millennium Development Goals (MA, 2005). Th e fundamental challenge for achieving both environmental and socioeconomic sustainabil ity lies in improving the understanding and management of human-nature interactions (Car penter et al., 2009; Liu et al., 2007a; Ostrom, 2009). ES science has recently become a popular paradigm toward improving the understanding and management of human-nature in teractions. Studies related to this field date back to at least the early 1980s (Fisher et al., 2009). But the paradigm began to form in late 1990s with the publication of the influential book fiNature™s Servicesfl (Daily, 1997) and the valuation of world™s ES and natural capital (Costanza et al., 1997), as well as the implementation of the payments for ES program in Costa Rica (Sanchez- Azofeifa et al., 2007). Publications using the term fiecosystem servicesfl, fienvironmental se rvicesfl or fiecological servicesfl are soaring exponentially after the monumental Millennium Ecosystem Assessment (Fisher et al., 2009; MA, 2005). The significance of this field has gained increasing acceptance and application since the 3 Millennium Ecosystem Assessment, as well as se veral important projects such as the Natural Capital Project (Kareiva et al., 2011), and th e Economics of Ecosystems and Biodiversity (TEEB) project (TEEB, 2010). The ES paradigm is characterized as assessi ng benefits from nature in biophysical and/or monetary values, and realigning socioeconomic be nefits/costs using market-based mechanisms through engaging different stakeholders (e.g., farmers, governments, enterprises, non- government organizations) at multiple scales (J ack et al., 2008; Tallis et al., 2008). In my opinion, the reasons for the popularity of the ES paradigm lie in two main aspects of them. First, the ecosystem services approach links natural systems and human systems together as coupled human and natural systems (CHANS) (Liu et al ., 2007a; Liu et al., 2007b). Specifically, it links ecosystem changes to human well-being and thus provides an operational template for decision making and real-world practices. Second, the ES framework provides an excellent platform for interdisciplinary studies engaging almost all disciplines ranging from natural science to social science and to the humanities. Despite of the rapid progress in ES research, so far most efforts have been made to assess and map ES for designing and implementing payments for ecosystem services (PES) programs (Daily and Matson, 2008; Engel et al., 2008; Liu et al., 2008; Liu et al., 2013; Nelson et al., 2009). The quantitative understanding of the linkages between ES and HWB remains poor, primarily due to the lack of qua ntitative methods and indicators as well as robust theories explaining how ecosystem changes affect human society (Carpenter et al., 2006; Carpenter et al., 2009; Dietz et al., 2009). The poor understanding of the linkages between ES and HWB poses serious threats to the rapidly expanding PES programs and other conser vation projects (e.g., Integrated Conservation 4 and Development Projects [ICDPs]). For instance , two thirds of ICDPs implemented between 1993 and 2007 by the World Bank had failed to achieve their conservation and development dual goals, largely due to the neglect of linkages between ES and HWB across scales (Tallis et al., 2008). The landmark work by Ostrom and others also pointed out that the pervasive failures of conservation policies were primarily due to igno rance of the complex linkages between ES and HWB (Ostrom, 1990, 2005, 2007; Os trom, 2009; Ostrom et al ., 2007). The PES programs are with no exception to such risk. First, economic va lues are only a subset of intrinsic values of ecosystem and an economically driven focus on the values of ecosystems for human needs may be detrimental to long-term survival of nonhuman components of ecosystems (Redford and Adams, 2009). Second, ES substantially but not exclusively contribute to HWB. There are many other factors (e.g., life experience, personality, culture, legal frameworks in which one lives) affect the subjective feelings of humans (Diener and Ryan, 2009; Diener et al., 1999). Third, payments to protect one service may adversel y affect the provision of other services. For example, payments for natural forest conserva tion may promote carbon sequestration, reduce soil erosion and encourage wildlife pr oliferation, but may also increase agricultural losses to wildlife disturbance (Liu et al., 2013). Fourthly, there is a distinction between protection of ES and biodiversity because there are ecosystems (e.g., ag ricultural systems) provide important services but are not priority biodiversity hotspots. Finally, the distribution of benefits provided by ES is often unequal across different population groups. In comparison to the affluent, poor people may be more dependent on the provision of ES but they usually have less control of them (Carpenter et al., 2006; Redford and Adams, 2009). Therefore, the effectiveness of PES programs and other conservation policies will largely depend on improving the understanding and management of the linkages between ES and HWB. 5 While numerous conservation and developm ent policies have been designed and implemented to protect ES and improve HWB, relatively few studies (Arriagada et al., 2009; Gross-Camp et al., 2012; Scu llion et al., 2011) have simulta neously evaluated both the environmental and socioeconomic outcomes. Perh aps one reason is because of the lack of understanding and consideration of human-nature interactions discussed above. Another reason perhaps is because of the lack of systematic collection of both environmental and socioeconomic data before the policy implementa tion. Finally, it is difficult to identify a counterfactual without-policy baseline scenario if the policy has already been implemented. Moreover, there is also a lack of systematic and quantitative studies on underlying mechanisms of how different drivers lead to corresponding environmental and socio economic outcomes. For instance, it is often assumed that conditional cooperators (i.e., me mbers who will contribute more for guarding public good under the condition that others also c ontribute more) and costly monitoring have positive effects to management outcomes of co mmon-pool resources. However, until recently this proposition and the underlying mechanisms ha ve not been confirmed with systematically collected data in real-world practice (Rustagi et al., 2010). The effects and underlying mechanisms of some other factors, such as soci al networks and group size (i.e., the number of group members), on collective action and commo n-pool resource outcomes remain elusive (Ostrom, 2005; Tucker, 2010). In addition, interactions among different policies are mostly i gnored in previous research (Liu et al., 2008; Liu et al., 2013). There are several good reasons to consider these interactions. First, there are usually concurrent conservati on and development policies across space and the evaluation of each policy may be biased if th ere are significant interaction effects among multiple policies (Liu and Yang, 2012; Ward a nd Pulido-Velazquez, 2008). Second, it will be 6 difficult to achieve different goals of multiple po licies simultaneously without considering their interactions. The above discussion on the trade-offs among different ES for PES programs is evidence of this in reality. Third, for policies with dual goals (e.g., ICDPs) both for ES and HWB, the linkages between ES and HWB mean that there coul d also be interactions among different policy implementation strategies that separately emphasize ES or HWB. Finally, interactions among multiple policie s may generate unanticipated consequences through feedback loops across scales. For instance, while an efficient irrigation program reduced water use on individual farms, it increased evapotranspiration (i.e., loss of water associated with plant use), reduced groundwater replenishment, and redistribut ed water supply at a water basin scale, which actually increased overall water use at the water basin scale and beyond (Ward and Pulido-Velazquez, 2008). A PES program that provided incentives encourage the expansion of the efficient irrigation program also cannot reduce overall water use alone. The solution is to combine the two policies with in stitutional innovations that account water use at multiple scales aiming to reduce overall water use (Ward and Pulido-Velazquez, 2008). 1.2 Research objectives To fill knowledge gaps discussed above, the overall goal of my dissertation is to improve the understanding and management of human-nature interactions. Specific research objectives include: (1) Developing an integrated approach to understand the linkages between ecosystem services and human well-being (Chapters 2, 3 and 4); 7 (2) Understanding the effects and underlying mechanisms of i ndirect and direct drivers, especially group size (i.e., the number of gr oup members), on collective action and ecosystem service management outcomes (Chapter 5); (3) Testing the interaction e ffects of different policies on human well-being (Chapter 6); (4) Understanding the effects of payments for ecosystem services programs on both ecosystem services and human well-being (Chapter 7); (5) Examining the effects of the post-disaster reconstruction policy on both ecosystem services and human well-being (Chapter 8). To achieve these objectives, I c hose Wolong Nature Reserve (WNR; N 30û45' Œ 31 û25', E 102û52' Œ 103 û24') for Giant Pandas ( Ailuropoda melanoleuca) as my focus study area. Per the needs of specific research questions in some chapte rs (Chapter 3 and 8), I also chose the adjacent Sanjiang Township (SJT) as a comparison site. Th ere are three main reasons for such selection. The first reason is that there are long-term exis ting data from household surveys and field plots dating back to 1998. There are also remotely se nsed images (e.g., Landsat TM and EM) tracing back to 1960s. The rich datasets and accumu lated local knowledge lay a good foundation for systematic experimental design and rigorous statistical analyses supporting reasonable causal inference. The second reason is that both WNR a nd SJT are of high ecological importance. They both belong to the Sichuan Giant Panda Sanctuaries of the United Nations E ducational, Scientific and Cultural Organization (UNESCO) World Herita ge System. The Sanctuary was established in 2006 to protect the giant panda habitat (Li et al., 2013). It also has been classified as one of the world™s top 25 Biodiversity Hotspots (Myers et al., 2000) and one of the Global 200 Eco-regions identified by the World Wide Fund for Natu re (WWF) (WWF, 2007). Th ird, both sites have 8 local residents whose ancestors settled the area hundreds of years ago. Local residents interact with nature and depend on many ES (e.g., cultivation of maize, pot atoes, yaks, pigs, cattle, and collection of fuelwood, mushrooms, and tradit ional Chinese medicines) long before the establishment of the Sanctuary. Since the late 1990s, several national and local conservation and development policies (e.g., Natural Forest Conser vation Program, Grain-to -Green Program, and Tourism Development Plan) have been impl emented in this area. On May 12, 2008, the devastating Wenchuan Earthquake (Ms 8.0 or Mw 7.9) also hit both sites, with the epicenter immediately near the two sites. In response to tremendous socioeconomic and environmental impacts caused by the earthquake, a national r econstruction plan (i.e., Post-earthquake Reconstruction Plan) has also been implement ed. All these conservation and development policies may lead to dramatic impacts on both lo cal ES and HWB, which again provide excellent opportunities to address my objectives. Because all the main chapters (i.e., Chapter 2 to 8) have been published in or have been prepared for peer-reviewed journals, more detailed background information of the study areas including details on different conservation and development policie s are provided in each of the main chapters. 9 CHAPTER 2 GOING BEYOND THE MILLENNIUM ECOSYSTEM ASSESSMENT: AN INDEX SYSTEM OF HUMAN DEPENDENCE ON ECOSYSTEM SERVICES In collaboration with Thomas Dietz, Wei Liu, Junyan Luo, Jianguo Liu 10 Abstract The Millennium Ecosystem Assessment (MA) estimated that two thirds of ecosystem services on the earth have degraded or are in decline due to the unprecedented scale of human activities during recent decades. These changes will have tremendous consequences for human well-being, and offer both risks and opportunities for a wide range of stakeholders. Yet these risks and opportunities have not been well managed due in part to the lack of quantitative understanding of human dependence on ecosystem services. Here, we propose an index of dependence on ecosystem services (IDES) system to quantify human dependence on ecosystem services. We demonstrate the construction of th e IDES system using household survey data. We show that the overall index and sub-indices can reflect the general pattern of households™ dependences on ecosystem services, and their vari ations across time, space, and different forms of capital (i.e., natural, human, financial, manuf actured, and social capit als). We support the proposition that the poor are more dependent on ec osystem services and further generalize this proposition by arguing that those disadvantaged groups who possess low levels of any form of capital except for natural capital are more depe ndent on ecosystem services than those with greater control of capital. The higher value of the overall IDES or sub-index represents the higher dependence on the corresponding ecosystem services, and thus the higher vulnerability to the degradation or decline of corresponding ecosystem services. The IDES system improves our understanding of human dependence on ecosystem services. It also provides insights into strategies for alleviating poverty, for targeting priority groups of conservation programs, and for managing risks and opportunities due to changes of ecosystem services at multiple scales. 11 2.1 Introduction The Millennium Ecosystem Assessment (MA) was designed to assess the consequences of ecosystem change and provide scientific information that could aid in sustainably managing ecosystems for human well-being (MA, 2005). Although the intended audience was decision-makers, MA also provided a conceptual fram ework for studying interactions among four key components (i.e., indirect drivers, direct drivers, ecosystem serv ices, and human well-being) of coupled human and natural systems (CHANS) (Liu et al., 2007a), and identified future research needs (Carpenter et al., 2006). Of the interactions between the four co mponents in the MA framework, the linkage between ecosystem services and human well-bein g is perhaps least understood. The relationship between human well-being and the social factors that influence it has been extensively studied (Abdallah et al., 2008; Campbell, 1976; Diener, 2000; Diener et al., 1999; Grant et al., 2009). Through efforts like land change science and structural human ecology, the relationships between changes of ecosystems serv ices and factors that influence them have also begun to be documented from local to regional and global scale (Rosa et al., 2010; Turner et al., 2008). It has been recognized that humans substantially depend on ecosystem services, which range from basic provisions of food, fresh water and fuel, through regulation of water and air quality, to cultural services like ecotourism (Daily, 1997; MA , 2005). The MA also established that during recent decades, two thirds of these services ha ve degraded or are in decline due to the unprecedented scale of human activities (MA, 2005) . But what are the consequences of such dramatic degradation to short-term and long-te rm human well-being? The risks of ecosystem degradation and their consequences for human well-being, including nonlinear or abrupt changes, 12 are poorly quantified. On the one hand, there is a lack of a robust theory that links ecological diversity to ecosystem dynamics and ecosystem se rvices (Carpenter et al., 2006). On the other hand, the scientific community lacks understandi ng of how and to what extent humans depend on ecosystem services. For instance, it has been widely recognized that the poor are most dependent on ecosystem services and most vulnera ble to the degradation of ecosystem services (MA, 2005); however, it is generally not known how such dependence differs across time, space and various population groups (e.g., across income levels). To better understand, monitor and manage such dependences, a quantitativ e approach is urgently needed. From our perspective, there are at least four reasons to quantify human dependence on ecosystem services. First, th e relationship between ecosystem services and poverty seems obvious but the dependence of the poor on ecosystem services is ra rely quantified, which leads to a pervasive tendency to overlook it in statistics, poverty assessment s and natural resource management decisions (Shackleton et al., 2008). A quantitative measurement of such dependence and its integration into decision maki ng could reverse inappropriate strategies that could otherwise lead to further marginalization of the poor and increased pressure on ecosystem services. Second, benefits provided by ecosystem servi ces are often unequally distributed across different population groups and there may be trade-offs among groups. With better understanding of the distribution of benefits from ecosystem services (e.g., fuelwood, clean water, non-timber forest products, and tourism) across differe nt population groups, conservation and development programs may be better designed to guide the flow of benefits from ecosystem services to target priority population groups. The Wolong Nature Reserve of China provides a compelling example. Most benefits obtained fr om ecotourism flow to the outside tourism 13 development companies rather than local hous eholds (He et al., 2008; Liu et al., 2012), a common phenomenon in many other areas (Kiss, 2004). Government policies should encourage household relocation closer to t ourism facilities and provide more support to local households (e.g., provide training to improve human capital and offer favorable loan opportunities) to enhance their capacity for participating in tourism businesses (He et al., 2008; Liu et al., 2012). In doing so, more benefits from tourism would flow to local households and substantially reduce their pressure on provisioning services by which local ecosystems provide fuelwood, bamboo shoots, and traditional Chinese medicine. Third, a better understanding of such depe ndence would draw attention to currently unmanaged risks and unrealized opportunities that come with ecosystem change. For example, agricultural supply chains can be tightly dependent on ecosystem services and thus are vulnerable to dramatic ecosystem degradation. Unprecedented human activit ies would likely lead to more frequent extreme climate events and natural disasters (e.g., storms, floods, droughts, and landslides), cause tremendous destruction to eco systems and their services, and threaten the livelihoods of those people who are highly dependent on corresponding ecosystem services (MA, 2005; Rosa et al., 2010). Yet few managers or policy analysts understand this dependence and related unintended consequences, and even fewer manage the potential risks and opportunities (Grigg, 2008; Liu and Yang, 2012). Fourthly, a quantitative measurement of such dependence would improve the understanding of human-nature in teractions. One of the major advances and challenges of the CHANS approach for studying human-nature interactions is to construct coupled models by integrating sub-models of both human and natu ral subsystems (McConnell et al., 2011). The key of such integration requires good understanding of the interactions between human and natural 14 subsystems. Currently, there are few coupled models integrating drivers, ecosystem services, and human well-being to systematically understand human-nature intera ctions (Carpent er et al., 2006; Carpenter et al., 2009; Yang et al., 2013a). The quantification of human dependence on ecosystem services could potentially serve as a proxy to facilitate such integration and understanding. The objectives of this study were to (1) propose the conceptual basis of an index of dependence on ecosystem services (IDES) system to measure the degree of human dependence on ecosystem services; (2) demonstrate the construc tion of the IDES system with empirical data; and (3) illustrate advantages and applications of the proposed IDES system. Specifically, we first provided the conceptual basis of an IDES system, including an overall index and sub-indices for different categories of ecosystem services based on the widely accepted MA framework. We then delineated the process of estimating the in dices at Wolong Nature Reserve. We examined temporal changes of the overall IDES and shifts in of the structure of the IDES system (i.e., changes of sub-indices). We compared the overall index with an alternative indicator (i.e., the commonly used agricultural income share) to i llustrate the advantages of our proposed index system. Moreover, we assessed the dependence of the poor on ecosystem services. In particular, we analyzed how households™ dependences on ecosystem services differ across different degrees of access to capitals (i.e., natura l, human, financial, manufactured, and social capitals). We also evaluated the spatial heterogeneity of the overall IDES. 15 2.2 Methods for developing an index system of human dependence on ecosystem services 2.2.1 Conceptualization of the index system The term ecosystem services is defined and used in a variety of ways (Boyd and Banzhaf, 2007; Costanza et al., 1997; Daily, 1997; Farber et al., 2002; MA, 2005; Wallace, 2008; Wallace, 2007). Here we aligned with the definitions of the MA as the benefits that people obtain directly or indirectly from ecosystems, including both na tural systems or highly managed systems (MA, 2005). In particular, we included agricultural products as part of ecosystem services. We acknowledge that some literatures might exclud e products from highly managed systems (e.g., agro-ecosystems and constructed wetlands) and rest rict ecosystem services to goods and services provided by natural systems only. But since the logic of our analysis is driven by the MA, we felt it appropriate to adhere to the definition by the MA. Our proposed index system of dependence on ecosystem services includes an overall inde x and three sub-indices. The overall index of human dependence on ecosystem services is define d as the ratio of net benefits obtained from ecosystems to the absolute value of total net be nefits that derived from ecosystems and other socioeconomic activities (e.g., migr ant work, and small business unrelated to ecosystem services, see Table S2.1). In addition to the overall index, a sub-index can be calculated for each category of ecosystem services under the MA framewor k (i.e., provisioning, regulating services, and cultural services) (MA, 2005). Because supporting se rvices are the bases for other three types of services, following the common pr actice in ecosystem service assessment, they are not included in IDES to avoid double accounting. As shown by the definition, the higher value of the overall index or sub-index represents the higher dependence on the corresponding ecosystem services, and thus the higher vulnerability to the degr adation or decline of corresponding ecosystem services. The equations for calculating three sub-i ndices and the overall IDES are given as below. 16 ||/31iiiiSNBENBENBIDES (2.1) 31iiIDESIDES (2.2) where i is the category of ecosystem services (i.e., provisioning, regulating, and cultural services); IDESi is the sub-index for category i; ENBi is the total net benefit obtained from category i ecosystem services; SNB is the total net benefit obtained from socioeconomic activities; IDES is the overall index. There are four reasons for us ing net benefits instead of gross benefits. First, ecosystems generate both services and dis-services to humans. Dis-services may include pests and diseases causing reduction in agricultural production and other unintended negative health consequences for organisms including humans (Zhang et al ., 2007). Second, the generation and delivery of ecosystem services may entail costs (e.g., costs of seeds, fertilizers, and pesticides for agricultural products). Using the gross benefits could potentially mislead decision making (Naidoo and Ricketts, 2006). One might opt for a pr ogram that has the largest increase in gross benefits when another program has a larger yield of net benefits, thereby choosing an inefficient program. Third, using net benefits allows the inclusion of trade-offs between different ecosystem services (Nelson et al., 2009). Such trade-offs would not be correctly represented if gross benefits are used without considering the costs of delivering t hose services. Finally, using net benefits facilitates cross-contex t comparisons. Few previous ecosystem service assessments have evaluated net benefits (Birch et al., 2010; Chang et al., 2011; Yang et al., 2008). Many previous studies have evaluated only the gross benefits so results from different studies are not 17 comparable because ecosystem dis-services and costs of generating ecosystem services can be substantial and vary considerably across contexts. Both the sub-indices and overall index can be negative. This is because net benefits are not necessarily positive. Total ne t benefits from each category and all categories of ecosystem services summed can be negative. The ecological and economic meaning of an index with negative value is that the gross benefit obtained from ecosystem services is lower than the sum of costs for generating the corresponding ecosystem se rvices and costs of ecosystem dis-services. For example, the gross benefits of producing agri cultural products may be lower than the total costs of seeds, fertiliz ers, and pesticides. 2.2.2 Methods for constructing the index system The index system is constructed to asse ss net benefits of a unit of analysis (e.g., household). The procedures for this approach are in some ways similar to that of many Cost- Benefit Analyses (CBA) (Boardman et al., 2006; Hanley et al., 2001) and Ecosystem Service Assessments (ESA) (Chang et al., 2011; MA, 2005; Nelson et al., 2009; Yang et al., 2008) where data from a variety of sources are aggregated in to an integrated assessment and where the unit of analysis for which the calculation is done must be specified. Fo r CBA and ESA this is often a region or nation, while here we will work at the household level. Where markets for the gross benefits a nd costs exist, assessments are relatively straightforward and simple. It is easy to appl y market-based valuation methods such as the market price method, the apprai sal method, and the avoided cost method (Barbier, 2011; Chee, 2004; Scott et al., 1998; Yang et al., 2008). Otherwise, when mark et data are not available, 18 nonmarket valuation methods such as the contin gent valuation method, th e travel cost method, the stated preference method, and the hedonic pri ce method can be used (Barbier, 2011; Bateman et al., 2011; Scott et al., 1998; Yang et al., 2008).There are also cross-cutting methods, such as the benefit transfer method and unit-day value method, which combine both market-based and nonmarket methods (Ready and Navrud, 2006; Shre stha et al., 2007; Wilson and Hoehn, 2006). Recently, integrated approaches such as the Integrated Valuation for Ecosystem Services and Tradeoffs (InVEST) have focused on assessing ecological production and then applying economic valuation methods (Kareiva et al., 2011; Nelson et al., 2009). A va riety of reviews and guidelines have discussed these economic valu ation methods in detail (e.g., (Barbier, 2011; Boardman et al., 2006; Hanley et al., 2001; Richar d et al., 2001)). A summary and critique of the use of these methods was presented by Batema n (Bateman et al., 2011) and thus we do not discuss the use of these economic valuation methods in detail here. We provided an example of how different types of data could be collected through various economic valuation methods to assess the net benefits for constructing the ID ES system. The following empirical study will demonstrate the integration of different data sources and valuation methods in detail. Consider a rural household living in a forest ar ea as an example. Costs and benefits from agricultural products and other socioeconomic activities are pa rts of the household™s income and expenditures and could be captured in a survey with relative ease, using best practices for economic surveys. But when benefits or avoide d costs that do not involve market transaction (e.g., non-timber forest products such as fruits, herbal medicine, and fuelwood), they are not shown in the household™s income and expenditure s as conventionally defined and thus are not captured by conventional economic survey methods . If there are established payments for ecosystem services (PES) programs, then the obtained benefits (e.g., payments) and associated 19 costs (e.g., labor costs for monitoring forests) ha ve market values. If such PES programs are not in place, an ESA can be conducted by adding corresponding survey questions, for example by using contingent valuation met hod (see case studies in (Hanley et al., 1998; Yang et al., 2008)). An ESA can also be conducted using integrated tools such as InVEST for the entire study area (see case studies in (Kareiva et al., 2011; Nelson et al., 2009)) and then disaggregating to the household level (e.g., divided by total number of households in the entire study area or calculated by defining a buffer zone of accessibility to certain ecosystem services based on each household™s location, see an example of fuel wood collection in (H e et al., 2009)). 2.3 Empirical demonstration of constructing the index system 2.3.1 Description of the demonstration area Here we provide an example to demonstrat e the index system at Wolong Nature Reserve (N 30û45™ - 31û25™, E 102û52™ - 103û24™, Fig. 2.1) in China. We choose Wolong Nature Reserve as our study area for three reasons. First, situated in the transi tion from Sichuan Basin to the Qinghai-Tibet Plateau, it is within one of the top 25 global biodiversity hotspots endowed with enormous ecosystem services (Liu et al., 2003a ; Myers et al., 2000). Second, it is one of the earliest nature reserves established in China (Liu et al., 2001). Like many ot her protected areas, there are human residents living inside who depend on many types of ecosystem services. Third, our research team has been conducting studies in this area over the past 18 years and has accumulated extensive datasets and local knowle dge that give us a well-grounded basis for testing the IDES concept, methods and applications. 20 The primary purpose of Wolong Nature Re serve is to protect giant pandas (Ailuropoda melanoleuca) as well as regional forest ecosystems a nd rare plant and animal species (Wolong Nature Reserve, 2005). When it was established in 1963, its initial size was ~20,000 ha but was expanded to its current size of ~200,000 ha in 1975 (Wolong Nature Reserve, 2005). It is home to ~10% of the total wild giant panda population (Wolong Nature Reserve, 2005). Currently, there are ~4,900 local human resi dents, distributed in ~1,200 households in two townships (i.e., Wolong and Gengda Townships) within the Reserve (Fig. 2.1). The majority of local residents are farmers involved in subsisten ce activities such as cultivating maize and vegetables, raising livestock (e.g., pigs, cattle, yaks, and horses), collecting traditional Chinese medicine, keeping bees, and collecting fuelwood for cooking and heating (Table S2.1). In response to the massive droughts in 1997 and floods in 1998, the Chinese government started to implement a series of ecosystem service policies (Liu et al., 2008; Liu et al., 2013), including two of the world™s la rgest payments for ecosystem services (PES) programs: the Natural Forest Conservation Program (NFCP) and the Grain-to-Green Program (GTGP) (Liu et al., 2008). These PES programs aim mainly to impr ove regulating services such as soil erosion control, water conservation, carbon sequestration, and air purification. From 2000, Wolong Nature Reserve started to implement GTGP, NF CP, as well as a local PES program called the Grain-to-Bamboo Program (GTBP) (Wolong Nature Reserve, 2005; Yang et al., 2013e). NFCP aims to conserve and restore natural forests through logging bans, afforestation, and monitoring, using PES approach to motivate conservation behavior (Liu et al., 2008; Yang et al., 2013e). GTGP and GTBP aim to convert cropland on steep slopes to forest/grassland, and bamboo forest by providing farmers with subsidies, respectiv ely (Liu et al., 2008; Wo long Nature Reserve, 2005; Yang et al., 2013e). 21 2.3.2 Data and methods The household survey data used here were collected with the permission from the Wolong Administration Bureau of Wolong Nature Reserve. A verbal consent process from interviewees was used due to the low level of e ducation of our interviewees. The verbal consent script was read to the selected interviewees before conducting the survey. Only when they agreed to participate in our survey, we then continued to ask questions in the designed survey instruments. Or else, we did not collect any information but switched to the next selected interviewee. The Institutional Review Board of Michigan State University (http://www.humanresearch.msu.edu/) approved the verbal consent pr ocess, verbal consent script, and survey instruments. For this study we used household survey data to estimate the obtained net benefits from ecosystem services (or equivalen tly gross benefits and costs) for households. This allows us to construct the IDES estimates for households. Ou r surveys were conducted in the summer of 1999 and the end of 2007 to obtain data covering activities in 1998 and 2007. We tracked the same randomly sampled 180 households so the data constitutes a panel. Usually the household heads or their spouses were chosen as interviewees because our past experience indicates that they are the decision makers and are most familiar with household affairs (An et al., 2001). To facilitate cross-context comparisons, we used the categor ies for household income and expenditure data that are consistent with those of the National Bu reau of Statistics of China (National Bureau of Statistics of China, 2011) and thus with sta ndard economic survey methods. We used the MA classification for ecosystem services to generate sub-indices. It is important to note that it is impractical, if not impossible, to assess all the ecosystem services in a study area. This analysis 22 only attempts to include as many major ecosystem services as possible using the best available data in our study area. As the term implies direct ecosystem benefits are those that are used directly in generating human well-being. For example, agricultu ral products are provisioning services that provide direct benefits from ag ricultural ecosystems. Other serv ices contribute indirectly to human well-being. Sometimes indirect benefits ar e only one step removed from direct benefits (i.e., first-order indirect benefits) and sometimes they are more distantly linked (i.e., secondary or more distant indirect benefits). For example, local households do not directly partake in the ecotourism activities but they one-step indirectly benefit from the cultural services of ecotourism through providing transportation, food and acco mmodation services to eco-tourists. But ecotourism may also enhance the development of infrastructure (e.g., road construction), create more job opportunities, and thus provide indirect benefits several steps removed from the cultural services. Generally, the challenge in id entifying benefits for CBA is to separate the genuine indirect effects from those that ar e double accounting (De Rus, 2010). Usually, if there is not a strong rationale, only direct benefits and costs are included to avoid double accounting (Boardman et al., 2006). However, in our study ar ea, first-order indirect benefits capture an important part of benefits from ecosystem se rvices and the inclusion of them do not cause double-accounting (Table S2.1). As a first approximation, here we included direct benefits and first-order indirect benefits in our calculations because these capture d the majority benefits in our study area (Table S2.1). We adapted the MA cl assification for types of related ecosystem services (Table S2.1) to make it appropriate fo r our study area. For some specific services, the classification may differ from what would be appropriate in other areas. But this does not affect 23 the comparisons using the overall index and s ub-indices of IDES, which are based on the generalizable MA framework. Some households obtained nega tive net benefits in agricultural operating income when the total gross agricultural income was lower th an total agricultural expenditure, due to pests, diseases, natural disasters (e.g., storms and landslides ), and/or low prices of agricultural products. Most income from ecosystem services comes fr om provisioning services, but some households also have income from ecotourism, which we categorized as benefits related to cultural services, and income from PES programs, which we categori zed as benefits related to regulating services (Table S2.1). The benefits that households obtained from ecosystems include not only the benefits reflected in their income, but also the avoided costs not reflected in their expenditures. Two major items of avoided costs were assessed here . One is the reduced electricity fees through a subsidized electricity price. B ecause the conservation of forest s also dramatically conserves watersheds in our study area, local households were given a reduc tion of electricity price of 0.07 yuan per kilowatt-hour in both 1998 and 2007 (yuan: Chinese Currency, 1 USD = 7.52 yuan as of 2007). Thus, the avoided electricity fees c ould be calculated by multiplying their consumed electricity amount and the reduced price. Another item of avoided cost is from fuelwood collection for energy use. Households would need to pay for alternative energy sources (e.g., electricity, coal) if they do not collect fuel wood. Because households need to spend labor collecting fuelwood, in the past when one household did not have enough laborers in the fuelwood collection season, one migh t exchange laborers or hire laborers from other households. Thus, the monetary value of collected fuelwood can be estimated as the mark et value of the labor spent on collecting it. In our household survey, we measured the collected amount of fuelwood 24 and total labor spent in collecting it. We th en calculated the shadow price of fuelwood (approximately 0.10 yuan per kilogram in 1998 a nd 0.20 yuan per kilogram in 2007). Data for each household on each of these sources of net in come and avoided costs were then used to construct the index system. 2.3.3 Results of the index system Table 2.1 showed the results of net benefits from different sources and the overall IDES and corresponding sub-indices in both 1998 and 2007. Our results showed a dramatic increase of net benefits from all categories of ecosystem services and socioeconomic activities. From 1998 to 2007, the total net benefit from ecosystem se rvices has increased from an average of approximately 1,723 yuan to 12,972 yuan (both values were in present values for 1998). Meanwhile, from 1998 to 2007, the total net benef it from socioeconomic activities also has dramatically increased from an average of approximately 2,456 yuan to 12,350 yuan. Table 2.1 also showed that the overall i ndex of households™ dependences on ecosystem services has increased from approximately 0.42 in 1998 to 0.61 in 2007. The average overall IDESs were 0.45 in 1998 and 0.61 in 2007, indicating that approximately 45% and 61% of total net benefits to households came from ecosy stem services in 1998 and 2007, respectively. Approximately 54% and 63% households had an ove rall IDES larger than 0.50, and 9% and 16% households had an overall IDES of 1.00 in 1998 and 2007, respectively. Overall these results suggested that most households in our study area were highly dependent on ecosystem services and some were essentially completely dependent on them. 25 The percent of households obtained positive ne t benefits from provisioning services were 89% and 85% in 1998 and 2007, respectively. Almost all households benefited from regulating services in both 1998 and 2007. Perhaps most inte resting, almost all households in 2007 acquired positive net benefits from regulating services through the PES programs (i.e., NFCP, GTGP, and GTBP). These programs were the major reason for the dramatic increase of net benefits from regulating services. However, almost no hous ehold in 1998 and only 11% households obtained positive net benefits from cultural services such as ecotourism. 2.3.4 Advantages and applications of the index system Our IDES is better than the agricultura l income share in reflecting households™ dependences on ecosystem services. Agricultural income share, or the ratio of agricultural income to total income, is a commonly used i ndicator that can approximately reflect a rural household™s dependence on ecosystem services. Although the proposition that the poor are more dependent on ecosystem services is rarely exam ined quantitatively, it is a widely accepted notion (Carpenter et al., 2006; Shackleton et al., 2008). Here, we comp ared the overall IDES with agricultural income share by examining their re lation to overall household income. Our results suggested that the overall IDES were negatively associated with household income in both 1998 and 2007 (Table 2.2).That is, higher income households make less use of ecosystem services. These results confirmed the common view that low incomes households are more dependent on ecosystem services. However, the association of household income with agricultural income share was significant only in 1998 but not in 2007 (Table 2.2). These results indicated that our overall IDES was better than the agricultural income share as a measure of rural households™ 26 dependences on ecosystem services. In our study area, income had become decoupled from income share from agriculture but not from us e of ecosystem services by our second survey. Comparing the results of the index system in 1998 and 2007 (Table 2.1), the reasons that IDES is a better measure than agricultural inco me share are easy to see. In 1998, most of the household income was from agriculture, which was cl assified as benefits related to provisioning services (Table S2.1). The overall IDES was almost equivalent to the sub-index for provisioning services (Table 2.1) and thus was similar to ag ricultural income share in reflecting a household™s dependence on ecosystem services in 1998. In 2007, household income sources became more diverse and included many non-agricultural items (T able S2.1). Therefore, unlike the overall IDES, agricultural income share no longer accu rately reflected households™ dependences on ecosystem services. These results demonstrate some of the uses of the index system. We further illustrated its applications by examining how the benefits of ec osystem services were unequally available to households. Because ecosystem services flow from natural capital, households who possess substantial access to natural capi tal should obtain more benefits from ecosystems and thus would be more dependent on ecosystem services than those with less access to natural capital. The positive associations between indices of dependence on ecosystem services and the area of cropland supported this argument (Table 2.3). Alt hough poverty is often defined in terms of low access to financial capital, households in financ ial poverty often have limited access to human, manufactured, and social capital. Thus, the pr oposition that the poor depend more on ecosystem services may be generalized from poverty in fi nancial capital to povert y in human, manufactured, and social capital. Table 2.3 supported such negative associations between indices of high 27 dependence on ecosystem services and a number of measures of low access to different forms of capital. It should be noted that the relationship be tween dependence on ecosystem services and lack of access to forms of cap ital can change rather rapidl y. In 1998, the local economy at Wolong Nature Reserve was relatively closed and mainly relied on agricu lture. Since 2000, the NFCP, GTGP, GTBP, and tourism development ha ve led to more non-farm income. The local economy became more open. For example, thousands of tourists visit Wolong Nature Reserve every year to view giant pandas. As a result, local farmers began to grow vegetables selling to outside markets, and some young farmers enroll ed their cropland in GTGP and GTBP and migrated to urban areas for work. These were the reasons why some indicators of capital such as the distance to the main road, area of cropland, soci al ties were not as impo rtant in 1998 as they were in 2007 (Table 2.3). Furthermore, most households in 1998 lived in low quality wooden or stone sheds, while in 2007, with more money av ailable, some of them lived in high quality concrete houses. Households actively strategize how to substitute one form of capital for another in order to achieve access to needed resources and enhance well-being (Chen et al., 2012a). 2.4 Discussion and conclusions This proposed index system is a step forward in quantifying human dependence on ecosystem services. As mentioned above, it is imp ractical, if not impossible, to capture all ecosystem services in any study. However, we ha ve implemented the inde x system with existing household survey data and were able to capture ma jor ecosystem services in our initial estimates. Our results suggest that such an approximation is a viable approach that reveals useful 28 information on human dependence on ecosystem se rvices. The overall inde x and its sub-indices can reveal the general pattern of households™ dependences on ecosystem services, and the variations across time, space, and different le vels of access to multiple forms of capital. Given the fact that different individuals, households and communities rely on different ecosystem services, in practice it is likely that the measurement of benef its, costs and IDES will depend on relatively detailed data that is collected with an understa nding of the local context. To facilitate comparisons across time, space and different institutional levels, it is necessary that different studies use a common platform such as the generally accepted MA framework for classification. While it will usually not be possible to capture all benefits and costs associated with ecosystems services, care should be taken to accurately estimate the most important benefits and costs in the local context in order to construct overall IDES estimates. To avoid misinterpretation in some circumstances (e.g., the effect of Integrated Conservation and Development Projects on changes in househol ds™ dependence on ecosystem services), such comparisons should not only focus on the overall ID ES index but also consider the sub-indices as well as the structure of dependence on ecosystem services (i.e., the distribution of three sub-indices). It should be noted that IDES measures th e relative importance of ecosystem services, with comparison to other socioeconomic activities, in providing benefits directly and indirectly to humans. If one wants to compare the absolute values of benefits from ecosystem services across different areas, one should use the net benefits obtained from ecosystem services, which are also provided through construction of the IDES system. 29 It should also be noted that there is a s ubstantial difference between being dependent on ecosystem services and being dependent on PES. The PES program compensates some of the forgone benefits that local households enjoyed before the implementation of conservation policies. But the PES program does not necessarily compensate all the forgone benefits. For example, in our study area, the main purpose of paym ents from NFCP is to protect natural forests. As a result payments from NFCP mostly co mpensate the forgone pr ovisioning (e.g., timber harvest) services whereby regulating services (e.g., soil erosion control, carbon sequestration, and water conservation) are increased. Cultural services (e.g., recreation and ecotourism) are not included. In our study, we therefor e included benefits from ecotour ism which are not captured in the PES from the NFCP. Using the overall IDES, we confirmed the proposition that the poor are more dependent on ecosystem services, and thus are more vulnerable to degradation or decline of the corresponding ecosystem services. More importan tly, we generalized this proposition to those disadvantaged groups who possess less access to mu ltiple forms of capital (i.e., human, financial, manufactured, and social capit al) and found they too were more dependent on ecosystem services than the affluent. Although we demonstrated the construction of the IDES system and its applications based only on data from the Wolong Nature Re serve, the conceptual basis of IDES and methodology we used were designed to be gene ralizable. While we examined households, the unit of analysis could range from individuals to communities, regions, and nations. Our analysis here is a proof of the concept of IDES. Further elaborations are warranted and could potentially improve the estimates of human dependence on ecosystem services. 30 We believe the IDES index system presented here has some major advantages to advance the understandings of linkages between ecosystem services and human well-being and support decision-making. First, this paper empirically de monstrated how the index system could better reflect human dependence on ecosystem services than the other commonly used indicator (i.e., agricultural income share) at the household level. Second, the index system provides both a composite index and sub-indices. This allows the quantitative analysis of the structure of human dependence on ecosystem services and the quantitative examination of the interwoven linkages between different types of ecosystem services and different components of human well-being. Future studies could combine IDES with indicators of indirect drivers, direct drivers, and human well-being to construct integrated models based on the MA framework to better understand complex interactions among human and natu ral components (e.g., to assess how human dependence on ecosystem services may affect human well-being). The improved understanding may help to develop theories on the complex ity of CHANS and inform decision making in a rapidly changing global environment. Third, th e improved understanding of linkages between ecosystem services and human well-being, if in tegrated into decision-making, may avoid some inappropriate strategies that aggravate the margin alization of disadvantaged groups. This in turn could reduce the pressure these groups place on eco systems to obtain services critical to them. However, a distinction should be made: dependen ce on ecosystem services is not equivalent to pressure on ecosystems because there are often su stainable ways to extract ecosystem services. Acquisitions of regulating and cultural services (e.g., air purification and ecotourism) are often non-consumptive, while many uses of provisioni ng services (e.g., timber, fuelwood) are often consumptive and may or may not be sustainabl e. The reduction of pressure or impacts on ecosystem services can be realized through reduction of overall dependence on ecosystem 31 services or through a shift of the structure of dependence to different types of ecosystem services such as a shift from high dependence on provisi oning services to high dependence on regulating and cultural services. It can also be achieved by extracting provisioning services in ways that do not harm the ecosystem. Fourthly, improved unders tanding may also enhance the effectiveness and long-term viability of conservation and development programs by targeting priority population groups such as those with limite d access to capital and high dependence on provisioning services. Finally, such understanding could draw stakeholders™ attention to the unmanaged risks and unrealized opportunities associated with ecosystem service changes. Climate change and other global changes are causing rapid shifts in ecosystem structure and function and may threaten continued flow of services to those most dependent upon them. Taking our study area as an example, conserva tion and development efforts such as NFCP, GTGP, GTBP, and tourism development have already reduced many households™ dependences on provisioning services; however, the very uneven distribution of benefits from ecosystem services may create potential risks and impede the future success of such policies. 32 Table 2.1. Net benefits, overall IDES and sub-indices in 1998 and 2007. Net benefits/Indices 1998 2007 Mean (S.D.) Range (Minimum: Maximum) * Mean (S.D.) Range (Minimum : Maximum) Net socioeconomic benefit (yuan) 2456.38 (3315.50) (0 : 16,600) 12,350.10 (21,027.75) (0 : 186,046) Net benefit from provisioning services (yuan) 2308.77 (2506.89) (Œ 2671 : 13,676) 8544.97 (14,063.43) (Œ 2620 : 107,003) Net benefit from regulating services (yuan) 77.60 (92.88) (0 : 544) 2900.64 (2003.59) (0 : 10,448) Net benefit from cultural services (yuan) 3.33 (44.72) (0 : 600) 1526.10 (13,476.34) (0 : 177,626) Total net benefit from ecosystem services (yuan) 2389.71 (2527.22) (Œ 2620 : 14,182) 12971.70 (19,043.19) (Œ 27 : 181,801) Sub-index for provisioning services 0.4131 (0.8627) (Œ 7.0351 : 0.9973) 0.3754 (0.3131) (Œ 0.1750 : 1) Sub-index for regulating services 0.0340 (0.0714) (0 : 0.7518) 0.2112 (0.2026) (0 : 1) Sub-index for cultural services 0.0003 (0.0038) (0 : 0.0513) 0.0257 (0.1143) (0 : 0.8568) Overall IDES 0.4473 (0.8237) (Œ 6.9019 : 1) 0.6123 (0.3055) (Œ 0.0015 : 1) Notes: Monetary values for net benefits in 2007 were discounted into present values of 1998 for comparison. * Negative value of an index means that th e gross benefit from ecosystem services is lower than the sum of costs from ecosyst em dis-services and co sts of generating the corresponding ecosystem services. 33 Table 2.2. Comparison of overall IDES and agricultura l income share for thei r associations with gross household income. Household income in 1998 Household income in 2007 Agricultural income share Œ 0.355*** Œ 0.012 Overall IDES Œ 0.194** Œ 0.405 *** Notes: Numbers are Spearman™s rhos. Tota l samples are the same 180 randomly sampled households across years. *p < 0.05; **p < 0.01; *** p < 0.001. 34 Table 2.3. Regression of sources of variation on overall IDES. Variable IDES 1998 IDES 2007 Natural capital Area of cropland (Mu, 1 Mu = 1/15 ha) 0.020 (0.014) 0.042*** (0.007) Human capital Household size Œ 0.077 (0.049) Œ 0.037* (0.016) Number of laborers Œ 0.070 (0.056) Œ 0.080*** (0.018) Average education of adults (year) Œ 0.032ƒ (0.017) Œ 0.032 *** (0.009) Average age of adults (year) 0.011* (0.005) 0.008 *** (0.002) Financial capital Household income (yuan, log) 0.071 (0.075) Œ 0.152*** (0.025) Per capital income (yuan, log) 0.143 (0.095) Œ 0.126*** (0.028) Manufactured capital Type of house (0 for low quality non-concrete sheds and 1 for high quality concrete house) Œ 0.022 (0.139) Œ 0.189*** (0.048) Distance to the main road (meter, log) Œ 0.029 (0.027) 0.042*** (0.010) Social capital Social ties to local township and reserve level officials (0: low; 1: high). 0.065 (0.129) Œ 0.188** (0.065) Spatial heterogeneity Township (0: Gengda; 1: Wolong) 0.098 (0.133) Œ 0.101* (0.047) Notes: Numbers outside and inside parentheses are coefficients and robust standard errors of bivariate regressions, respectively. Dependent variables are overall IDES in 1998 and 2007, respectively. Total samples ar e the same 180 randomly sampled households across years. ƒp< 0.01; *p < 0.05; **p < 0.01; *** p < 0.001. 35 Figure 2.1. Wolong Nature Reserve in Sichuan Provi nce, southwestern China. For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this dissertation. 36 CHAPTER 3 GOING BEYOND THE MILLENNIUM ECOSYSTEM ASSESSMENT: AN INDEX SYSTEM OF HUMAN WELL-BEING In collaboration with Thomas Dietz, Daniel Boyd Kramer, Xiaodong Chen, Jianguo Liu 37 Abstract Understanding the linkages between ecosyst em services (ES) and human well-being (HWB) is crucial to sustain the flow of ES for HWB. The Millennium Ecosystem Assessment (MA) provided a state-of-the-art synthesis of such knowledge. However, due to the complexity of the linkages between ES and HWB, there are still many knowledge gaps, and in particular a lack of quantitative indicators and integrated models based on the MA framework. To fill some of these research needs, we developed a quantitative index system to measure HWB, and assessed the impacts of an external driverŠ the 2008 We nchuan EarthquakeŠ on HWB. Our results suggest that our proposed index system of HWB is we ll-designed, valid and could be useful for better understanding the linkages between ES and HWB. The earthquake significantly affected households™ well-being in our demonstrat ion sites. Such impacts differed across space and across the five dimensions of the sub-index (i.e., the basic material for good life, security, health, good social relations, and freedom of ch oice and action). Since the conceptual framework is based on the generalizable MA framework, our methods should also be applicable to other study areas. 38 3.1 Introduction Understanding the linkages between ecosyst em services (ES) and human well-being (HWB) is crucial to sustain th e flow of ES for HWB (MA, 2005). The Millennium Ecosystem Assessment (MA) was intended to provide a state- of-the-art synthesis of such knowledge. ES are defined as the benefits human directly and i ndirectly obtained from ecosystems (Daily, 1997; MA, 2005). The MA suggested that ecosystems pr ovide services that are of importance for improvements of HWB at multiple scales. These services range from provisioning services such as clean water, food, and forest products, through regulating services such as flood control, soil retention, and air purification, to cultural services such as ecotourism, aesthetic appreciation, and a sense of place (MA, 2005). However, during th e past five decades, such improvements of HWB were achieved at escalating co sts due to the decline or degradation of more than 60% of ES across the globe. This decline or degradation in ES may increase th e risks of nonlinear or abrupt changes, and may lead to further margin alization of some groups of people (MA, 2005). Although the MA is a monumental work, the linkages between ES and HWB are complex and remain poorly understood (Butler a nd Oluoch-Kosura, 2006; Carpenter et al., 2006; Carpenter et al., 2009). There are four major challenges in de veloping better understanding of such linkages. First, HWB itself is an evolving and complex concept (Butler and Oluoch-Kosura, 2006). It is difficult to provide a universally acceptable definition of HWB. In the MA, HWB has five constituents: the basic material for a good li fe, security, health, good social relations, and freedom of choice and action (MA, 2005). Seco nd, ES substantially, but not exclusively, contribute to HWB. We interpret MA™s definiti on of HWB as the satisfaction of human needs (Maslow, 1943) to achieve a state of being well (i.e., healthy, happy, and prosperous), both physically and mentally. While ES substantia lly satisfy many human needs (MA, 2005), there 39 are many other influences on well-being, such as personal factors (e.g., personality, self-expectations), demographic factors (e.g., age, and gender), institutional factors (e.g., legal frameworks in which one lives), life experience (e.g., traumatic or disruptive events), and other contextual factors that may affect the subjective feelings of humans (Diener, 2009; Diener et al., 1999). These factors may be affected by ES indirectly rather than directly. For example, threat of violent conflicts may lead to a lack of a sense of security and armed conflicts may be the result of degradation of food supply or other natural resources. Third, the concept of ES is also an evolving concept that changes as we develop new understandings of nature. For example, human society began to appreciate the carbon sequestration capacity of ecosystems only after recognizing that the increasing carbon emission si nce the Industrial Revolution is leading to problematic global warming (Ros a et al., 2010). Meanwhile, it has been widely recognized that the fiwin-winfl solutions are rare and often ther e are trade-offs among different ES that each contributes to HWB (McShane and Wells, 2004; Tallis et al., 2008). Finally, the linkages between ES and HWB are bidirectional and dyna mic across space and time (Carpenter et al., 2009). Even the most simplified version of MA conceptual framework has demonstrated feedback loops among the four components (i.e., i ndirect drivers, direct drivers, ES, and HWB) (MA, 2005). So far there have been relatively few studies quantitatively integrating the four components of the MA conceptual framework to study the linkages between ES and HWB (MA, 2005; Nelson et al., 2009; Pereira et al., 2005). Existing quantitative indicators and models of ES were designed under other conceptual framewor ks for particular sectors (e.g., land use and land cover change, water supply) or to address the intersections between sectors (e.g., biodiversity and land use and land cover change) (Carpenter et al., 2009). Before the MA, 40 measures of HWB mostly focused on the economic, social-psychological, and health dimensions and did not acknowledge ES as driving forces of HWB (MA, 2005; Summers et al., 2012). These indices include the World Health Organization™s Quality of Life measure (WHOQOL), the Genuine Progress Index (GPI), the Happy Planet Index (HPI), the Human Development Index (HDI), the Life Satisfaction Index, and other differe nt indices of Quality of Life (QoL) (Camfield and Skevington, 2008; Diener, 1994, 2000; Diener and Ryan, 2009; Kahneman and Krueger, 2006). Since the MA, it is becoming widely accepte d that HWB cannot be separately considered from ES (Abdallah et al., 2008; Summers et al., 2012; Vemuri and Costanza, 2006). Furthermore, many quantitative studies of HWB do not cove r all five components of HWB in the MA framework, and thus are inappropriate for the integration of the ES and HWB components. Although qualitative measures of ES and HWB are useful for some studies at the local level (Pereira et al., 2005), they are inadequate to overcome the major cha llenges discussed above. Rather we require indicators suitable for quantitative analyses (e.g., system modeling and simulation, and detailed statisti cal analysis of causes and effects). Therefore, developing quantitative indicators and models matching the MA framework is a top priority if we are to understand the linkages between ES and HWB (Car penter et al., 2006; Carpenter et al., 2009). In recent years, a substantial amount of effo rt has been made to quantify various ES at multiple scales and assess the trade-offs and synergies that occur in both natural ecosystems and constructed/artificial ecosystems (Chang et al., 2011; Kareiva et al., 2011; Nelson et al., 2009; Power, 2010; Raudsepp-Hearne et al., 2010; Swallow et al., 2009; TEEB, 2010; Yang et al., 2008). The Natural Capital Project (Kareiva et al., 2011; Nelson et al., 2009) and the Economics of Ecosystems and Biodiversity (TEEB) project (TEEB, 2010) are examples of such efforts that have substantially advanced our understanding of these issues. A few recent studies (Abdallah et 41 al., 2008; Dietz et al., 2009, 2012; Jordan et al., 2010; Summers et al., 2012; Vemuri and Costanza, 2006) have discussed in detail how ES contribute to HWB and provided some new insights to improve the understanding of the linkages between ES and HWB. For example, Vemuri and Costanza (2006) and A bdallah et al. (2008) examined how different forms of capital, including natural capital, might e xplain the life satisfaction at the country level (Abdallah et al., 2008; Vemuri and Costanza, 2006). Jordan et al . (2010) provided a conceptual framework to construct a composite index of HWB, including basic human needs, environmental needs, economic measures and happiness (Jordan et al ., 2010). Dietz et al. (2009, 2012) proposed a model of efficient well-being to assess national efficiency in enhancing HWB through the use of different forms of capital (Die tz et al., 2009, 2012). Summers et al. (2012) comprehensively reviewed the components of HWB with an emphasis on the contri bution of ES (Summers et al., 2012). However, relatively less attention has been paid to developing quantitative indicators and models of HWB based on the MA framework, nor has there been much work on empirically integrating HWB indicators with indicators and/or models of ES. For instance, quantitatively we know little of how changes in ES, human use of ES, and/or dependence on ES may affect HWB, nor how different population groups have been aff ected by changes in ES and how have they in turn responded (Carpenter et al., 2006; Carpenter et al., 2009; Yang et al., 2013c). While theories of how human activities drive environmental cha nges are progressing stead ily, the understanding of how environmental changes may affect hu mans lag far behind (Dietz et al., 2009). In response to some of thes e research needs, in this study, we attempt to (1) develop a new index system to quantify HWB based on the MA framework and (2) empirically demonstrate the application of the index th rough assessing the impacts of the 2008 Wenchuan Earthquake on HWB. 42 3.2 Methods 3.2.1 Ethics statement We obtained the permission from the Wol ong Administration Bureau of Wolong Nature Reserve for conducting household surveys inside the reserve and outside the reserve at Sanjiang Township through the Sanjiang Conservation Station. Since many of our interviewees are not literate or have very low level of education, a verbal consent process was used. We first read the verbal consent script to the selected intervie wees. Once they agreed, we then continue to interview them. If consent was not obtained, we di d not collect any further information from that interviewee and switched to the next selected interviewee. The survey instruments, verbal consent process, and script were approved by the Institutional Review Board of Michigan State University (http://www.humanresearch.msu.edu/). 3.2.2 Development of the human we ll-being index (HWBI) system We developed a new instrument to measure HWB based on the MA conceptual framework. To do this, we first reviewed previous literature and selected a list of indicators for each of the five dimensions. Second, we refined th e measures from the literature to situate them in the MA framework, in some cases adding new measures. Third, we pre-tested the indicators with respondents from outside of our research samples and revised them. Finally, we examined the internal validity of the items using item-total co rrelations to check if any item is inconsistent with the average response across all items. The final list of indicators, the specific asked questions, and the results of internal validity checks are shown in Table S3.1. Throughout the 43 instrument development process, we followed st andard guidelines for using multiple indicators to develop measures of composite variable s (Brown, 2006; Rowe, 2006). Specifically, we pretested the wording to ensure each indicator measured a sing le, observable outcome. We used positive nomenclature for all the wording of indicat ors because the technical literature has shown that ratings on negatively worded items or indicators are signifi cantly less reliable than those positively worded (Rowe, 2006; Rowe and Rowe, 1997; Sandoval, 1981). The response to each item was measured with a five category Likert-style scale. After preliminary reliability analysis, we us ed Confirmatory Factor Analysis (CFA) to construct the overall index and sub-indices. CFA addresses several problems in this type of data analysis. First, it allows us to avoid the pitfall of assuming indivi dual items are identical with the underlying theoretical variables (Eagly and Kulesa, 1997). CFA allows the boundaries distinguishing the five MA dimensi ons to be fluid and is open to the possibility that some items may tap multiple underlying variables. For exampl e, a higher satisfaction with housing condition may not only reflect a higher satisf action with the adequacy of mate rial goods but also a stronger feeling of safety. Second, CFA is a special form of structural equati on modeling that handles both the measurement model, that is, the relatio nship between indicators (or observed measures) and factors (or latent variables), and the casual model linking latent variables to each other and to observed variables (Brown, 2006). Unlike traditional methods (e.g., principal component analysis), CFA handles easily both the situations of multiple indicators for one factor and one indicator for multiple factors. Third, results of CFA can provide compelling evidence for construct validity (Brown, 2006). Finally, CFA, un like traditional methods, allows for hypothesis tests regarding unequal contributions of indicators to the measured factors, minimizes the 44 problem of non-normal and non-continuous distri butions of indicators, and adjusts for measurement errors (Brown, 2006; Rowe, 2006). 3.2.3 Description of the demonstration sites There are four reasons we chose the Wolong Nature Reserve (WNR) and the adjacent Sanjiang Township (SJT) in Wenchuan County of Sichuan Province, southwestern China (Fig. 3.1) as our sites to demonstrate th e utility of our approach. First, they are sites of great ecological importance. Both WNR and SJT belong to the Si chuan Giant Panda Sanctuaries of the United Nations Educational, Scientific and Cultural Organization (UNESCO) World Heritage system. The Sanctuary was established in 2006 to promote the conservation of the giant panda habitat (Li et al., 2013). It also has been classified as one of the world™s top 25 Biodiversity Hotspots (Myers et al., 2000) and one of the Global 200 Eco-regions defined by the World Wildlife Fund (WWF, 2007). Second, human resi dents settled the area hundreds of years before the establishment of the sanctuaries and developed wa ys of life adapted to the local environment. Local residents™ well-being substantially depends on many ES. Because human and natural systems are coupled as a result of the current situ ation as well as a long history (Liu et al., 2007a; Liu et al., 2001), successful conservation of the giant panda habitat and associated ecosystems and services will not be achieved if local resident s™ well-being is ignored. Third, the destructive Wenchuan Earthquake provides a dramatic, if tragic , natural experiment to examine the impacts on HWB. Finally, during the past 18 years, our research team ha s been working in WNR and has collected extensive data both before and after the earthquake. These formal data are matched with accumulated local knowledge of this area that helped us to design our surveys and interpret our results. 45 WNR is approximately 2,000 km 2 with approximately 4,900 local rural residents from about 1200 households. SJT is 491 km2 of which 344 km2 is enclosed in WNR, and all local residents now live in the remaining 147 km 2 zone outside WNR (Fig. 3.1). SJT has approximately 4,000 local rural residents distri buted across 1,100 households. The majority of households at WNR and SJT earn their livelihood main ly through agricultural activities (e.g., growing maize and vegetables, ra ising livestock, and collecting ma terials for traditional Chinese Medicine), and partly through temporary local jobs (e.g., road construction), small tourism businesses (e.g., selling souvenirs), and migrant work in the cities outside the local area (Yang et al., 2013e). The epicenter of the devast ating 2008 Wenchuan Earthquake (M s 8.0 or Mw 7.9) was close to our demonstration sites (Fig. 3.1). Th e earthquake caused tremendous socioeconomic and ecological impacts. By September 25, 2008 it was reported that 69,227 people died, 374,643 were injured, 17,923 were missing, and a to tal of 1,486,407 victims were evacuated and temporally resettled (Xinhua News Agency, 2008a). Rough estimates of the direct and indirect economic losses (e.g., damages to infrastructure, croplands, and tourism) were over one trillion yuan (Xinhua News Agency, 2008b). The earthquake also has caused huge impacts to ecosystems and wildlife habitat. Approximately 122,136 ha (3.40% of the total area of natural ecosystems) were affected including 97,748 ha of forest, 18,021 ha of shrub, 4,919 ha of meadow, 1,157 ha of barren land, 242 ha of water bodies, and 50 ha of snow-covered land (Ouyang et al., 2008). Approximately 65,584 ha of panda habitat (5.92% of the total panda habitat) were damaged with 34,737 ha and 30,847 ha distributed inside and outside nature reserves respectively (Ouyang et al., 2008). A lthough both WNR and SJT were affected by the 46 earthquake, the impacts at SJT were less severe th an those at WNR. Forty-eight local residents of WNR and seven of SJT died in the earthquake. Several additional hundreds of workers and passengers died along the road within the reserve. Infrastructure such as the roads, residential houses, schools, hospitals, and tourism facilities were destroyed at WNR but were less damaged at SJT. In fact, after the earthquake, since the main road of WN R was blocked while the road of SJT was accessible, many people inside WNR fled using trails to SJT. The variation in earthquake impacts between WNR and SJT allows us to examine the differential effects of the 2008 Wenchuan Earthquake on HWB. 3.2.4 Household surveys During the summer of 2010 we randoml y sampled approximately 15% of local households both inside and outside the reserve for a total of 169 households at WNR and 157 households at SJT. Because our past experience in this area suggests that household heads or their spouses are usually the decision makers ab out household affairs and thus most familiar with the questions we were asking (An et al., 2001), we chose them as interviewees. For the collection of retrospective data, we followed standard practices of life history calendars to enhance respondents™ recall accuracy (Axinn et al., 1999; Freeman et al ., 1988). Before conducting the formal household interviews, we first explained th e meaning of each indicat or to our local field assistants, and pretested and revised the survey instrument with households outside of our sample area. Because most of the interviewees were farmers with low literacy, we implemented the interviews face-to-face using local languages. 47 3.3 Results 3.3.1 Internal consistency of the HWBI system The combination of indicators we used app ears to be an appropriate measure of HWB. The item-total correlations for each indicator are reasonably strong, ranging from 0.30 to 0.74 before and 0.33 to 0.75 after the earthquake (Table S3.1). The Cronbach™s alpha values are highŠ0.92 and 0.91 for before and after the earthquake, respectively. Moreover, the deletion of any of the indicators reduces the value of alpha for both before and after the earthquake. Thus these items appear to have r easonable internal validity. 3.3.2 CFA results of the HWBI system Model fit statistics show that the goodness-of-fit of our CFA is high regardless of the criterion used (Table 3.1). The ratio of Chi-Square to the degrees of freedom (df) is 1.6, which is lower than the commonly used maximum of 3 as a criterion for adequate fit (Gefen et al., 2000). The Comparative Fit Index (CFI) and Tu cker-Lewis Index (TLI) are 0.976 and 0.971 respectively, again indicating good fit. Both the Root Mean Square Error of Approximation (RMSEA) and Standardized Root Mean Square Residual (SRMR) are lower than 0.05. No modification indices are above the default threshold of 10 suggested by Muthen and Muthen (Muthén and Muthén, 1998-2010). In addition to the overall fit statistics, the significance tests of coefficients for each path and the test of signi ficance for each path™s contribution to model fit also show high goodness-of-fit and construct validity for each indica tor (p < 0.05, Table S3.2). Our results also suggest that all five latent variables representing the five dimensions of HWB have significant coefficients and significantly contribute to the model fit (p < 0.001, Fig. 48 3.2 and Table S3.2). These results are consistent with the MA structur e of five different dimensions of HWB. However, our results (Fig. 3.2, Table S3.2) also suggest that the dimension of basic material for good life is significantly positively a ssociated with the dimension of freedom of choice and action (p < 0.001), and the dimension of security is signi ficantly positively associated with the dimension of good social relations (p < 0.001). This evidence suggests that the five dimensions are not fully independe nt (as we suspected would be the case while conceptualizing the five dimensions), and thus it is appropriate to use CFA instead of the principal component analysis that usually assumes ort hogonality of latent variables. 3.3.3 Impacts of the earthquake on HWBI Table 3.2 provides descriptive st atistics of sub-indices and overall HWBI both inside and outside the reserve before and after the earthquake. Our results show that overall HWBI and sub-indices inside the reserve were all significantly higher than those outside the reserve both before and after the earthquake (Table 3.2). However, the impacts of the ea rthquake on overall HWBI and sub-indices differed from inside to outside the reserve, as indicated by the interaction term in the regressions estimating the effects of the ear thquake (Table 3.3). All the coefficients of interaction terms between the pre-earthquake value and the research site were negative, and all but the term for good social relati ons were statistically significant (Table 3.3). It appears that the decreases in the sub-indices and overall HWBI insi de the reserve were larger than those outside the reserve. Of the coefficients of the pre-eart hquake index values, only the sub-index of freedom of choice and action and overall HWBI were significant (p < 0.05) and both are positive. This suggests that households with higher freedom of choice and action or overall HWBI pre-49 earthquake decreased less in freedom of choi ce and action and in overall HWBIŠhigh values seemed to buffer against the a dverse impacts of the earthquake. 3.4 Discussion We proposed a HWBI system based on the MA framework and empirically demonstrated its construct validity. Further, the difference in the effects of the earthquake on HWB indices between households outside and inside the reserve is evidence that the observed impacts on HWB are consistent with what we would expect as a result of the earthquake and short-term post-disaster situation. So we believe we have a strong case for both the internal and external validity of our proposed measure. Compared to outside the reserv e, the larger decreases in th e overall index and sub-indices of HWB inside the reserve are probably because pre-earthquake the overall index and sub- indices were higher inside the reserve and because there were more severe damages, especially destruction to the main road connecting the reserve with the outside world. The significant decreases in sub-indices for security and health indicate that the earthquake caused not only physical damages affecting local households™ livelihoods but also had negative impacts on their mental health. Nevertheless, the short-term post-di saster reconstruction effo rts seemed to turn the disaster into opportunities to improve local households™ welfare. Outside the reserve, the sub-indices for basic material for good life and freed om of choice and action actually increased significantly (p < 0.001 and p < 0.05, respectively) after the earthquake. One major reason is undoubtedly post-disaster road construction. Due to the implementation of the Wenchuan Earthquake Reconstruction Plan (State Planning Group of Post-Wenchuan Earthquake Restoration and Reconstruction, 2008), road cond itions outside the reserve have dramatically 50 improved and all nine villages at SJT now have cement pavement roads connecting to the main road. In contrast to the construction efforts in the STJ, those inside the reserve suffer from frequent associated disasters after the earthquake (e.g., mud-rock flows and landslides) that are less problematic for those outside. Our results also suggest that households w ith higher overall HWBI or less freedom of choice and action pre-earthquake were less affected than those with lower indices. These results are consistent with findings from other studies such as the Hurricane Katrina (Elliott and Pais, 2006). It is probably because those households with lower over all HWBI or less freedom of choice and action lack adaptive capacity and t hus were more vulnera ble to the disaster (Gunderson, 2010). This pattern holds both inside and outside the reserve. It indicates that it is not caused by different socioeconomic contexts inside and outside the reserve. Unfortunately the post-disaster reconstruction policy had not addressed this problem by the time of our investigation in the summer of 2010. This suggests that post-disaster policies should give priority to those households with lower overall HWBI or less freedom of choice and action. The post-disaster reconstruc tion outside the reserve was almost completed by May of 2010 but is still ongoing inside the reserve with completion planne d for 2015. At this time it is difficult to predict how the earthquake-induced changes in HWB ma y in turn affect households™ socioeconomic activities and use of ES in the lo ng run. But according to information gained from regular monitoring by the local government and our own field investigation, starting shortly after the earthquake there seemed to be dramatic incr eases of illegal logging and poaching outside the reserve and increases in poaching inside the reserve. This indicates that post-disaster reconstruction must consider hous eholds™ use of and dependence on ES and their interactions 51 with HWB. Priorities should be given to helping local households to build capacity and find alternative income sources that do not ha rm or offset conservation efforts. We believe the HWB index systems we developed has some major advantages compared to other approaches that have been proposed. First, our index system is based on the general MA framework and explicitly consider s the contribution of ES to HWB. Second, its construct validity has been confirmed empirically in our demonstrati on sites. Therefore, it c ould easily be applied to other study areas and across different scales with some modifi cations if necessary. Third, our index system is developed using CFA techniques that examine the relationships of multiple indicators to multiple factors (i.e., dimensions) a nd the correlations among different dimensions. This allows a nuanced assessment of the measurem ent properties of the index and its components. Finally, our index system provides both a composite index and s ub-indices. This allows the quantitative examination of the interwoven linkages between different types of ES and different components of HWB. The value of having bot h aggregate and disaggregate measures was evident when we considered the effects of the earthquake, which differed across sub-indices. However, similar to other composite indices but unlike single question indices, we acknowledge that this advantage is achieved at the costs of more data being needed and a more effort in constructing the measure. 3.5 Conclusions and implications We developed an index of HWB based on th e MA framework and applied it to a region in which ES are very important for HWB. Our results suggest that our HWBI system has reasonable internal and external validity. Our index was able to detect the Wenchuan earthquake™s impact on household well-being, and show that the estimated impacts differed 52 between households inside and outside the reserve as well as acro ss the five dimensions of the sub-index. Human and natural systems are complex and coupled (Liu et al., 2007a). Human use of and dependence on ES affects HWB, and changes in HWB may in turn affect human use of and dependence on ES. Our analysis points to some pract ical implications of this coupling. If post- disaster reconstruction policies do not adequately address the negative impacts of the earthquake on local households, especially those with less freedom of choice and action and lower overall HWBI, many households may be forced to find alternative income sour ces including illegal logging and poaching to main tain basic livelihoods. Our proposed HWBI system seems to be a viable approach and could be useful for further research to better understand the linkages between ES and HWB. We demonstrated the development of the HWBI system and its app lication in assessing the impacts of 2008 Wenchuan Earthquake on households™ well-being. Since the conceptual framework is based on the generalizable MA framework, our methods should also be applicable to other study areas. 53 Table 3.1. Summary of model fit information for the confirmatory factor analysis. Fit statistics Value Ratio of Chi-Square to df (2/df) 1.6 CFI (Comparative Fit Index) 0.976 TLI (Tucker-Lewis Index) 0.971 RMSEA (Root Mean Square Error of Approximation) 0.030 SRMR (Standardized Root Mean Square Residual) 0.035 Notes: *** p < 0.001. The Chi-square value is for the MLR estimator (maximum likelihood estimation with robust standard errors) in Mplus, which is not used for Chi-square difference testing in the regular way. No modification indices are above th e default threshold of 10 in Mplus. All observed variables and latent variables are tested to significantly (p < 0.05) contribute to model fit. 54 Table 3.2. Descriptive statistics of sub-indices and overall HWBI both inside and outside the reserve before and after the earthquake. Human well- being Before earthquake After earthquake Inside Outside t value Inside Outside t value Basic material for good life (Q1) 0.461 (0.018) 0.239 (0.014) 9.870*** 0.439 (0.018) 0.267 (0.017) 7.039*** Security (Q2) 0.692 (0.010) 0.463 (0.015) 12.459*** 0.603 (0.013) 0.445 (0.018) 7.002*** Health (Q3) 0.668 (0.010) 0.467 (0.013) 12.234*** 0.589 (0.012) 0.403 (0.016) 9.094*** Good social relations (Q4) 0.685 (0.010) 0.465 (0.014) 12.623*** 0.642 (0.012) 0.474 (0.017) 8.111** Freedom of choice and action (Q5) 0.387 (0.018) 0.140 (0.013) 11.006*** 0.364 (0.018) 0.151 (0.016) 8.946*** Overall HWBI 0.640 (0.009) 0.422 (0.013) 13.746*** 0.566 (0.012) 0.375 (0.016) 9.646*** Notes: Numbers outside and inside parentheses are means and st andard errors for changes of overall indices and sub-indices, respectively. The numbers of observations are 169 and 157 inside and outside the reserve both before and after the earthquake, re spectively. The overall index and sub-indices are respectively normali zed into the range from 0 to 1 using the maximum-minimum normalization method. 55 Table 3.3. Impacts of the earthquake on sub-indices and overall HWBI inside and outside the reserve. Independent variables Q1 Q2 Q3 Q4 Q5 Overall HWBI Pre-earthquake index 1.082 (0.055) 1.029 (0.043) 1.031 (0.048) 1.048 (0.043) 1.111* (0.049) 1.104* (0.067) Site (0: outside; 1: inside reserve) Œ0.002 (0.014) 0.032 (0.044) 0.080ƒ (0.044) Œ0.012 (0.036) 0.019 (0.012) 0.067ƒ (0.036) Pre-earthquake index × Site Œ0.145* (0.060) Œ0.159* (0.071) Œ0.151* (0.075) Œ0.074 (0.060) Œ0.207** (0.060) Œ0.182** (0.064) Constant 0.009 Œ0.031 Œ0.079** Œ0.013 Œ0.004 Œ0.091*** Notes: Dependent variables are corresponding sub-indices or overall HWBI post-disaster respectively. Numbers outside a nd inside parentheses are coeffi cients and standard errors, respectively. The number of total observation is 326, including 169 and 157 observations inside and outside the reserve, respectively. ƒp < 0.1; *p < 0.05; **p < 0.01; *** p < 0.001. 56 Figure 3.1. Wolong Nature Reserve and adjacent Sanjiang Township in Wenchuan County, Sichuan Province, southwestern China. 57 Figure 3.2. Path diagram of the confirmatory factor an alysis model for HWBI. Notes: HWBI: Human well-being index; Q1: Basic material for good life; Q2: Security; Q3: Health; Q4: Good social relations; Q5: Freedom of choice an d action. Single-headed arrows indicate the direction of causal influence, and double-headed arrows represent covariance between two latent variables. Number on each path represents the standardized coefficient estimated by the confirmatory factor analysis model. Paths of th e structural model are not shown here. Detailed description of observed indicators and model results are sh own in Table S3.1 and S3.2. 58 CHAPTER 4 AN INTEGRATED APPROACH TO UNDERSTAND THE LINKAGES BETWEEN ECOSYSTEM SERVICES AND HUMAN WELL-BEING In collaboration with Thomas Dietz, Daniel Boyd Kramer, Zhiyun Ouyang, Jianguo Liu 59 Abstract Understanding and managing human-nature intera ctions is the fundamental challenge for sustainability. The ecosystem services (ES) paradigm has recently emerged as a promising approach for such efforts through quantifying ES and designing and implementing payments for ES programs. However the core of this paradigm, the linkages between ES and human well- being (HWB), remain poorly understood, primarily due to lack of quantitative indicators, integrated models, and adequate data. Here we combine data on our recently developed index systems of human dependence on ES and of HWB w ith data on indirect and direct drivers of changes collected both from hous ehold surveys and field measures at Wolong Nature Reserve, China. Using these data, we systematically examined how human dependence on ES and HWB might be affected simultaneously by other driv ers, and how human dependence on ES and other drivers might affect HWB. Our results show that external driv ers (i.e., 2008 Wenchuan Earthquake) significantly affected both households™ dependence on ES and their well-being. Such impacts differed across various dimensions indicated by sub-indices. Our results suggest that diversifying human dependence on ES helps to mitigate disaster impacts on HWB. Those disadvantaged households with lowe r access to multiple forms of capital, more property damages, or larger revenue reductions also experienced greater losses in HWB. Our findings suggest that the construction of quantitative indicators for ES and HWB and integrated models of them provides a viable approach in forwarding the understanding of linkages between ES and HWB, building coherent theories on human-nature feedb acks and on vulnerability/resilience of Coupled Human and Natural Systems, and ma naging human-nature interactions. 60 4.1 Introduction Understanding and managing human-nature intera ctions is the fundamental challenge for both environmental and socioeconomic sustainabi lity (Carpenter et al., 2009; Liu et al., 2007a; Ostrom, 2009). The scope and intensity of human-na ture interactions have been increasing in an unprecedented way since the Industrial Revolut ion (Liu et al., 2007b). The phenomenon of humans interacting with natu re was recognized long ago, but the understanding of underlying patterns and processes has been evolving slowly. This evolution in understanding human-natur e interactions has led to changes in environmental management (e.g., increases in the amount and targets of funds used for biodiversity conservation). While there are various perspectives for characterizing environmental management approaches (Adams et al ., 2004; Colby, 1991; Hughes and Flintan, 2001; Jeanrenaud, 2002; Locke and Dearden, 2005; Tallis et al., 2008), from our perspective there are currently three main coexisting paradigms in real -world practices. They are distinguished by the way of understanding human-nature interactio ns, and by the extent to which different stakeholders are engaged in designing and impl ementing programs. The first paradigm is the implementation of conservation set-aside programs (or ‚fines and fences™ approach), including protected areas and some restoration programs (e .g., croplands converting to forests/grasslands programs). As the name indicates, this paradigm vi ews human as the threat to nature and isolates human from nature rather th an acknowledges human needs a nd engages stakeholders for conservation and development. It is characte rized by setting aside areas of environmental importance (e.g., biodiversity conservation, water c onservation, and erosion control) to prevent intensive human activities resulting in local, regi onal and global pressures. The origins of this paradigm date back to 1860s when the Portuguese colonial government of Brazil created what is 61 now Tijuca National Park in 1861, followed by the establishment of the Yosemite National Park in California in 1864 and Yellowstone National Park in 1872 (Phillips, 2003). Currently, this paradigm still dominates the flows of c onservation funds. By 2011, there are over 130,700 national protected areas (an additional 43,674 without a known year of establishment were not included in this estimate) and 27,200 internationa l protected areas established in 236 countries, covering more than 24 million km 2 of surface area on the earth (IUCN and UNEP-WCMC, 2012). The second paradigm is the community conser vation approach, which has diverse labels including the Integrated Conservation and Development Projects, Community-based Conservation, Community Conser vation, and Community-based Natural Resources Management [see detailed discussion in (Berkes, 2007; Meguro, 2009)]. An underlying assumption of this paradigm, although not necessarily holds due to cross-scale impacts, is that local communities are the major threats to local eco system degradation. Thus, this paradigm is characterized as appreciating the development needs of local communities and passively engaging them in conservation policy design and implementation, a lthough the extent of engagement varies from one place/project to another. The origin of this paradigm dates back to 1980s when the World Wide Fund for Nature (WWF) attempted to addre ss some shortcomings of the conservation set aside programs (Hughes and Flintan, 2001). The third paradigm is the recently emerged ecosystem services (ES) paradigm, including pr ograms for the valuation and mapping of ES programs, and for providing payments for ES (Daily and Matson, 2008; Kareiva et al., 2011; Yang et al., 2008; Yang et al., 2013e). This paradigm recognizes that human and natural systems are coupled and there are cross-scale threats to local ecosystems. It is characterized as quantifying environmental benefi ts in biophysical and/or mone tary values, actively engaging 62 different stakeholders at multip le scales, and realigning socioe conomic benefits/costs through market-based mechanisms (Jack et al., 2008; Tallis et al., 2008; Yang et al., 2013e). The origins of this paradigm trace back to late 1990s with the publication of the influential book fiNature™s Servicesfl (Daily, 1997) and the valuation of world™s ES and natural capital (Costanza et al., 1997), as well as the implementation of paymen ts for ES programs in Costa Rica (Sanchez- Azofeifa et al., 2007). It became popular af ter the monumental Millennium Ecosystem Assessment (MA, 2005), followed by several importa nt projects such as the Natural Capital Project (Kareiva et al., 2011), and the Economics of Ecosystems and Biodiversity (TEEB) project (TEEB, 2010). The ES paradigm has two major advantages. Fi rst, it takes a holistic analytical approach and provides an integrated interdisciplinary ES framework for studying the linkages between ES and human well-being (HWB) across scales (Carpenter et al., 2006; Carpenter et al., 2009; MA, 2005). Second, it provides an operational template linking theoreti cal understandings with real-world practices for decision making (Daily a nd Matson, 2008; Tallis et al., 2008). Under this paradigm, the linkages between ES and HWB ar e the core of human-nature interactions, determining the effectiveness of those PES pr ograms (Tallis et al., 2008; Yang et al., 2013e). However, so far most efforts have been made to evaluate and map ES so as to design and implement payments for ES programs. Very litt le attention has been paid to improve the understanding of the linkages between ES and HWB largely due to the lack of quantitative methods and indicators (Carpenter et al., 2006; Carpenter et al., 2009; Yang et al., 2013a). As far as we know, no study has integrated the four components (i.e., indirect drivers, direct drivers, ES, and HWB) of the MA framework to qua ntify how ES interact with HWB. 63 To fill some of these knowledge gaps, our research team has developed two index systems (Yang et al., 2013a; Yang et al., 2013c) based on the MA framework to quantify human dependence on ecosystem services and its relationship to human well-being. The index system of human dependence on ES (IDES) includes an overall index and three sub-indices for provisioning, regulating, and cultural services. A higher value of the overall index or sub-index indicates a higher dependence on the corresponding ES, and thus higher vulnerability to the damages or losses of the corre sponding ES (Yang et al., 2013c). The index system of HWB (HWBI) includes an overall index and five sub- indices, based on MA categories, for basic material for good life, security, health, good soci al relations, and freedom of choice and action, respectively. A higher value of the overall index or sub-indices of HWB suggests a higher satisfaction of corresponding human needs (Yang et al., 2013a). We provided methods for constructing the two index systems and confirmed their validity with empirical data from our long-term research site at Wolong Nature Rese rve (Fig. 4.1), southwestern China. The Reserve contains the largest wild giant panda ( Ailuropoda melanoleuca) population (~10%) in the world, and currently has ~4,900 local human reside nts distributed in ~1,200 households. Local households™ well-being substantially depend on many ES such as agricultural and forest products (e.g., maize, cabbage, yaks, pigs, cattle, fuelw ood, mushroom, and traditio nal Chinese medicine plants), water retention, erosion control, air purification, and ecotourism, as well as other socioeconomic activities such as local and migr ant labor work (Yang et al., 2013c; Yang et al., 2013e). The Wenchuan Earthquake (M s 8.0 or Mw 7.9) occurred on May 12 of 2008, with the epicenter adjacent to the Reserve (Fig. 4.1). The Ea rthquake can be viewed as a tragic external driver that caused tremendous environmental an d socioeconomic impacts both outside and inside 64 the Reserve (Yang et al., 2013a). Thus it provides a useful finatural experimentfl of the effects of an external force on both indirect and direct drivers (e.g., economic, demographic, and social conditions of households) which in turn may infl uence local households™ dependence on ES, and their well-being. For instance, the earthquake destroyed approximately 5,200 ha of forest, accounting for 6.5% of total forest area in 2007 (Fig. 4.1). The earthquake also killed forty-eight local residents and several hundreds of workers and transport passengers within the Reserve and caused severe destruction to infrastructure such as the main road, tourism facilities, residential houses, schools and hospitals (Yang et al., 2013a). The large changes in indirect and direct drivers, dependence on ES, and well-being in a short time, though traumatic, provide an excellent opportunity to (1) test how human dependence on ES and HWB might be affected simultaneously by other drivers; and (2) ex amine how human dependence on ES and other indirect and direct drivers might affect HWB. 4.2 Materials and Methods To achieve these two objectives and push forward the understanding of linkages between ES and HWB, we combined data on IDES, HWBI, and indirect and direct drivers of changes collected both from household surveys and fiel d measures at Wolong Nature Reserve. We evaluated the simultaneous impacts of the eart hquake both on IDES and HWBI at the household level. We also constructed regression mode ls to examine how IDES affects HWBI while controlling for other indirect and direct drivers. We meas ured IDES and HWBI based on previously published methods as citied (Yang et al., 2013a; Yang et al., 2013c). For indirect and direct drivers, we considered economic, demogr aphic, and social conditions of households, as well as the level of earthquake damages using an indicator of the damage to the house of each 65 surveyed household. We also controlled for the age and gender of respondents. The descriptive statistics of variables used in regression anal yses are summarized in Supporting Information Table S4.1. We used data collected in two kinds of household surveys, a basic household survey and a HWB survey, from our long-term research site at Wolong Nature Reserve. We chose household heads or their spouses as interviewees to represent each household since our previous experience in this area suggests that they are the decision makers who are most familiar with household affairs (An et al., 2001). For the basic household survey, we random ly sampled approximately 20% of local households, that is, 220 households at WNR in 1999 measuring conditions in 1998 (An et al., 2001). We tracked the same households for repeated surveys before the earthquake at the end of 2007 (measuring conditions in 2007) and in 2010, after the earthquake, measuring conditions in 2009.. Information elicited includes household size, demographic information on each household member (e.g., age, gender, and education), housin g conditions (e.g., type and area), household income and expenditure, and social ties to lo cal leaders (people w ho work for the local government or enterprises are regarded as local elites). The overall index and sub-indices of IDES are constructed based on the basic household survey data with detailed methods shown in the cited literature (Yang et al., 2013c). In previous work both the overall index and sub-indices of IDES were found to have high validity and re liability, reasonably revealing the general pattern of households™ dependences on ES, and the variations across time, space, and different levels of access to multiple forms of cap ital (Yang et al., 2013c). For the HWB survey, we randomly sampled 169 households measuring the conditions in 2007 and 2009. The overall index and sub-indices of HWBI are developed based on the HWB 66 survey data using confirmatory factor analysis with detailed methods illu strated in Yang et al. (2013a). The overall index and sub-indices of HWBI also have high validity and reliability. The item-total correlations range from 0.30 to 0.75 a nd the Cronbach™s alpha values are 0.92 and 0.91 for 2007 and 2009, respectively (Yang et al., 2013a). The model fit statistics for the confirmatory factor analysis of the indices also show high goodness-of-fit, with overall fit statistics all above 0.97 and significant coefficients (p < 0.05) for all paths (Yang et al., 2013a). There were 101 households that were samp led both in the basic household survey and HWB survey in both 2007 and 2009. We therefore used data from these 101 households for our analyses. We discounted all monetary values used for analysis into constant values of 2007. We performed all statistical an alyses in STATA 12.0 (StataCorp LP, Texas, USA). 4.3 Results 4.3.1 Simultaneous impacts of the earthquake on IDES and HWBI. The earthquake changed both local households™ dependence on ES and their well-being (Fig. 4.2 and 4.3). The average overall IDES decreased significantly from 0.634 in 2007 to 0.331 in 2009 (t = 7.190, p < 0.001). Sub-indices of provisioning and cultural services significantly decreased from 0.366 and 0.035 in 2007 to 0.051 and 0.015 in 2009, respectively (Fig. 4.2). However, there was no significant change in the sub-index for regulating services (Fig. 4.2). Additional analyses (Table S4.2) explain the differences in changes across sub-indices of IDES. On average, the earthquake reduced by 70% and 43% local households™ net benefits from provisioning and cultural services, respectively, but did not significantly affect net benefits from regulating services. Meanwhile, between 2007 and 2009, net socioeconomic benefits (those not 67 derived from ecosystems services) increased on aver age by a factor of three, largely due to the increase of temporary work for local labor after the earthquake. The earthquake significantly reduced both ove rall HWBI and its sub-indices of local households (Fig. 4.3). But the magnitude of impa cts differed across sub-indices. The earthquake significantly reduced all sub-indices of basic material for good life, security, health, good social relations, and freedom of choice and action by 5.2%, 13.2%, 12.9%, 7.2%, and 7.4%, respectively, leading to a reduction of overall HWBI by 12.6%. Additional analyses also suggest that the overall HWBI values of affluent households were significantly higher than those of poor households both before and after the earthquake (Table S4.3). 4.3.2 Effects of IDES on change of HWBI. Our results show that IDES in 2007 is positiv ely associated with the change of HWBI (Table 4.1) and that households more dependent on agriculture had larger decreases in HWBI. Because overal HWBI declined due to the eart hquake (Fig. 4.3), the negative association between agricultural income share in 2007 and the change in HWBI indicates larger decreases in HWBI for households more dependent on agriculture in 2007. Meanwhile, because the dependence on ES varied substantially across households in 2007 and many households no longer heavily depended on ES from agriculture in 2007 (Yang et al., 2013c), the positive coefficient of IDES in 2007 indicates that households more dependent on multiple ES (i.e., provisioning services from agriculture and other non-market forest resources, regulating services, and cultural services) suffered less from the earth quake in terms of HWBI. However, our results show that change of IDES is not significantly associated with the change of HWBI (Table 4.1). 68 This may be because the effect of IDES in 2007 overwhelms the effect of its change or because there is a time lag before change of IDES aff ect HWBI (Liu et al., 2007a; Yang et al., 2013a). Our results also confirm that the observed si gnificant effect of IDES on change of HWBI cannot be detected by using the specific indicators constituting it (Table S4.4). Controlling for the same independent variables in the model (Table 4.1), we trie d to replace IDES in 2007 with indicators constituting it (i .e., the corresponding net benefits obtained from provisioning, regulating, or cultural services for households in 2007). However, none of the coefficients of the alternative indicators were significant (p > 0.05, Table S4.4). These results suggest that IDES as a composite index has a combination effect on th e change of HWB that is distinct from the effects of separate indicators constituting it, which is strong additional evidence of the validity and utility of the IDES. 4.3.3 Effects of other indirect and direct drivers on change of HWBI. Our results also show that disadvantaged hous eholds with poorer economic, demographic, or social conditions or those who suffered from a higher level of earthquake damages had significantly larger decreases in HWBI (Table 4.1). Specifically, households with fewer household members, less income af ter the earthquake, and weaker social ties to local leaders reported larger decreases in HWBI. Households suffering from higher levels of house damage and more reduction of per capita income also reported larger decreases in HWBI. 69 4.4 Discussion For the first time in the literature, we have in tegrated indicators of indirect drivers, direct drivers, ES, and HWB to assess how dependence on ecosystem services and other factors affect the change of HWB after a natural disaster. Ou r results suggest that households dependent on multiple ES had smaller decreases in HWB wh ile those dependent primarily on provisioning services saw larger decreases in HWB. Our resu lts suggest that disadvantaged households with lower access to multiple forms of capital, more property damages, or larger revenue reductions experienced greater losses in HWB. If this patter n generalizes, then priorities for disaster aid assistance and post-disaster reconstruction should be given to disadvantaged or more severely affected households. In the long run, diversifying the types of ES on which households are dependent will also help to enhance resilience to disasters. Our results suggest that the earthquake significantly affected both households™ dependence on ES and their well-being. Such impacts differed across various dimensions captured by our sub-indices. Our results suggest that inside the reserve, there were significant impacts of the earthquake on the overall index and all sub-indices of households™ dependence on ES except that for regulating services (Fig. 4.2). We offer two explanations for this. On one hand, due to the massive associated disasters after the earthquake (e.g., landslides, mud-rock flows), the main road connecting the reserve with the outside world was frequently damaged or blocked. Poor transportation prevented local households from selling their agricultural products outside the reserve and dramatically reduced tour ists to the reserve. Thus, the net benefits obtained from provisioning and cultural services to local households drama tically decreased after the earthquake (Table S4.2). Nevertheless, most of the benefits from regulating services were realized through payments for ES programs, and thus did not significantly change. This 70 highlights a not much emphasized benefit of paym ent for ESŠit provides an external link that buffers local and regional disasters. On the other hand, the socioeconomic net benefits to local households increased largely due to post-disaster reconstruction efforts, which provided many governmental subsidies and temporary local employment opportunities unrelated to ES (e.g., labor work for road and housing construction). In addition, besides the significant decreases in the overall HWBI and sub-indices for basic materi al for good life, security, health, and freedom of choice and action, the significant decrease in sub-index of good social relations may be surprising. Based on our communications with interv iewees, the deterioration of social relations is due to the increasing conflicts between local households and local governments, and among local households in addressing post-disaster reconstruction issues such as the distribution of donations and government subsid ies, post-disaster relocation, and compensation for land acquisition. Our findings suggest that, to avoid or reduce social conflicts during post-disaster reconstruction, decision-making processes shou ld engage local people significantly by addressing their needs and being more transparent at multiple institutional scales (e.g., village, township, and reserve levels). Perhaps clear pl ans for the distribution of benefits could be developed before a natural disaster in order to reduce post-disaster conflicts. A significant decline in social relations may not only reduce local households™ life satisfaction but also increase transaction costs and limit the beneficial outcomes of both conservation and development programs, a resu lt suggested in many other studies (Anthony and Campbell, 2011; Bouma et al., 2008; Liu et al ., 2012; Pretty, 2003; Yang et al., 2013f). For example, in our study area our previous analysis (Yang et al., 2013f) suggests that good social relations among household group members enhanced collective action (e.g., forest monitoring) and beneficial resource outcomes (e.g., forest co ver preservation). While social ties to local 71 leaders in this study helped to reduce the earthquake impact on households™ well-being, our previous study (Yang et al., 2013f) also suggest that strong social ties in the community discourage illegal logging and reduce the amount of formal forest monitoring efforts needed. Strong social ties also increase the probability of households™ participation in ecotourism businesses (Liu et al., 2012), mitigate their de pendence on provisioning ES (Yang et al., 2013c), and reduce their environmental impacts (Yang et al., 2013c). Such evidence suggest that changes in HWB (e.g., social relations) may lead to ch anges in both socioeconomic and environmental behaviors of humans, forming feedback loops and in turn affecting indirect and direct drivers as well as human dependence on ES. Our findings suggest that the construction of quantitative indicators for ES and HWB and integrated models using them is a viable appr oach for forwarding the understanding of linkages between ES and HWB. Our integrated approach is based on the generalizable MA framework and can be applied to multiple scales and units of analysis (Yang et al., 2013a; Yang et al., 2013c), it may also be adapted to other areas and issues to test interesting hypotheses, answer important questions, and address pressing problems for sustainabi lity (Table 4.2). The improved understanding will help to build coherent theories on human-nature feedbacks, and vulnerability/resilience of CHANS (Liu et al., 2007a; Liu et al., 2007b) or Social-Ecological Systems (Collins et al., 2011; Ostrom, 2009) to guide the management of human-nature interactions. 72 Table 4.1. Factors associated with the change of HWBI before and after the earthquake. Characteristics Independent variables Coefficient (Robust S.E.) Initial HWBI HWBI in 2007 Œ 0.111 (0.078) IDES IDES in 2007 0.096* (0.045) Change of IDES 0.018 (0.026) Household economic conditions Agricultural income share in 2007 Œ 0.118* (0.051) Household income in 2009 (thousand yuan) 0.405 e-03** (0.122 e-03) Change of per capita income (thousand yuan) Œ 0.980 e-03* (0.395 e-03) Household demographic conditions Household size in 2007 0.016** (0.006) Number of seniors (age >= 60 years) Œ 0.028 (0.018) Average education of adults in 2007 (year) 0.007 (0.004) Female adult share in 2007 Œ 0.113 (0.072) Household social conditions Social ties to local leaders (0: weak, 1: strong) 0.052* (0.025) Earthquake damage House damage (0: low, 1: high) Œ 0.049** (0.018) Respondents™ characteristics Gender of interviewee (0: female, 1: male) 0.031 (0.019) Age of interviewee (year) 0.105 e-04 (0.001) Constant Œ 0.052 (0.077) Notes: We compared different combinations of independent variables shown in Table S4.1 and presented the final model here based on theore tic interest and the highest goodness-of-fit. Dependent variable is the change of HWBI before and after the earthquake, which is HWBI in 2009 subtracting HWBI in 2007. R-squared of the ordinary least square regression is 0.272. The number of observations is 101. *p < 0.05, **p < 0.01, *** p < 0.001. Variance inflation factors are tested to be less than 5. 73 Table 4.2. Example hypotheses and questions for fu ture research on human-nature interactions. Hypotheses/Questions Heterogeneity Human dependence on ecosyste m services and human well-being vary across time and space in all couple d human and natural systems. Contextual effects or path dependence Agents (e.g., individual, household, enterprise) have high dependence on ecosystem services predisaster are also more likely to do so postdisaster. Nonlinear effects and thresholds The effects of disasters on human de pendence on ecosystem services and human well-being are nonlinear. When the magnitude of impacts cross certain thresholds, irreversible shifts may occur (e.g., relocation of human settlements). Time lags and legacy effects There is a time lag or legacy effect of changes in human dependence on ecosystem services on human well-being. Spillover effects Changes in human dependence on ecosystem services or human well-being (e.g., social relations) of one agent may also affect human dependence on ecosystem services and/or human well -being of its surrounding agents (e.g., neighbors, relatives, friends). Reciprocal effects and feedback loops Changes in indirect and direct drivers may affect human dependence on ecosystem services and human well- being. In turn, changes in human dependence on ecosystem services and human well-being may alter people™s behaviors (e.g., energy use, land use practices), affect indirect and direct drivers (e.g., changes in climate, land use and land cover), and thus form feedback loops. Policy How do institutional or technology innovations affect human dependence on ecosystem services and human well-being? How does the implementation of polic ies (e.g., Integrated Conservation and Development Projects) affect human dependence on ecosystem services? What are the drivers behind changes in human dependence on ecosystem services? There are interaction effects among different policies in changing human dependence on ecosystem services or human well-being. One policy may enhance (i.e., synergic effect), offset (i.e., trade-off effect ), or even reverse the effect of another policy. When the underpinning mechanisms of such effects are not well-understood, we ma y regard them as unanticipated outcomes or surprises. 74 Figure 4.1. Wolong Nature Reserve in Wenchuan C ounty, Sichuan Province, southwestern China. 75 Figure 4.2. Impacts of the earthquake on huma n dependence on ecosystem services. Notes: Unit of analysis is the household. N = 101. *p < 0.05, *** p < 0.001. 76 Figure 4.3. Impacts of the earthquake on human well-being. Notes: Unit of analysis is the household. N = 101. **p < 0.01, *** p < 0.001. 77 CHAPTER 5 NONLINEAR EFFECTS OF GROUP SIZE ON COLLECTIVE ACTION AND RESOURCE OUTCOMES In collaboration with Wei Liu, Andrés Viña, Mao-Ning Tuanmu, Gu angming He, Thomas Dietz, Jianguo Liu 78 Abstract For decades, scholars have been trying to determine whether small or large groups are more likely to cooperate for collective action and successfully manage common-pool resources. Using data from the Wolong Nature Reserve si nce 1995, we examined the effects of group size (i.e., number of households monitoring a single fo rest parcel) on both collective action (forest monitoring) and resource outcomes (changes in forest cove r) while controlling for potential confounding factors. Our results demonstrate that group size has nonlinear effects on both collective action and resource outcomes, with intermediate group size contributing the most monitoring effort and leading to the biggest forest cover gain. We also show how opposing effects of group size directly and indirectly affect collective action and resource outcomes leading to the overall nonlinear relationship. Our findings suggest why previous studies have observed differing and even contradictory group -size effects, and thus help guide further research and governance of the commons. They also suggest that it should be possible to improve collective action and resource outcomes by altering factors that lead to the nonlinear group-size effect, including punishing free riding, enhancing overall and within-group enforcement, improving social capital across groups and among group members, and allowing self-selection during the group formation process so members with good social relationships can form groups autonomously. 79 5.1 Introduction Groups are basic units for collective acti on and may achieve outcomes that individual efforts cannot (Esteban and Ray, 2001). But the threat of free riding imp lies that the optimal amount of collective action does not always occur, a nd has led to a substantial literature trying to understand what factors facilitate or block the emergence of co llective action. Because collective action is needed to manage many common-pool re sources, understanding the mechanisms that shape collective action and resource outcomes is a critical challenge for sustainability (Ostrom, 1990; Poteete et al., 2010). From Pareto in 1906 (Pareto, [1906] 1927) and especially since the influential work by Olson in 1965 (Olson, 1965), group size has been hypothesized as a crucial factor affecting collective action and resource outcomes. (We note that Olson used an unusual definition of figroup sizefl: the potential number of group memb ers. Here we follow conventional practice and consider the actual number of participants.) Howe ver, the debate on group-size effect continues with some researchers arguing that it is linear and negative (Baland and Platteau, 1999; Olson, 1965; Ostrom, 2005), others arguing for linear a nd positive (Agrawal and Chhatre, 2006; Haan and Kooreman, 2002; Isaac et al., 1994; Zhang a nd Zhu, 2011), and still others insisting it is curvilinear (Agrawal, 2000; Agrawal and Goyal, 2001; Poteete and Ostrom, 2004), ambiguous (Chamberlin, 1974; Esteban and Ray, 2001; Oliver and Marwell, 1988; Pecorino and Temimi, 2008), or nonsignificant (Gautam, 2007; Rustagi et al., 2010; Todd, 1992). Even in the most recent work (Boyd et al., 2010; Carpenter, 2007; Gautam, 2007; Hwang, 2011; Mathew and Boyd, 2011; Pecorino and Temimi, 2008; Zhang a nd Zhu, 2011), a consensus on the nature of the effect or even its exis tence still remains elusive. 80 Previous literature indicates that ther e are two hypothetical opposing forces through which group size affects collective action and resource outcomes (Fig. 5.1). Group members play different roles in collective action, ranging from free riders (i.e., members who enjoy group benefits without paying for the costs) and conditional cooperators (i.e., members who will contribute more when others contribute more ) to altruists (i.e., me mbers who contribute regardless of others™ behaviors), as well as variou s roles mixing these strategies (Fischbacher et al., 2001). Group size can have diverse effects. On one hand, members tend to free ride as the group becomes larger (Baland and Platteau, 19 96; Olson, 1965). As group size increases, transaction costs (e.g., communication costs, costs of monitoring to maintain a necessary level of excludability) may rise sharply (Agrawal, 2000; Agrawal and Goyal, 2001; Esteban and Ray, 2001; Ostrom, 2005; Pecorino and Temimi, 2008); t hus, the larger the group, the more difficult to detect and reduce free riding. If the common good has any degree of rivalry, average individual payoff will shrink as group size in creases, which further aggravates free riding (Chamberlin, 1974; Oliver and Ma rwell, 1988; Pecorino and Temimi, 2008). On the other hand, small groups often lack the resources (e.g., labo r, time, funds) that large groups can deploy (Agrawal, 2000; Agrawal and Goyal, 2001; Ostrom, 2005; Tucker, 2010). When available resources are limited, it is difficult to devote ad ditional resources to collective action (Esteban and Ray, 2001; Pecorino and Temimi, 2008). Taking advantage of more resources, large groups may enhance enforcement through monitoring and punishment to reduce free riders and thus improve collective action and resource outco mes (Agrawal, 2000; Agrawal and Goyal, 2001; Boyd et al., 2010; Chhatre and Agrawal, 2008; Mathew and Boyd, 2011; Rustagi et al., 2010). Ostrom scrutinized previous evidence and poi nted out the problem of focusing on group size itself without considering factors that influence or are influenced by group size (Ostrom, 2005). 81 She then suggested further research to focus on the hypothesized curvilinear effects of group size (Ostrom, 2005). A few previous studies qualitatively described the curvilinear or nonlinear effects of group size (Baland and Platteau, 1996; Poteete and Ostrom, 2004; Wade, 1988) , and some claimed a nonlinear relationship by simply plotti ng collective action against group size without controlling other factors (Agrawal, 2000; Agrawal and Goyal, 2001). However, none has provided a quantitative analysis of field evidence while controlling potential confounding factors as suggested by Ostrom. Furthermore, there is litt le empirical examination of the mechanisms of nonlinear group-size effects, which is essential to guide commons governance. To fill these knowledge gaps, we used empirical data from our long-term studies (An et al., 2006; An et al., 2003; Bearer et al., 2008; Chen et al., 2012b; He et al., 2009; He et al., 2008; Hull et al., 2011; Linderman et al., 2005b; Liu et al., 2001; Liu et al., 1999a; Liu et al., 2012; Tuanmu et al., 2010; Viña et al., 2008; Yang et al., 2013a; Yang et al., 2013c) in Wolong Nature Reserve, Sichuan Province, China (N 30 û45' Œ 31 û25', E 102 û52' Œ 103 û24'; Fig. 5.2). Wolong Nature Reserve is home to ~10% of the total wild giant panda ( Ailuropoda melanoleuca) population, and home to ~4,900 local human reside nts distributed in ~1,200 households. In response to degradation of forest and panda habitat due to huma n activities since the 1970s (Liu et al., 2001), the Reserve implemented the Natura l Forest Conservation Program (NFCP) in 2001. NFCP is a nationwide conservation program that aims to conserve and restore natural forests through logging bans, afforestation, and monito ring, using a payments-for-ecosystem-services scheme to motivate conservation behavior (Liu et al., 2008). Of the total ~120,500 ha in the NFCP monitoring area in Wolong, ~40,100 ha were assigned to ~1,100 rural households while the remaining areas were monitored by the st aff of the reserve™s administrative bureau. 82 Meanwhile, the bureau set two timber checkpoints at the two ends of the only main road crossing the reserve (Fig. 5.2). The common-pool resource in question in the Reserve is the forest (an essential component of the panda habitat) assigned to households . Because logging is largely the action of local residents (Supporting Information [S I] Section 2.4.1), collective action (i.e., forest monitoring) has the potential to reduce ille gal logging and improve resource outcomes (i.e., changes in forest cover). The bureau administering NFCP has assigned the forest parcels to household groups of various sizes ranging from 1 to 16 (Table S5.2). Parcels distant from households were assigned to large groups with slightly higher payments (Table S5.2). Households could not choose which parcel to monitor or which household groups to participate in. Our analyses indicate that the distance from a household to its monitored parc el and NFCP payment do not affect the group- size effects (SI Section 2.4.3). Thus, the current distribution of group size is suitable for examining the group-size effects and mechan isms. Each assigned household group decides autonomously on its monitoring strategies (e.g., monitoring frequency, duration, and whether to subdivide to monitor in turns). The bureau ev aluates the monitoring performance based on field assessments of illegal activities (e.g., logging) and rewards people who report illegal activities (in cash). All households within a group share the same monitoring respon sibility and suffer the same payment deduction when any illegal activit ies are detected by the bureau in their co- monitored parcel. However, they are exempt from penalties if they report lawbreaker(s), in which case the corresponding lawbr eaker(s) is punished instead. 83 5.2 Materials and Methods To understand the group-size effects and the underpinning mechanisms, we combined data on characteristics of households, household groups, and monitored parcels (SI Section 1). We acknowledge that conflicts with regard to monitoring might occur within a household, but because the policy is designed to treat househol ds, not individuals, as monitoring units, the common practice of treating households as the unit of analysis is appropriate here. We measured household monitoring efforts by the total amount of labor input (one unit of labor input is defined as one laborer working for one day, SI Section 2.1) through surveys. We measured resource outcomes as changes in forest cover derived from previously published forest-cover maps (SI Section 1.1.1). We also measured factor s that might explain the mechanisms, including free riders (i.e., households that did not participate in monitoring), the level of within-group enforcement (i.e., strong enforcement if there are punishment measures for free-riding members within the group; otherwise, weak enforcement), and within-gr oup division (i.e., whether groups divide into subgroups to conduct monitoring in turns) (SI Section 2). Some other contextual factors shown in previous studies to affect gr oup size, collective action, or resource outcomes were used as control variables (SI Section 2.3). We acquired the map of household monitori ng parcels and associated documentation (e.g., the number of households that monitor each forest parcel) from the administrative bureau of Wolong Nature Reserve. To estimate forest -cover change, we used previously published forest-cover maps derived from Landsat imager y in 2001 and 2007 (Viña et al., 2007; Viña et al., 2011). These maps included two main land-cover cla sses (i.e., forest and nonforest) with overall accuracies between 80% and 88% using independ ent ground-truth data. Topographic data such as elevation, slope, and the Compound Topographic Index, a relative measure of wetness 84 (Gessler et al., 1995), were obtained from a digital elevation m odel at a spatial resolution of 90 m/pixel (Berry et al., 2007). We measured all household locations (~2,200 households) inside and surrounding the Reserve using Global Posi tioning System receivers. We calculated geographic metrics of forest pa rcels and households using the software of ArcGIS 10.1 (ESRI Inc., California, USA). These metrics include pa rcel size, parcel size per household, average elevation, average slope, average wetness, distance between each parcel and the nearest household, distance between each parcel and the main road, distance between each household and its monitored parcel, distance between each hous ehold and the main road, initial forest cover in 2001, and the percent of forest-cover change from 2001 to 2007. To understand the NFCP planning, impleme ntation, evaluation, and decision-making processes and to prepare for the household interview, we invited eight Reserve administrative staff for focus group interviews and five officials who were and/or are in charge of the NFCP for personal interviews. We used best availa ble household survey data containing NFCP implementation information in 2007 and 2009 from our long-term study in the Reserve, which has been tracking ~220 randomly sampled house holds across the years since 1998 (An et al., 2001). The panel survey elicited basic informati on such as demographic status, socioeconomic conditions, and energy use (An et al., 2002). In the 2007 and 2009 surveys, besides basic information from panel surveys, we also aske d questions regarding NFCP implementation (e.g., NFCP payments, monitoring frequency, time spen t for each monitoring, monitoring strategy [e.g., within-group division], and within-group enforcement). A total of 156 randomly sampled NFCP participating households in 2007, covering the full range of group size (i.e., 1 to 16), were used to examine how group size affect s collective action (i.e., household forest monitoring). The 113 85 households who monitored NFCP parcels with gr oup size larger than one (i.e., 2 to 16) in 2009 were used to examine the mechanis ms of nonlinear group-size effects. We first used a Tobit model to examine th e effect of group size on monitoring efforts at the household level. We then used a spatial auto regressive model to examine the effect of group size on forest-cover change at the parcel level. Finally, we conducted the path analysis to test the two hypothetical, opposing forces on the mechanisms of nonlinear group-size effects. Detailed descriptions of data collection, processing, and model specification and construction were provided in SI. 5.3 Results Our results show that group size has a non linear effect on the monitoring efforts per household, with an intermediate group size cont ributing the most (Table 5.1 and Fig. 5.3A). These results are consistent wh ether or not we include the households who monitored parcels individually (i.e., group size of one) and when using different comb inations of control variables (Table S5.13). The effect peaks at a size of eight or nine households where a household spends 9.2 labor units per year monitoring its forest parcel. Our results also indicate that some other factors besides group size matter substantially. Th e level of social ties to local leaders has a significantly negative effect on per household monitoring efforts (Table 5.1). When all other variables are at their mean values, households with strong social ties to local leaders on average input 54% less labor units than households with weak social ties to local leaders. Our experience in the Reserve helps explain this effect. The st aff members in the administrative bureau who are in charge of combatting illegal logging activities are hired from outside the Reserve, and anyone 86 can report illegal logging and receive a cash rewa rd from the administrative bureau. We are also not aware of a single case in which staff member s turned a fiblind eyefl to illegal logging so households with strong ties could avoid monitori ng or sanctions. Rather, additional analyses (SI Section 2.4.2) reveal that, compared to households with weak social ties to local leaders, households with strong social ties often have more social re lationships, power, knowledge, and experience. Our extensive fieldwork experience at th e site indicates that these social ties provide social capital and reputation that discourages others from conducting illeg al activities in their monitoring parcels and thus reduce the need for them to spend efforts on formal monitoring. The distance between each household and the main road has a positive effect on a household™s monitoring efforts, with distant households doing more monitoring (Table 5.1). The average household that lives 1 km further from the main road on average spends 33% more labor units in forest monitoring. Additional analys es (SI Section 2.3) suggest that households far from the main road are closer to the par cels they monitor (Spearman™s = Œ 0.201, p < 0.05). Our results demonstrate that group size also has a nonlinear effect on changes in forest cover, with an intermediate group size leading to the biggest gain (Table 5.2 and Fig. 5.3B). These results are consistent whether we include the parcels monitored by single households (i.e., group size of one) or not (SI Section 2.5.2). The effect peaks at a size of nine households where the forest cover increases 15.8% in comparison to the reference level in 2001. The effects of slope, wetness, initial forest cover in 2001, and spatial error correlation are also significant (Table 5.2). We accounted for as many as possible alternat ive explanations of the observed nonlinear group-size effects based on systemat ic quantitative and qualitative analyses. No factor other than group size seems to account for the observed nonlin ear effects. First, correlation tests (Table 87 S5.2) show that except for the two criteria used for household group assignment (see details at SI Section 1.2) by the administrative bureau (i.e., distance between each household and its assigned parcel and received NFCP payment), no other f actors were significantly associated with group size and thus are implausible as possible alternative explanations for the group-size effects. We used two additional approaches to ensure that the observed nonlinear effects were not caused by the two criteria used for household group assignments (SI Section 2.4.3). We examined the associations between the two criteria used for household group assignment and household monitoring efforts and we estimated two-step Tob it models of monitoring effort. Using either approach, all hypothesized alternatives to group size were linearly associated with household monitoring efforts, and thus could not lead to the observed nonlinear effects. Our path analysis (Table 5.3) confirms that group size has effects through the two opposing forces (Fig. 5.1). If the balance between positive and negative effects shifts with group size, it can yield the observed nonlinear pattern. On one hand, group size has a significantly positive effect on the probability of a household free riding (p < 0.01, Table 5.3). With all other relevant factors controlled at their mean va lues, an increase of group size by one household increases the free-riding probability by 15%. On the other hand, group size has a significantly positive effect on within-group enforcement (p < 0.01), which significantly reduces free riding (p < 0.01, Table 5.3). Again, controlling all other rele vant variables at their mean values, an increase in group size by one household strengthens within-group enforcement by 10%, whereas a shift from weak to strong within-group enfo rcement reduces free riding by 52%. Additional analyses (SI Section 2.4.4) suggest that as groups become la rger, a group member would face higher pressure of deteriorating social relationships with the other members in each group, which enhances within-group enforcement and thus redu ces free riding. This result is consistent with 88 the significant effect of social ties on household monitoring efforts (Table 5.1), indicating that social capital plays an important role in affecting conservation behaviors of households. It follows that collective action might be easier to maintain when social relationships among group members are improved or members with good so cial relationships can form their groups autonomously. 5.4 Discussion The coexistence of two opposing forces may also explain why previous studies found different group-size effects. If, as we argue, the net effect of group size is determined by the dynamics (e.g., strength and varia tion with group size) of the two opposing forces, the optimum point of the net effect (or the necessary range of group size to observe a nonlinear effect) would be dependent on the context (Agrawal, 2000). The range of group size in our study area may appear to be small. However, th e nonlinear pattern we observed m eans that such a range is large enough to exhibit the nonlinear effect in our context. One of the reasons we find such effects with only moderate variation in group size may be because our study area is a flagship nature reserve for giant pandas. As a result, the loca l administrative bureau has relatively abundant resources to allocate payments for household groups to mon itor parcels and evaluate their performance biannually (SI Section 1.2). A nd many household activitie s are substantially affected by kinship and leadership, so it is not surprising that social capit al matters substantially in household monitoring efforts and resource ou tcomes. Neither of these conditions might hold in other contexts where official engagement is less pronounced and social capital is of less importance. In our context, the optimum point can be detected even though no group is larger than 16. In other contexts, a larger range of gr oup size might be necessary to detect nonlinear 89 effects, which raises an important issue for future investigation: what elements of context influence the optimum point in the relationships between group size and either provision of collective action or resource outcomes? Our study uses intensive analyses based on quantitative and qualitative data buttressed by years of fieldwork at the site to examine the effect of group size on per household effort and resource outcome. We acknowledge that the optimal group size may vary across contexts. In some commons management regimes, the vari ation in group size may not be great enough to demonstrate the nonlinear effect. The approach we have used could readily be applied to other contexts. When a literature based on analyses like ours at other sites emerges, comparison across studies would allow the identification of what aspects of context influence optimal group size, something that cannot be done in a single study. Randomized experiments are sometimes seen as the figold standardfl for research on causal mechanisms. But there have been no randomized experiments at our site, nor are there likely to be because of its status as a showcase for conservation efforts. In addition, in the real world, there is no randomized or even quasi-rando mized field experiment in this field of study. The best that can be done in many real-world re source management situations is to be careful with regard to inference. Our analyses show that significant advances in understanding can be made through careful analyses of nonexperiment al data by drawing on historical data. Such efforts of ongoing programs provide a useful co mplement to field experiments in building a cumulative literature and forwarding the impor tant work on collective action and resource management. 90 Our findings also suggest that by regulating factors interacting with group size, it should be possible to improve collective action and re source outcomes. For instance, all groups of various sizes can stimulate group members to c ontribute and protect common-pool resources by punishing free riding and enhancing overall and within-group enforcement. Overall enforcement can be enhanced not only through intensifying costly monitoring efforts but also via improving social capital across groups. The within-group enforcement and outcomes may also be enhanced by improving social capital among group members or allowing self-selection during the group formation process so members with good social relationships can form groups autonomously. Unprecedented deterioration of global commons requires better understandings of the mechanisms shaping collective action and resource outcomes. Due to the complexity of coupled human and natural systems (Liu et al., 2007a), improving such understandings is challenging and requires efforts to integrate da ta and methods from multiple disciplines. The struggle to understand the group-size effects is one example showing the importance of such efforts. Our findings help disentangle the puzzle of group-size effects and guide solutions to pressing problems of coupled human and natural systems (Liu et al., 2007b) as well as the design of commons governance policies. 91 Table 5.1. Coefficients of the Tobit model for the nonlinear effect of group size on collective action. Variable Coefficients (Robust S.E.) Marginal effects Intercept 8.921*** (2.360) Œ Quadratic term of group size Œ 0.128** (0.041) Œ Group size 1.331** (0.408) 0.767 Social ties to local leaders (binary: 0 for weak social ties; 1 for strong social ties) Œ 5.377** (1.920) Œ 3.012 Distance between each household and the main road 2.787* (1.216) 1.749 Additional controls Not significant (Table S5.9) Œ Notes: Unit of analysis is the household. Dependent variable is total labor input for monitoring per year. Additional controls include household si ze, number of household laborers, education of adults, household income, and percentage of agricultural income (T able S5.9). Log pseudo-likelihood is Œ 390.962. Total number of observations is 156. Independent variables were mean centered before entering the mode l. *p < 0.05; **p < 0.01; ***p < 0.001. 92 Table 5.2. Coefficients of the spatial autoregressive error model for the n onlinear effect of group size on resource outcomes. Variable Coefficients (S.E.) Intercept 0.146*** (0.015) Quadratic term of group size Œ 1.056E-03* (4.800E-04) Group size 7.205E-03* (3.643E-03) Slope 0.339** (0.121) Wetness 0.048*** (0.012) Initial forest cover in 2001 Œ 0.269*** (0.030) Additional controls Not significant (Table S5.16) (Coefficient of spatial error correlation) 0.561*** Moran™s I 0.021 Notes: Unit of analysis is the forest parcel. Dependent variable is the percent of forest-cover change from 2001 to 2007. Additional controls in clude parcel size, pa rcel size per household, elevation, distance between each parcel and the nearest house hold, and distance between each parcel and the main road (Table S5.16). Total number of observations is 151. Log likelihood is 170.281. Independent variables were mean cente red before entering the model. Detailed discussion of the spatial autoregressive models were in SI Section 2.5.2. *p < 0.05; **p < 0.01; *** p < 0.001. 93 Table 5.3. Path analysis of the two opposing forc es through which group size affects collective action. Path analysis Unstandardized coefficient (S.E.) Dependent variable: Free ride r (binary: 0 for a household that does not free ride; 1 for a household that free rides) Group size 0.146** (0.051) Within-group enforcement (binary: 0 for weak enforcement; 1 for strong enforcement) Œ 0.522** (0.184) Dependent variable: Within-group enforcement Group size 0.103** (0.038) Within-group division (bin ary: 0 for no within-group division; 1 for within-group division) 0.376 (0.266) Group size × Within-group division Œ 0.050 (0.061) Dependent variable: Group size Social ties to local leaders (bin ary: 0 for weak social ties; 1 for strong social ties) 0.052 (0.651) Distance to main road (log) Œ 0.067 (0.136) Number of laborers Œ 0.051 (0.350) Household size 0.027 (0.243) Education of adults 0.016 (0.117) Household income (log) Œ 0.093 (0.311) Percentage of agricultural income 1.839 (0.946) Notes: Unit of analysis is the household, but both characteristics of households and their assigned groups are considered. Continuous independent variables are mean centered. All goodness-of-fit indices show that the model fit is respectably high (Table S5.5). Total number of observations is 113 households. **p < 0.01.94 Figure 5.1. Hypothetical effects of free riding, within-group enforcement, and group size on collective action and resource outcomes. Notes: Both free riding and within-group enforcement are hypothesized to be positively related to group size. But free riding is hypothesized to be negatively related to within-group enforcement. The combined effects of free ri ding and within-group enforcement on collective action and resource outcomes are not expected to be additive because of interactions between within-group enforcement and free riding. The ne t effect of group size is determined by the dynamics (e.g., strength and vari ation with group size) of free riding and within-group enforcement, which may form a nonlinear pattern. Free riding/Enforcement Free ridingEnforcementCollective action/Outcomes Group size Group size 95 Figure 5.2. Map of the location, main road, fo rest cover in 2007, and household monitoring parcels of Wolong Nature Reserve in Sichuan Province, China. 96 Figure 5.3. The nonlinear group-size effects on collective action and forest outcomes. Notes: This figure shows the predicted monitoring effort ( A) and forest-cover change (B) from 2001 to 2007 under different group sizes (i.e., number of households monitoring a single forest parcel). The graphs show the net effects of group size on per household monitoring effort and on change in forest cover, while controlling the other variables in Tables 5.1 and 5.2. The blue line is the predicted fit based on group size, and the orange dots are the actual observations. One dot may represent several overlappi ng observations. Except for linear and quadratic terms of group size, all other independent variables were controlled as their mean values (Tables S5.1 and S5.3). In B our conclusion still holds as the nonlinear effect is still si gnificant even when excluding the parcels with group size of one, or the two parcels with group sizes of 15 and 16 (see details in SI Section 2.5.2). However, for A and B, the observations do not visually fit the predicted lines in the same way as the observations in ordinary least-squares regressions (Daniel et al., 1999) because these models are not the ordinary least-sq uares regressions (see details in SI section 2.5). 97 CHAPTER 6 INTERACTION EFFECTS OF CONSERVA TION AND DEVELOPMENT POLICIES ON RURAL HOUSEHOLD INCOME AND INCOME STRUCTURE In collaboration with Thomas Dietz, Zhiyun Ouyang, Jianguo Liu 98 Abstract Numerous conservation and developm ent policies have been implemented simultaneously around the world, but li ttle attention has been given to how these policies interact with each other. Using data collected since 1995 in Wolong Nature Reserve, we examined both the main and interaction effects of conserva tion and development policies on rural household income and income structure. Our study conf irmed the existence of both synergistic and antagonistic effects among differe nt conservation and developmen t policies, which are seldom given much attention to during the design, implementation and/or evaluation of conservation and development policies. Such interaction effects also led to unanticipat ed consequences. The payments from two conservation policies, when examined separately, negatively affected household income, but jointly led to a positive ef fect. Our findings suggest that conservation and development policy studies should use an integr ated paradigm by accounting for complex policy interactions, and jointly designing, implementi ng, monitoring, and evaluating conservation and development policies. 99 6.1 Introduction To mitigate widespread ecosystem de gradation and improve human well-being, numerous conservation and development policie s have been designed and implemented and often there are multiple policies at any given locale. Increasingly, conservation and development are seen as linked, a point first made by the IUCN in 1980 (IUCN, 1980). Two distinct courses may characterize such policy intervention e fforts. On one hand, conservation policies increasingly seek to integrate development goals (e.g., poverty allevi ation) to gain public support; on the other hand, development po licies seek to incorporate conservation goals under the slogan of sustainable development (Tallis et al., 2008). Despite the effort s to integrate conservation and development goals, policies were often evaluated se parately and little atten tion has been given to how conservation and development polic ies interact with each other. Considering the potential in teractions among policies is cr itically important for several reasons. First, ignoring the intera ctions of multiple policies may l ead to biased estimates of the effects of each policy. Second, ignoring the inter action effects may lead to strategies that enhance the performance of one po licy but compromise that of anot her policy, or in some cases, may lead to failures of both policies. This is a key issue for the recently popular payments for ecosystem services (PES) programs because paymen ts for protection of one service may affect the provision of other services(Foley et al., 2005; Kareiva et al., 2011; Kinzig et al., 2011; Millington et al., 2013; Nelson et al., 2009). Third, for policies with dual or multiple goals such as the Integrated Conservation and Development Programs (ICDPs), there could also be interaction effects among differe nt policy implementation strategi es. The pervasive failures of ICDPs are in part largely due to lack of understanding such complex interactions (Liu et al., 2007a; Liu et al., 2013). Fourthly, different policies may interact w ith each other, form feedback 100 loops, and generate unanticipated or unintended consequences and these effects may differ across scales. A program to construct efficient irrigation infrastructure (e.g., drip irrigation) decreased water use on individual farms but incr eased evapotranspiration (i.e., loss of water associated with plant use), reduced groundwater recharge, and redistributed water supply at a water basin scale, which increased overall water us e at large scales and affected distant holders™ water rights (Ward and Pulido-Velazquez, 2008) . Providing incentives for more efficient irrigation infrastructure alone also cannot gua rantee less overall water use (Ward and Pulido- Velazquez, 2008). Only the combination of the two policies with other institutional innovations that account water use at multiple scales may reduce overall water use (Ward and Pulido- Velazquez, 2008). Finally, in every location on the globe, there are concurrent conservation and/or development policies being implemented at local, regional to national and global scales (Liu and Yang, 2012). While in some cases a single policy may be the major driver of social and environmental change, in most cases the influence of interaction effects of multiple policies cannot be dismissed wit hout careful analysis. Theoretically, the outcome of interactions among different po licies can present in any of the forms shown in Fig. 6.1. [We are aware that be sides multiplicative interaction terms there are other statistic forms of interactions (B lalock, 1965; Southwood, 1978). Here we use multiplicative interaction terms because they are the most common statistical specification and well capture the sorts of effects we anticipate (Aiken and We st, 1991; Friedrich, 1982; Whisman and McClelland, 2005).] When the total effect of two policies is greater than the sum of their individual effects, one will see a synergistic effect (Fig. 6.1a, 6.1c, and 6.1e). On the contrary, when the total effect of two policies is smaller than the sum of their individual effects, one will see an antagonistic effect (Fig. 6.1b, 6.1d, and 6.1f). When such different forms of complex 101 interactions are not considered, the outcomes may come as surprises, unanticipated and unintended consequences (Kinzig et al., 2011; Liu et al., 2007a). While there is some theoretic discussion of the interaction effects among policies (Liu et al., 2008; Liu et al., 2013), systematic empirical examinations of the interactions of conservation and development policies are altogether missing from the current literature. To fill these knowledge gaps, we used data from a long-term project and extensive local knowledge at Wolong Nature Reserve in China, where we have worked since 1995, to estimate both the main and interaction effects of conservation and development policies. 6.2 Methods In this study we considered the four most important conservation and development policies implemented in the reserve, where we judge importance in terms of implementation scale, payment, duration, and potential imp acts. Based on their primary goals, here we considered the Natural Forest Conservation Program (NFCP) and Grain-to-Green/Bamboo Program (GTGB) as conservation policies and the Electricity Subsidy Program (ESP) and Tourism Development Program (TDP) as deve lopment policies. Th e general background information of these four policies is shown in Table 6.1. Because all the four policies have dual conservation and development goals, which ma y profoundly affect local households™ livelihood, dependence on ecosystem services, and thus biodiversity conservation (Adams et al., 2004; Yang et al., 2013c), here we focus on household income and income structure (i.e., percentage of agricultural income) as indi cators of policy outcomes. 102 We constructed two types of econometric m odels to examine the main and interaction effects of conservation and development polic ies on changes in total household income and income structure. Controlling other factors, one type of model used the amount of payments or subsidies; the other used the percentage of income from pa yments or subsidies (i.e., the corresponding payment or subsidy divided by th e total household income). We presented the main results from the models usi ng the percentage of payments or subsidies in Tables 6.2 and 6.3. The complementary results from the models us ing the amount of payments or subsidies are shown in Tables S6.2 and S6.3. The descriptive statistics of our data are summarized in Supplementary Information Table S6.1. 6.2.1 Study area Wolong Nature Reserve (N 30 û45' Œ 31 û25', E 102 û52' Œ 103 û24'; Fig. 6.2) for Giant Pandas (Ailuropoda melanoleuca), a flagship protected area, is one of the earliest nature reserves established in China. It is situated in Sichua n Province, Southwestern China, in the transition between the Sichuan Basin and the Qinghai-Tibet Plateau. It is established in 1963 and expanded to its current size of 2,000 km2 in 1975. Currently, there are ~4,900 local residents in 1,200 households across two townships (i.e., Wolong and Gengda). It is characterized by a high variation in topographic, climatic and hydrol ogical conditions with very high level of biodiversity, supporting the largest wild population of giant pandas (~ 10% of the world™s total) and over 6,000 other plant and animal species (Schal ler et al., 1985). It is identified as one of China™s first three UNESCO Biosphere Reserves in 1979 (Li and Zhao, 1989), part of the Southwestern China Mountains biodiversity hotspot in 2000 (Myers et al., 2000), and part of the UNESCO World Heritage Site Š Giant Panda Sanctuary Š in 2006 (Liu et al., 2012). 103 Wolong Nature Reserve is also a place that has been experiencing dramatic economic and environmental transformation since the late 1990s. In the 20 th century, this area was relatively inaccessible to the outside world and was char acterized by a subsistence-based agricultural economy (Liu et al., 2012). The vast majority of local residents were and are farmers primarily involving in such agricultural activities as growing maize and ve getables, cultivating livestock (e.g., pigs, cattle, yaks, and goats), keeping bees, harvesting timber, and collecting non-timber forest products such as traditional Chinese medicine plants, bamboo shoots, mushrooms and fuelwood (Yang et al., 2013c; Yang et al., 2013e). During the 1980s, the central government started to improve infrastructure and construct conservation stations both inside and outside the reserve (Wolong Nature Reserve, 2005). In early 1980s international collaboration between the State Administration of Forestry of China and World Wildlife Fund helped to establish the world™s largest in-captivity panda breeding and research facility in the reserve, the Center for Research and Conservation of Giant Pandas (CRCGP). International aid from the World Food Programme of United Nations (US$887,000) from 1984 to 1986 and some matching fund from the central government further improved local infrastructure and converted 113 ha of cropland into forest (Wolong Nature Reserve, 2005). By the middle of 1990s the steadily increase of pandas born in captivity at CRCGP had put the re serve in the international limelight and it became a global tourism destination. In 1999, a pr ovincial highway connecting the reserve to the outside was completed and markedly improved the accessibility of this area. The annual number of tourists had increased from 20,000 in 1996 to 108,100 in 2000, with a peak of 235,500 in 2006, followed by a valley of 13,000 in 2008 due to the 2008 Wenchan Earthquake (Liu et al., 2012). However, even during the peak year in 2006, only 24% of local households received direct or 104 indirect benefits from the tour ism industry(Liu et al., 2012). As much as 96% of tourism revenue was obtained by the local government and an ou tside tourism company (Liu et al., 2012). Since 2000, in response to the rapid degradation of forest and panda habitat in the reserve (Liu et al., 2001), several national and local conservation policies were implemented, which also have important potential socioeconomic impacts. The four most important conservation and development policies in terms of implementation scale, payment, duration, and potential impacts are the Natural Forest Conservation Progr am (NFCP), Grain-to-Green/Bamboo Program (GTGB), Electricity Subsidy Program (ESP), and Tourism Development Program (TDP; Table 6.1). All the four policies included both conservation and development goals. 6.2.2 Data collection and analyses In our study area, households are the basic units of decision making for most socioeconomic activities such as agricultur al production, small business operation, local temporary work, and migrant work as well as for environmentally significant behaviors such as fuelwood collection, enrollment in GTGB, and fore st monitoring for NFCP (An et al., 2001; Liu et al., 2012; Yang et al., 2013c; Yang et al., 2013e). Therefore, we collected data at the household level. Usually household heads or thei r spouses are decision makers who are familiar with household affairs (An et al., 2001), and thus we chose them as interviewees. In 1999, our research team conducted the fi rst household survey at Wolong Nature Reserve to collect data for the year 1998 (An et al., 2001). We conducted stratified random sampling of 220 households (i.e., ~20% of all the households) based on administrative groups (an administrative unit under the village leve l). Since then, we have been maintaining good collaborative relationship with local governments and communities. Because the overall education level of local residents is low, we c onducted face-to-face interviews using interviewers 105 fluent in the local dialect and accompanied by a local assistant. Our household questionnaires elicited information of demographic status (e.g., household size, birth year, gender, and education level), and basic socioeconomic activi ties (e.g., income, expenditure, and energy use). In the end of 2007 and early of 2008, we re visited the households sampled in 1999 to collect data for the year 2007. We were able to re-interview 183 households, of which four households were excluded from analyses becau se their household regi strations were later changed to non-rural households and did not quali fy for receiving NFCP payment. The losses of other previously sampled households are due to a diversity of reasons including deaths, migration to outside areas, di vorce, and temporarily working out side the study area during our survey period. We randomly added 9 other hous eholds into the 2007 survey. In addition to repeating the questions surveyed in 1999, we also asked for information on participation in and payments from conservation policies (i.e., NFCP and GTGB). In addition to the household surveys, we also measured the locations of households using a Global Positioning System device and calculated the distance from each household to the main road using the software ArcGIS 10.1 (ESRI Inc., California, USA). 6.2.3 Econometric Models We propose that changes in rural households ™ total income and income structure are determined not only by policy inte rvention but also by different forms of capital (Li et al., 2011; Liu et al., 2012) as shown in Eq. 6.1: Y = 0 + P1 1 + P22 + C f f + Ch h + Cn n + Cb b + Cs s + Cc c + , (6.1) Where Y is the changes in total household income or changes in the percentage of agricultural income; P1 is a vector of different policy intervention variables; P2 is a vector of interaction 106 terms of different policy intervention variables; Cf is a vector of indicators of financial capital; Ch is a vector of indicators of human capital; Cn is a vector of indicators of natural capital; Cb is a vector of indicators of built-up capital; Cs is a vector of indicators of social capital; Cc is a vector of indicators of other contextual factors; is are parameter vectors to be estimated; and is the error term that has a normal distribution with mean of zero. To improve the interpretabil ity of coefficients and reduce the collinearity among the linear and interaction terms (S upporting Information [SI] Appendi x E, supporting text), all continuous independent variables were mean-cente red prior to their input into the regression models (Schielzeth, 2010). All the statistical an alyses were performed using the software of STATA 12.0 (StataCorp LP, Texas, USA). Models were estimated via centered polynomial regression. 6.3 Results Our results show that NFCP payment percen tage and GTGB payment percentage are both negatively associated with changes in total household income, wh ile the coefficient of their interaction term is positive a nd significant (p < 0.001, Table 6.2). These results are consistent with the form of interaction effect shown in Fig. 6.1e, suggesting an antagonistic effect. As the results in Table 6.2 are centered polynomial re gression (Supporting Information [SI], supporting text), when all other variables are controlled at their mean values, increasing the NFCP and GTGB payment percentages by 1% and 1% actua lly on average reduce the increase in household income by 1,288 yuan (or 5.9% of the mean value) and 155 yuan (or 0.7% of the mean value), 107 respectively. The coefficient of the interacti on between NFCP payment percentage and tourism participation was negative and significant (p = 0.022, Table 6.2). This is consistent with the form of hypothesized policy interaction effect in Fig. 6.1d, also indicating an antagonistic effect. Again, controlling all other relevant variables at their mean valu es, households participating in tourism businesses actually on average had 5,274 yuan (or 24.0% of the mean value) more increase in household income than those did not participate. In addition, the initial electricity subsidy percentage, increase of agricultural income, number of laborers in 1998, and increase in the number of laborers all positiv ely contribute to the increase of total household income. But the initial cropland area and distance from each househol d to the main road both negatively affect the increase of total household income (Table 6.2). Additional analyses suggest that households own more initial cropland or loca te further away from the main road are less likely to expand their income sources such as particip ating in tourism businesses (Spearman™s = Œ 0.242, p = 0.001 and = Œ 0.218, p = 0.003, respectively). Our results also show that GTGB payment pe rcentage and its interaction term with NFCP payment percentage are both ne gatively associated with changes in agricultural income percentage (p = 0.011 and p = 0.001, respectively), but the coefficient of NFCP payment percentage is not significant (Table 6.3). These results are consistent with the forms of interaction effect shown in Fig. 6.1b, indicating a synergetic effect. When all other variables are controlled at their mean values , increasing the GTGB payment pe rcentage by 1% helps to reduce agricultural income percentage by 0.3% (or 1.6% of the mean va lue). The coefficient of the interaction between NFCP payment percentage and tourism participation was negative and significant, while neither of their main effects are significant (Table 6.3). These results are still consistent with the form of hypothesized interact ion effect in Fig. 6.1e or Fig. 6.1f, in which the 108 main effects of both policies ar e not significant while the intera ction effect is negative and significant. In addition, the increase of agricultural income, total area of cropland in 1998, and the distance from each household to the main road all have positive coefficients, and thus have negative effects on the reduction of the percentage of agricultura l income, while the effects of initial electricity subsidy percentage, initial percentage of agricultural income in 1998, and social ties to local government s are opposite in sign. 6.4 Discussion Our study confirmed the exis tence of both synergistic and antagonistic effects among different conservation and development policies, which are seldom given much attention to during the design, implementation and/or evaluation of conservation and development policies. It should be noted that it is difficult, if not impo ssible, to distinguish empirically all the possible forms of interaction effects in any single st udy. Rather, the intention of this study is to empirically substantiate the im portance of taking interaction effects into account in policy research and practice. Our findings also suggest that synergistic and antagonistic effects not only emerge between conservation and development policies, but also exist among different conservation policies or among various development policie s. The environmental and socioeconomic effectiveness and efficiency of different policies often are evaluated separately, thus ignoring other simultaneously implemented policies (Liu et al., 2008; Yang et al., 2013e). The results of such evaluations could be biased and mislead ing. Improved understand ing of complex policy 109 interactions may also reverse the currently common failures of ICDPs to achieve both environmental and socioeconomic sustainability. Interaction effects lead to unanticipated consequences when the context or underlying mechanisms are not well understood. In our case, NFCP or GTGB payment percentage separately led to a negative impact on the changes in total household income, while jointly they had a positive effect. These results may seem to be surprising but are reasonable if the context is considered. For households affected only by NFCP or GTGB, the policy impacts might not be large enough to draw the attention of some households (e.g., NFCP payment accounts for less than 5% of total household income for 53% of households) or they might adopt incremental adaptation strategies (Kates et al., 2012) to offset the negative policy impacts. For example, based on our over 18 years™ local knowledge, to offset the negative impacts of NFCP or GTGB on their household income , many households adopted incremental adaptation strategies by adjusting their traditional agricultural activities such as switching from low income crops (e.g., corn and potato) to high income crops (e.g., cabba ge and traditional Chinese medical plants) and intensifying agricultural practices (e.g., increasing the use of fertilizers and chemical pesticides)(Yang et al., 2013c; Yang et al., 2013e). However, for many households who were both affected by NFCP and GTGP, it might be difficult for them to raise their families with the small amount of remaining cropland (on averag e less than 40% of cropland remained) even if they adopted the incremental adaptation strategies mentioned above. Therefore, they might be forced to adopt transformational adaptation strategies (Kates et al., 2012) such as substantially reducing or abandoning agricultural activities and going outside the reserve for migrant work or business opportunities in cities(Chen et al., 2012a). Such transformational adaptation strategies increased their total household income (Chen et al., 2012a) but might decrease their well-being 110 because in urban areas they may lack of quality health insurance cover and medical care, face substantial educational expenses for their child ren, and experience discrimination from urban residents, high stress and depression (Wong et al., 2007; Yang et al., 2013a). In conclusion, our findings emphasize that the design, implementation and evaluation of conservation and development policies are complex and thus require a systematic and integrated approach (Liu et al., 2007a; McConnell et al., 2011). To appropriately evaluate the effectiveness and efficiency of policies and avoid unanticipated or unintended negative consequences, different conservation and development policie s should take the interaction effects into consideration and should be jointly designed, implemented, monitored and evaluated. 111 Table 6.1. General information of the four c onservation and development policies in Wolong Nature Reserve. Natural Forest Conservation Program (NFCP) Grain-to-Green/Bamboo Program (GTGB) * Electricity Subsidy Program Tourism Development Program Aim Protect and restore natural forests through logging ban, afforestation, and payments Convert cropland to timber or bamboo forests or shrub lands with payments and alleviate poverty Encourage the switch from use of fuelwood to electricity and improve local livelihood Provide alternative income sources to local households and generate additional funds for conservation Type Conservation policy with development goal Conservation policy with development goal Development policy with conservation goal Development policy with conservation goal Implementation method Incentive-based mechanism Incentive-based mechanism Incentive-based mechanism Partnership between local government and a tourism company Initial year 2001 Green-to-Green: 2000; Green-to- Bamboo: 2002 2002 2002ƒ Duration 10 years (renewed in 2010 for another 10 years till 2020) 8 years (Grain-to- Green: renewed in 2008 for another 8 years; Green-to- Bamboo: ended in 2010) Continual unless being terminated Continual unless being terminated 112 Table 6.1 (cont™d) Natural Forest Conservation Program (NFCP) Grain-to-Green/Bamboo Program (GTGB)* Electricity Subsidy Program Tourism Development Program Implemented area 120,500 ha of land below treeline (i.e., 3600 m above sea level) Grain-to-Green: 367.3 ha from 969 households; Grain-to-Bamboo: 81.9 ha from 530 households Experimental zone of the reserve Experimental zone of the reserve Average payment rate ~900 yuan per household per year (almost tripled after 2010)⁄ Grain-to-Green: 3600 yuan per ha per year (halved after 2008); Grain-to-Bamboo: 13,500-18,000 yuan per ha per year 0.07 yuan per kilowatt-hour Not applicable Notes: * The Grain-to-Bamboo is a local program complementing th e Grain-to-Green Program and attempts to convert cropland to bamboo forest s for giant pandas. Due to the similarities in terms of aim and implementation method, together here we viewed them as one integrated cropland returning policy (i.e., GTGB). ƒ Tourism development has been proposed and adopted in Wolong Nature Reserve since 1980s. But the Ecotourism Development Plan in Wolong Nature Reserve was officially approved by the State Forestry Administ ration of China in 2002. ⁄ As officially announced by the Wolong Admini strative Bureau, the payment rate for NFCP was changed from 900 yuan per household per year into 600 yuan per capita since 2011. Only household members with an age of 7 years or above were qualified for head count. 113 Table 6.2. Effects of conservation and devel opment policies on changes in total household income. Variables Unstandardized coefficients Standardized coefficients Robust S.E.Policy intervention NFCP payment percentage Œ128.811*** Œ0.286 *** 35.194 GTGB payment percentage Œ15.535* Œ0.087 * 7.707 Initial electricity subsidy percentage 63.921* 0.097 * 28.004 Tourism participation 5.274ƒ 0.086 ƒ 3.188 NFCP payment percentage × GTGB payment percentage 387.458*** 0.263 *** 89.473 NFCP payment percentage × Initial electricity subsidy percentage Œ188.653 Œ0.042 133.730 NFCP payment percentage × Tourism participation Œ228.758* Œ0.201 * 98.906 Financial capital Initial total household income Œ0.175 Œ0.032 0.294 Changes in total agricultural income 1.114*** 0.633 *** 0.251 114 Table 6.2 (cont™d) Variables Unstandardized coefficients Standardized coefficients Robust S.E.Human capital Number of laborers 2.767* 0.148 * 1.283 Changes in number of laborers 2.161* 0.143 * 0.977 Education 0.119 0.015 0.407 Natural capital Cropland area Œ1.451* Œ0.130 * 0.578 Built-up capital Distance to the main road, log Œ1.783** Œ0.132 ** 0.563 Social capital Social ties to local governments (1: yes; 0: no) 0.319 0.004 3.919 Township 1: Wolong; 0: Gengda 4.253 0.078 2.671 Constant 14.601*** Š 2.788 Notes: Dependent variable is the change of total household income from 1998 to 2007 (i.e., total household income in 2007 subtracting total household income in 1998). All continuous independent variables were mean-centered before entering the model. The number of observations is 179. The R-squared is 0.728. Variation Inflation Factors (VIFs) were tested to be less than 5. The interaction terms between GTGB payment percentage and initial electricity subsidy percentage, and between GTGB payment pe rcentage and tourism pa rticipation were not included to avoid multi-collinearity and biased co efficient estimates. But our conclusions were consistent regardless of including or excluding these two interaction terms. ƒ P < 0.1; * p < 0.05; ** P < 0.01; *** p < 0.001. 115 Table 6.3. Effects of conservation and developmen t policies on changes in agricultural income percentage. Variables Unstandardized coefficients Standardized coefficients Robust S.E. Policy intervention NFCP payment percentage 0.904 0.141 0.663 GTGB payment percentage Œ0.347* Œ0.136 * 0.135 Initial electricity subsidy percentage Œ1.218** Œ0.129 ** 0.347 Tourism participation Œ0.037 Œ0.043 0.041 NFCP payment percentage × GTGB payment percentage Œ3.660** Œ0.174 ** 1.078 NFCP payment percentage × Initial electricity subsidy percentage 4.182 0.066 4.052 NFCP payment percentage × Tourism participation 2.398** 0.148 ** 0.870 GTGB payment percentage × Initial electricity subsidy percentage Œ1.204 Œ0.020 3.615 GTGB payment percentage × Tourism participation Œ0.218 Œ0.038 0.263 Financial capital Init ial total household income 1.14e-03 0.014 3.60e-03 Initial agricultural income percentage Œ0.859*** Œ0.685 *** 0.055 Changes in total agricultural income 6.86e-03** 0.274 ** 2.51e-03 116 Table 6.3 (cont™d) Variables Unstandardized coefficients Standardized coefficients Robust S.E. Human capital Number of laborers Œ1.75e-04 Œ6.59e-04 0.021 Changes in number of laborers Œ0.018 Œ0.083 0.015 Education Œ0.005 Œ0.041 0.005 Natural capital Cropland area 0.028*** 0.179 *** 0.007 Built-up capital Distance to the main road, log 0.022* 0.114 * 0.009 Social capital Social ties to local governments (1: yes; 0: no) Œ0.116* Œ0.096 * 0.046 Township 1: Wolong; 0: Gengda Œ0.038 Œ0.049 0.032 Constant Œ0.112** Š 0.042 Notes: Dependent variable is the change of agricultural income percentage from 1998 to 2007 (i.e., agricultural income percentage in 2007 subtracting agricultural income percentage in 1998). All continuous independent variables were m ean-centered before entering the model. The number of observations is 179. The R-squared is 0.771. Variation Inflation Factors (VIFs) were tested to be less than 5. ƒ P < 0.1; * p < 0.05; ** P < 0.01; *** p < 0.001. 117 Figure 6.1. Hypothetical outcomes of interac tion effects among di fferent policies. Notes: To simplify the illustration, here we considered interactions between two policies: (a) both policies have positive effects and the interaction effect is also positive; (b) both policies have negative effects and the interaction effect is also negative; (c) one policy has positive effect, another policy has negative effect, and the interaction effect is positive; (d) one policy has positive effect, another policy has negative effect, and the interaction effect is negative; (e) both policies have negative effects but the interaction effect is positive; (f) both policies have positive effects but the interaction effect is negative. Vertical dashed line indicates the initial implementation time of the policies. The blue li ne indicates the static baseline without policy intervention (although the baseline may not be static, here we used it to simplify the illustration). 118 The orange and red lines indicate positive and negative policy outcomes, respectively. If the effect(s) of one or two policies or their interactio n term is (are) not statistically significant, the orange or/and red line(s) will overlap with the no policy line, which is (are) just a special case(s) of the six forms. 119 Figure 6.2. Wolong Nature Reserve in Sichuan Province, southwestern China. 120 CHAPTER 7 PERFORMANCE AND PROSPECTS OF PAYMENTS FOR ECOSYSTEM SERVICES PROGRAMS: EVIDENCE FROM CHINA In collaboration with Wei Liu, Andrés Viña, Junyan Luo, Guangmi ng He, Zhiyun Ouyang, Hemin Zhang, Jianguo Liu 121 Abstract Systematic evaluation of the environmental and socioeconomic effects of Payments for Ecosystem Services (PES) programs is crucial for guiding policy design and implementation. We evaluated the performance of th e Natural Forest Conservation Program (NFCP), a national PES program of China, in the Wolong Nature Reserve for giant pandas. The environmental effects of the NFCP were evaluated through a historical trend (1965-2001) analysis of forest cover to estimate a counter-factual (i.e ., without-PES) forest cover baseline for 2007. The socioeconomic effects of the NFCP were evaluated using data collected through household interviews carried out before and after NFCP implementation in 20 01. Our results suggest that the NFCP was not only significantly associated with increases in forest cover, but also had both positive (e.g., labor reduction for fuelwood collection) and negative (e.g., economic losses due to crop raiding by wildlife) effects on local households. Results from this study emphasize the importance of integrating local conditions and understanding underlying mechanisms to enhance the performance of PES programs. Our findings are useful for the design and implementation of successful conservation policies not only in our study area but also in similar places around the world. 122 7.1. Introduction Much of the unprecedented degradation of im portant ecosystem services due to human activities (MA, 2005) has been explained using the classic ‚tragedy of the commons™ framework (Hardin, 1968; Jack et al., 2008). However, the landmark work by Elinor Ostrom and collaborators (Ostrom, 1990; Ostr om, 2009; Ostrom et al., 1999; Os trom et al., 1994; Ostrom et al., 2007) challenged this view and suggested that common pool resources can be successfully managed even without government intervention or privatization. Nevertheless, in many cases government intervention is still necessary, partic ularly when local management of common pool resources does not exist or is ineffective (Jack et al., 2008). Among the different types of government intervention, payment for ecosystem services (PES) programs have recently emerged for realigning economic and social costs/ benefits among different stakeholders through incentive-based mechanisms (Chen et al., 2012b; J ack et al., 2008; Scherr et al., 2006). A formal definition of PES as given by Sven Wunder (Wunder, 2005) as fia voluntary and conditional transaction over well-defined ecosyst em services between at least one seller and one buyerfl. During the past decade, hundreds of PES initia tives have emerged globally (Kalacska et al., 2008; Liu et al., 2008; Sanchez-Azofeifa et al., 2007; Wunder and Alban, 2008). Among these, China introduced two of the largest PES programs in terms of scale, total payments and duration (Liu et al., 2008; Task Force for Eco-Compensation Mechanisms and Polices in China, 2007). Because in many areas participation in thes e programs is not completely voluntary (Yin, 2009), they may not be regarded as PES programs but rather as PES-like programs. However, for consistency, in this article we refer to them as PES programs. One of them, the Grain-to-Green Program (GTGP, also known as the Sloping Land Conversion Program or the Farm to Forest Program), was initiated in 1999 at the national leve l to restore natural ecosystems and to mitigate 123 negative off-site effects (e.g., drought, flood, dust storm, sediment ation of reservoirs) caused by agricultural expansion onto marg inal and/or steep land (Liu, 2010; Liu and Diamond, 2005; Liu et al., 2013). By the end of 2009, GTGP had accumu lated an investment of ca. 31.7 billion USD (1 USD = 6.3 Chinese yuan, May 2012), with averages of 547.6 USD per ha and 381.0 USD per ha in the upper reaches of the Yangtze River basin and middle-upper reaches of the Yellow River basin, respectively (Liu et al., 2013). By 2009, ca. 32 million rural households enrolled ca. 8.8 million ha of cropland in the GTGP (Liu et al., 2013). The Natural Forest Conservation Program (NFCP) aims to conserve natural forests via logging bans and afforestation incentives. By the end of 2009, the NFCP had accumulated an investme nt of ca. 14.9 billion USD (Liu et al., 2013). To date both GTGP and NFCP have been funded entirely by the central and local governments. However, the GTGP is a direct PES program i nvolving local households, while the NFCP is an indirect PES program that is operated by local forestry bureaus and seldom engages local households. Only in very few areas such as in Wolong Nature Reserve for giant pandas, the decentralization of NFCP implementation and engagement of local households have been attempted. While several studies (Chen et al., 2009; Chen et al., 2010; Uchida et al., 2009; Uchida et al., 2005; Viña et al., 2013; Xu, 2002, 2004) have evaluated the effectiveness and efficiency of the GTGP, relatively few (Xu, 2002; Yin, 2009) have studied the effectiveness and efficiency of the NFCP. In addition, to our knowledge no study has evaluated the decentralized implementation of the NFCP (e.g., participation of local households) and its environmental and socioeconomic consequences. Moreover, very fe w previous studies on PES programs anywhere in the world (Arriagada et al., 2009; Gross-Ca mp et al., 2012; Sculli on et al., 2011) have addressed their environmental and so cioeconomic outcomes simultaneously. 124 Based on more than 30 years of remotely sens ed data combined with more than 10 years of household survey data, in this article we attempt to fill some of these knowledge gaps in the Wolong Nature Reserve for Giant Pandas. Our objectives are: (1) to evaluate the environmental outcomes of PES program implementation usi ng a counter-factual without-PES baseline, estimated using the historical trend of fore st cover change; (2) to capture short-term socioeconomic effects based on em pirical data before and after the implementation of the PES program; and (3) to summarize effective practices and learned experiences, as well as challenges and opportunities for guiding future conservation policy design and implementation. 7.2. Materials and methods 7.2.1. Study area We chose Wolong Nature Reserve for Giant Pandas ( Ailuropoda melanoleuca) as our study area not only because it is one of the earliest nature reserves established in China, but also because it has a relatively independent administ rative ability to design and implement its local PES programs. The reserve is characterized by a high biological diversity and supports approximately 10% of the tota l wild giant panda population (Zhang et al., 1997), making it a flagship reserve not only in China but also around the globe (Liu et al., 2003a). The Wolong Nature Reserve (N 30 û45™ - 31 û25™, E 102û52™ - 103û24™, Fig. 7.1) is located in Wenchuan County, Sichuan Province, China, in the transition between the Sichuan Basin and the Qinghai-Tibet Plateau. The reserve is characterized by a high variation in topography (e.g., elevations ranging from 1200 m to 6250 m above sea level), soils, climates and hydrological conditions(Viña et al., 2008). In 1963 (when it was first designated as a nature reserve), its size 125 was around 20,000 ha, but was expanded in 1975 to its current size of approximately 200,000 ha (Wolong Nature Reserve, 2005). Natural vegetation in the reserve is dominated by four types of forest, all of which are associated with understory bamboo species: subt ropical, evergreen broad-leaf forests (< 1600 m), evergreen and deciduou s broad-leaf forests (1600 Œ 2000 m), mixed coniferous and deciduous broad-leaf forests (2000 Œ 2600 m), and s ub-alpine coniferous forests (2600 Œ 3600 m) (Schaller et al., 198 5; Wolong Nature Reserve, 2005). In 1983, the State Council authorized the reserve to establish the Wolong Special Administrative Region with the purpose of protecting regional forest ecosystems and rare plant and animal species, but primarily for the conser vation of the iconic giant panda (Wolong Nature Reserve, 2005). The establishment of the Wol ong Special Administrati ve Region allowed the reserve to be a relatively independent administrative entity. Currently, the reserve has two townships (i.e., Wolong and Gengda), with a total human population of approximately 5700, including around 4900 local residents that are distributed in approximately 1200 households. Most local residents are farmers involved in so cioeconomic activities such as cultivating maize and vegetables, raising livestock, collecting medi cinal plants, keeping bees, collecting fuelwood, and cooking animal and human f ood (Yang et al., 2013c). Fuelwood in the reserve is primarily used for heating, cooking pig fodder, cooking human food, and smoking pork. It is a local tradition to raise pigs, smoke pork using fuelwood, and eat smoked pork. The establishment of the reserve and its expansion in 1975 did not mitigate the degradation of forest and panda habitat inside its borders (Liu et al., 2001). Therefore, several PES programs have been designed and implem ented. The first one was implemented in 1986 with funds from the World Food Programme of the United Nations, through which 113 ha of cropland were converted into forest (Wolong Nature Reserve, 2005). Since 2000, two national 126 (i.e., GTGP and NFCP) and one local (Grain-to-Bamboo Program) PES programs have been implemented in the reserve. Table 7.1 desc ribes general information on these three PES programs. The three PES programs were designed to target almost all lo cal rural households but not all households enrolled in those programs (Table 7.1). As opposed to other NFCP implementation areas in which afforestation prac tices are also included, the implementation of the NFCP in the reserve only included forest monitoring activities, targeting all the areas (approximately 120,500 ha) below the treeline (ar ound 3600 m above sea level) in the reserve. As a local PES program complementing the GT GP, the Grain-to-Bamboo Program was designed to increase the bamboo cover (giant panda™s ma in food) by enrolling cropland located within a 15m buffer zone at each side of the main ro ad (Fig. 7.1). The Grain-to-Bamboo Program has an annual payment rate of 2142.9 to 2857.1 USD per ha, depending on the distance to the main road and the cropland production relinquished. With comparison to GTGP, the payment rate of Grain- to-Bamboo Program is higher because it targets flatter and more produc tive cropland. To date, the GTGP and the Grain-to-Bamboo Program have enrolled 367.3 and 81.9 ha of cropland, respectively. But the total implemented area of GTGP and the Grain-to-Bamboo Program comprises only 0.37% of the area under the NFCP, thus direct environmental effects of these programs could be considered almost negligib le. However, the GTGP and the Grain-to-Bamboo Program may induce both direct and indirect soci oeconomic effects that should not be ignored (Chen et al., 2010). Therefore, we took the GTGP and Grain-to-Bamboo Program into account when specifically examining the social and ec onomic effects of conservation policies on local households. Besides the implementation of PES programs, the major economic development in the reserve during the past two decades has been tourism. A tourism development plan was proposed 127 and adopted as a tool to generate funds for conservation and to provide alternative income sources for local farmers. While it started in th e 1980s, this plan was not fully implemented until 2000 after the successful breeding of captive pandas and the comp letion of a provincial highway in the late 1990s (Liu et al., 2012). Thus, while only 4% of the households in the reserve participated in tourism activiti es in 1998, this increased to 28% in 2007 (Liu et al., 2012). Tourism development has grown in tandem with infrastructure development (e.g., road, hotels, and hospitals) and with government revenue. Howe ver, more than 96% of the total tourism revenue has been retained by the local government and an outside tourism company, limiting the benefits to local households (He et al., 2008; Liu et al., 2012). 7.2.2. Forest cover dynamics The dynamics of forest cover in the rese rve were obtained from forest cover maps developed in previous studies (Liu et al., 2001; Vi ña et al., 2007; Viña et al., 2011). These maps were developed using data from different sate llite platforms (i.e., Corona, Landsat MSS and Landsat TM) combined with ground-truth data acquired during field work. These maps depict information on forest cover in the re serve during the years 1965, 1974, 1987, 1994, 1997, 2001 and 2007. Image classification procedures ranged fr om visual interpretation (Liu et al., 2001) to unsupervised digital image classification (Viñ a et al., 2007; Viña et al., 2011) using the ISODATA technique, an iterative process for non-hi erarchical pixel classi fication (Jensen, 1996). Overall accuracies of these maps were between 78% and 88% using independent ground-truth data (Liu et al., 2001; Viña et al., 2007; Viña et al., 2011). Th ese maps included two main land cover classes (i.e., forest and non-forest). For a detailed description of classification procedures 128 and assessments of map accuracy please refer to the cited studies (Liu et al., 2001; Viña et al., 2007; Viña et al., 2011). 7.2.3. Focus group, individual and household interviews To understand planning, implementation, evaluation and decision making processes involved with the NFCP, we conducted focus group, individual, and household interviews. We invited eight staff members of the reserve for focus group interviews in 2002, 2007, 2009 and 2010. These people were selected because they were directly participating in the implementation and evaluation processes of the NFCP. These focus group interviews were organized to discuss forest monitoring activities before the NFCP, pe rception toward NFCP benefits and costs, and problems and challenges during the NFCP implemen tation. We also had individual discussions with five officials who were in charge of NFCP planning, implemen tation, evaluation and decision making in 2002, 2007, 2009 and 2010. The information obtained from focus and individual interviews were further verified from publis hed and unpublished governmental documents as well as through household interviews. For household interviews, we usually chose the household heads or their spouses as interviewees because they are the decision makers and are familiar with household affairs (An et al., 2001). For the household interview before th e NFCP, we used data acquired in 1999 (220 households through stratified random sampling, appr oximately 20% of the total households in the reserve) from previous studi es (An et al., 2002; An et al., 2003) by our research team. Data after the implementation of NFCP were acquired in 2002 (200 households), 2007 (192 households) and 2009 (207 households). About 170 randomly sampled households were in every wave across the different years (i.e., panel data). The panel surveys elicited basic information 129 such as demographic status, household socioeconomic activities (i.e., income and expenditure) and energy use (e.g., fuelwood and electricity). In the 2007 and 2009 surveys, besides basic information from panel surveys, we also adde d questions regarding NFCP, GTGP and Grain-to-Bamboo Program payments, and perceptions to ward NFCP implementation. All monetary measures used in analyses were discounted into constant values in the year 2007. 7.2.4. Local adaptation and implementation of the NFCP As in Costa Rica (Sanchez-Azofeifa et al., 2007), a national PES implementation authority exists for the NFCP in China. The annual budget for NFCP implementation in the reserve was 389,206 USD, of which 380,000 USD and 9206 USD were from the central and provincial governments, respectiv ely. With assistance from the Forest Inventory and Planning Institute of Sichuan Province (Fig. 7.2), th e Wolong Administrative Bureau designed and implemented a local NFCP. On the one hand, th e administrative bureau intended to reduce conflicts with local households on forest use and engage them in forest monitoring activities through economic incentives (i.e., payments). On the other hand, the administrative bureau hoped that the NFCP payment would compensate the foregone household income sources, reduce illegal logging and fuelwood collection, and even increase household income by encouraging households to switch from on-farm to off-farm economic activities. The NFCP payment rate for each household was fixed across years and was decided based on available funding from central and provincial governments, excluding admi nistrative and operational costs (e.g., costs for government patrolling and bia nnual NFCP evaluations). The administrative bureau also hoped that the switch from fuelwood to electricity would reduce the pressure on forests. Before the implementation of NFCP, th ere were eight small hydropower stations in the 130 reserve with a total capacity of 34 megawatts (Wolong Nature Reserve, 2005). But the electricity generated from these hydropower sta tions was mostly exported outside the reserve. To assist the implementation of NFCP and promote the switch from fuelwood to electricity, local power grids were upgraded and a so-called ecological hydropower station with a total capacity of 1600 kilowatts was constructed in 2002. This ecological hydropower station compensated the electricity price for local households by 0.01 USD per kilowatt-hour , leading to an actual unit residential electricity price of 0.03 USD per kilowatt-hour (Wolong Nature Reserve, 2005). Finally, an integrated local program of NFCP was implemente d, which included the logging ban, payments for household participator y forest monitoring activities, upgrading of rural power grids, compensation for electricity price, and the esta blishment of forestry police force to enhance policy enforcement. Approximately 40,100-ha forest (about one thir d of the total impleme ntation area in the reserve) was assigned to around 1130 households, with an annual average payment rate of approximately 143 USD per household (Table 7.1) , while the remaining area was monitored by the staff of the Wolong Administrative Burea u. Initially, the Wolong Administrative Bureau attempted to assign a single forest parcel to each household. However, this approach turned out to be very difficult to implement because of the difficulty in clarifying forest parcel boundaries. Therefore, the Bureau finally decided to assign large forest parcels, defined using natural boundaries (e.g., rivers, ridges, valleys), to groups ranging in size from 1 to 16 households (Yang et al., 2013f). Of the 40,100 ha of forest assigned to household monitoring activities, around 16,700 ha were assigned to individual households while the remaining areas were assigned to groups with two or more house holds. Through this approach, a household-group monitoring approach was created which binds households together for forest monitoring 131 activities. Each household in a group shared the same monitoring responsibility and received the same amount of payment as other household memb ers. The Bureau evaluates the performance of household forest monitoring activities through bia nnual field assessments of illegal activities (e.g., logging, hunting, mining, grazing in restricted areas, collection of seeds and bamboo shoots, and other activities considered to negatively affect soils, wildlife and natural vegetation) using a score-based quantitative evaluation standard for deducting payments. All local residents are encouraged to report illegal activities and will be rewarded in cash by the Bureau. If any illegal activity is detected in a forest parcel, all households in its corresponding group will have the same amount of payment deduction, unless they identif y the culprits of the illegal activity to the Bureau (Yang et al., 2013f). 7.2.5. Baseline for environmental benefits A fundamental question of any PES program is its ability to addr ess fiadditionalityfl (i.e., difference in ecosystem services provision between the with-PES scenario and a without-PES baseline). In other words, is the additionality sufficiently large to warrant implementation of a particular PES program (Wunder, 2007)? To answer th is question it is necessary to first identify a counterfactual without-PES baseline scenario. Such counterfactual scenario may take any of three forms: static, deteriorating, or improving (Wunder, 2007), which are difficult to identify if the PES program is already operational. However, it can be approximated us ing historical trends involving time periods before PES program im plementation, analyzing changes in similar without-PES study areas (e.g., adj acent sites) or analyzing changes in larger regions or administrative entities such as using entire countries as a baseline for local PES program implementations (Kalacska et al., 2008). In this st udy we chose to analyze the historical trend of 132 forest covers from 1965 to 2001 in the reserve to estimate the without-PES baseline in 2007. This is a common approach used by the REDD and REDD+ (Reducing Emissions from Deforestation and forest Degradation, conservation, sustainable forest management and enhancement of forest carbon stocks) program (FONAFIFO et al., 2012 ; Olander et al., 2008). We chose this approach for two main reasons. First, because NFCP is a national conservation policy, it was impossible to select appropriate without-PES baseline from similar study sites or use larger regional or administ rative entities (e.g., all of China). Second, before the local implementation of NFCP in 2001, forest dynamics in the reserve showed a deteriorating trend from 1965 to 2001 (Liu et al., 2001; Viña et al., 2007) while China overall had been experiencing forest gains since the middle of 1980s (State Fore stry Administration of China, 2005). In addition, we acknowledge that the rapi d socioeconomic development since the 1990s would also contribute to reduce the pressure on forests. But it should be noted also that external impacts (e.g., socioeconomic development) on th e reserve had existed since 1961, particularly with the construction of unpaved roads and had become more intens ive since 1992 with the construction of paved roads (Wolong Nature Reserv e, 2005). Thus, the historical trend in forest cover already included part of these macro-socioeconomic impacts. Other synchronous factors such as the GTGP implementation and tourism development may have interaction effects with NFCP on fo rest cover change. However, the GTGP only accounts for a very small percentage (< 1%) of the total land area of the reserve (Wolong Nature Reserve, 2005) and few local households have directly participated in, and benefited from, tourism related activities (He et al., 2008; Liu et al., 2012). Therefore, these factors were not likely to be a major driving force of land cove r change in the reserve. Moreover, the main activities responsible for deforestation in the reserve (i.e., logging and fuelwood collection) have 133 been reduced due to the implementation of NFCP (see Section 3.2 below). Our previous studies (Tuanmu, 2012; Viña et al., 2011; Yang et al., 2 013f) also suggest that the NFCP has been a major driving force of forest change in the re serve and surrounding areas since 2001. Viña et al. (2011) indicated that NFCP together with GTGP led to the forest transition in Wenchuan County, from a net forest loss in 1990s to a net fo rest increase from 2001 to 2007. Tuanmu (2012) suggested that controlling for other confounding factors different NFCP monitoring approaches and payment rates in different areas caused different change ra tes of panda habitat (of which forest is an essential component) across space. Ya ng et al. (2013b) showed that different NFCP monitoring efforts of household groups led to varied changes in forest cover. Therefore, using the historical forest trend from 1965 to 2001 to estimate the forest area in 2007 as a without-PES baseline constitutes a reasonable approximation fo r evaluating the additionality of the NFCP implementation. However, as the reserve adopted the integrated NFCP implementation approach, the NFCP additionality evaluated should be attri buted not only to the NFCP but also to other auxiliary measures (i.e., l ogging ban, upgrading of rural power grids, compensation for electricity price, and the estab lishment of forestry police force to enhance policy enforcement). 7.3. Results 7.3.1. Environmental effects of the NFCP implementation Implementation of the NFCP seems to have reduced illegal logging and increased forest cover in the reserve (Fig. 7.3 and Fig. 7.4a). All stakeholders in the focus group, individual and household interviews shared the opinion that illegal logging activities were reduced largely in response to the NFCP. 134 The equation of historical forest trend from 1965 to 2001 for linea r extrapolation of without-PES baseline in 2007 is given as: Y = Œ 857.74X + 1.79E-6 R 2 = 0.94 (7.1) where Y is the area of forest cover (ha), and X is year. An increase in forest cover is evident post-NFCP implementation. The with-PES forest cover in 2007 was significantly higher than the estimated without-PES fore st cover in 2001 (p < 0.05, Fig. 7.3). Before NFCP implementation, total forest cover decreased from approximately 106,000 ha in 1965 to around 70,000 ha in 2001, while it recovered to approximately 79,000 ha in 2007. Based on forest cover dynamics betw een 1965 and 2001, the estimated without-PES baseline for 2007 was approximate ly 68,000 ha. Therefore, the additionality of the PES program between 2001 and 2007 might be estimated as around 11,000 ha, which accounts for roughly 5.5% of the total land area in the reserve. Due to the differences in accuracy of land c over maps and other synchronous interaction effects (e.g., effects of GTGP, tourism development), the approxima tely 11,000 ha gain in forest cover cannot be completely attributed to th e NFCP. However, it constitutes a reasonable estimation of the additionality, particularly becau se of the lack of other major drivers of land cover change and of the observed transition from forest loss before NFCP to forest gain after NFCP implementation (See Section 2.5). From 2001 to 2007, the cumulative financial NFCP investment in the reserve by the central a nd provincial governments was 2.8 million USD (Wolong Nature Reserve, 2005). Considering only this direct investment a nd excluding indirect investments (e.g., tourism developm ent, donations to conservation), the cost-effectiveness ratio might be estimated as around 254.5 USD per ha of forest gained. 135 7.3.2. Socioeconomic effects of NFCP implementation The implementation of the NFCP may be triggering not only environmental but also socioeconomic effects. While many effects may not be measurable in the short term, some short- term effects are conspicuous and could be a ssociated with the NFCP implementation. For instance, a shift in the use of different energy sources (i.e., from fuelwood to electricity) was conspicuous shortly after NFCP implementati on. The amount of electricity consumption per household doubled, while the amount of labor for ce spent in collecting fuelwood almost halved after NFCP implementation (Table 7.2). In a ddition, the energy source for cooking and heating shifted significantly from high reliance on fuelwood to electricity, while the number of months using fuelwood for cooking pig fodder also decreased. Given that the energy use shift from fuelwood to electricity occurred shortly after NFCP implementati on but before the upgrading of rural power grids and the comple tion of the ecological hydropower station, this shift was most likely caused by other measures (i.e., the logging ban and payments for forest monitoring) in the integrated NFCP implementation approach. Fi nally, no significant relation between household income and labor force spent in fuelwood collection was found (Pearson™s r = 0.783, p > 0.1). Combined, these results suggest a general patter n of switching from fuelwood to electricity irrespective of household economic status. On average, total household income doubl ed from 1998 to 2001 and quadrupled from 1998 to 2007. Most (i.e., 89%, N = 183) interviewed households reported that the NFCP had brought more benefits than costs to them . Overall household perception is that the implementation of NFCP provided payment, im proved environmental quality, prevented water and soil erosion, landslides, and promoted tourism development (Fig. 7.4a). However, negative 136 effects brought by the NFCP implementation were also reported. The most reported issue was the restriction on forest use (Fig. 7.4b). Interviewees claimed that their ancestors lived in the region for hundreds of years before the establishmen t of the reserve. For several generations they have depended on forest resources (e.g., timber, wildlife for hunting, fuelwood, traditional Chinese medicinal herbs, mushrooms, bamboo shoots, and other non-timber forest products). The NFCP implementation has inhibited many of th ese activities, and thus cultural traditions and customs (e.g., using fuelwood to cook pig fodder, smoking pork with fuelwood, and eating smoked pork) have been affected. Other complain ts include the difficultyin collecting fuelwood, as 80% of the 183 interviewees reported that they had to search a larger area to collect fuelwood due to the logging ban established by the NFCP (Fig. 7.4b). Almost all households incurred wildlife-induced losses with different damage rates, although not all of them attributed such losses to the NFCP implementation. Approximately 65% of the 183 interviewees claimed that wildlife-induced losses (e.g., loss of crops and livestock due to wildlife predators such as wild boars, hog badgers, bears, and monkeys) have significantly increased after the NFCP implementation (Fig. 7.4c). Approximately 89% of the 183 interviewees were unsatisfied or very unsatisfied with government inac tion (i.e., no measures to redu ce wildlife-induced losses or no compensation for wildlife-induc ed losses) (Fig. 7.4d). 7.4. Discussion Before the NFCP implementation, the reserve suffered ecological degradation in the form of deforestation and giant panda habitat losses (Liu et al., 2001). Unlike many other areas in China, the implementation of NFCP in the reserv e engaged various stakeholders and developed a local program that integrates a logging ban with payments for monitori ng activities and included 137 multiple auxiliary measures (i.e., upgrading of rural power grids, compensation for electricity price, and the establishment of forestry police force to enhance policy enforcement). Environmental and socioeconomic outcomes of the NFCP implementation were mixed, having both positive and negative outcomes. While an increase in forest cover was observed in association with the implementation of the NFCP, it should be mentioned that little informati on is currently available for the entire reserve on the species composition of these areas of fore st gain. Field observations have shown that some of these areas seem to be dominated by exotic and potentially invasive fast-growing species (e.g., Japanese larch, Larix kaempferi). However, previous studies have shown that giant pandas and other native wildlife species including the Asiatic black bear (Ursus thibetanus) used some of these areas (Bearer et al., 2008) , sugge sting that secondary forests may also provide habitat for wildlife species and thus facilitate biodiversity conservation. In addition, the overall giant panda habitat suitability has increased, particularly in NFCP implementation areas (Tuanmu, 2012), suggesting that some of the forest areas restored have in fact become suitable habitat for the pandas. Therefore, while the enviro nmental benefits obtained from the increase in forest cover associated with the NFCP seem to be spatially heterogeneous, overall they seem to be positive, at least in te rms of habitat restoration for some wildlife species. The effects of different conservation and development policies on changes in total household income were also mixed. Results from econometric models developed by our research group (Yang et al., 2013d) suggest that GTGP and Grain-to-Bamboo Program payments, initial electricity subsidy, and tourism participation had positive effects on changes in total household income from 1998 to 2007, respectively. With all other variables controlled at their average values, a 1% increase of total GTGP and Grain-to-Bamboo Program paymen ts, or a 1% increase 138 of initial electricity subsidy would on average increase the total household income by 2% and 0.04%, respectively. Ceteris paribus, households participating in tourism activities on average had a 54% higher increase of total household income than those that did not participate. However, only less than 28% of the households participated either directly or indirectly in tourism activities. Meanwhile, NFCP payment did not have a significant effect on changes in total household income from 1998 to 2007 (p > 0.1), wh ile the interaction term between NFCP payment and tourism participation had a negative effect (p < 0.05). The drastic increase in total household income from 1998 to 2007 was largely due to selling agricultural produce (e.g., cabbages) outside the reserve, conducting local or migrant labor, and participating in tourism (Liu et al., 2012; Yang et al., 2013d). Below we summarize some of the lessons learned through the NFCP implementation in the reserve, as well as some challenges and opportunities. 7.4.1. Lessons learned through NFCP implementation Several reasons were given by the interviewed households for the significant reduction in forest cover losses (multiple response questio n), among which decentralization of monitoring activities to households (44% of 174 respondents) and payment fo r forest monitoring activities (40% of 174 respondents) were ranked as the two most important. Other reasons, such as the upgrading of rural power grids and subsidies on electricity, were also highlighted as important in reducing forest cover losses. The introduction of the forest police team for enhancing NFCP enforcement was also regarded as an important cont ributor to the reduction in forest cover losses. Such vigorous local enforcement has also been shown effective in other PES programs in 139 Ecuador (Wunder and Alban, 2008) and in several other places around the world (Chhatre and Agrawal, 2008). Vigorous local enforcement efforts together with decentra lization of monitoring activities stimulate greater participation and cooperation of service providers by curtailing expected returns from alternative illegal forest uses, hence render participation in PES programs more attractive (Chhatre and Agrawal, 2008). The decentralization of forest monitori ng to households complemented previous command-and-control measures. This decentrali zation encouraged the participation in, and compliance with, forest conservation policies. A previous study by our research group has found that panda habitat (for which fo rest cover is essential), recovered faster in household monitored parcels than in government monitored parcels, af ter controlling for other contextual factors (e.g., elevation, slope, aspect, distance to the main road) (Tuanmu, 2012) . We have also found that the effects of group size (i.e., the number of househol ds for monitoring a forest parcel) on household monitoring efforts and forest outcomes are nonlin ear, with intermediate group size (i.e., 9 in a range from 1 to 16 households) performing the best (Yang et al., 2013f). The optimum group size also shifts with context. These results conf irm the point of no ‚panacea™ or no ‚one-size-fit-all™ approach for successful management of common pool resources (Ostrom, 1990; Ostrom, 2009; Ostrom et al., 1999; Ostrom et al., 1994; Os trom et al., 2007), and explain why the practice of household-group monitoring in the reserve turned out to be quite effective. On the one hand, with smaller group sizes, households tend to free ride (i.e., do not participate in forest monitoring activities), particularly if there are inadequate punishment meas ures within the group. On the other hand, as group size increases, a household would face increasing pressure of deteriorating social relationships with other households in the group, and thus would be less likely to free ride (Yang et al., 2013f). Our previous study (Yang et al., 2013f) confirmed these two opposing 140 effects of group size and the balance between them led to the observed nonlinear response. However, such mechanisms were not consider ed by the corresponding policy makers during the initial NFCP design process. The decentralization and group size formation, although successful, were more random than planned. Therefore, fo r guiding future policy design and implementation it is essential to first consider the mechanisms underlying the pot ential success or failure of particular policy prescriptions. The synchronous reduction of fuelwood consump tion, increase of electricity consumption, and decrease of labor force for fuelwood collection activities shortly after the NFCP implementation and before the upgrading of ru ral power grids and the completion of the ecological hydropower station suggest that the logging ban and payments for forest monitoring were the most likely candidates in reducing the pressure on forests. Later, upgrading rural power grids and the ecological hydropower station may have also cont ributed to the reduction of logging and forest recovery. Such an upgrade combined with subsidies for electricity use provided an alternative energy source that simplified the switch from fuelwood to electricity. This suggests that auxiliary measures offeri ng alternative livelihoods are essential for the successful implementation of PES programs. Ra ther than relying on command-and-control measures, service buyers should then guide a nd facilitate service providers to change environmentally harmful behaviors with alternative solutions. The integration of quantitative evaluati on standards with the PES incentive and punishment mechanism served as a rigorous and si newy gate guard for policy enforcement. The essential difference between this evaluation appr oach and previous approaches was the emphasis on performance-based measures complementing with command-and-control measures. This specifically involved the design and execution of an evaluation criterion aimed at assessing the 141 outcomes of the NFCP implementation rather than the processes or procedures for household forest monitoring. Meanwhile, all the outcomes we re linked with the incentive and punishment mechanisms (i.e., PES distribution and relevant law enforcement). Without this integrated evaluation approach, the motivation for regular voluntary forest monitoring by households may largely diminish, as penalties w ould not be effectively executed. Without this performance-based evaluation measures the PES program ma y have been ineffective. 7.4.2. Challenges and Opportunities The overall environmental and socioeconomic performances of the NFCP in the reserve were mixed and there were still some unresolve d questions. For instance, could the PES program be more effective or efficient? Was it ethi cal to achieve conser vation goals with the socioeconomic costs on local communities? Here we argue that the PES program could be conducted in a more effective, efficien t, ethical and sustainable manner. First, fundamental questions on PES are, to whom, and how much should be paid? From an efficiency perspective, only those who constitute a credible threat to ecosystem services provision should be in the scope of a PES program (Wunder, 2005). But this perspective may be unfair to those who do not threaten ecosystem service provision. However, those who do not threaten the provision of ecosystem services usually do not have opportunity costs from foregone environmentally harmful activities (e.g., logging) and may even obtain benefits due to the externality of ecosystem serv ice provision. As pointed out by Wunder (2005), fithe ideal ecosystem service seller is, if not outright environmental nasty, th en at least potentially about to become sofl. The current NFCP payment rate in the reserve was designed in the year 2001 as a flat payment, which is becoming less attractive because of inflation and increasing opportunity 142 costs. Thus, ideally those who do not or poten tially will not be involve d in any threatening deforestation activities should not be included in the NFCP. Even with a fixed budget, savings from the reduction of participants may be used to appropriately increase payment rates, to make them more attractive and competitive. However, it is necessary to first develop strategies for targeting an optimal pool of par ticipant households, while keeping the selection of participants voluntary, objective, and transparent. Theoreticall y, such targeting strategies and payment levels should be jointly designed to maximize environm ental benefits with a given budget (Alix-Garcia et al., 2008; Wunscher et al., 2008). Lessons and experience on designing targeting strategies (e.g., inverse auction systems) and payment pr ograms (e.g., discriminative payment levels) may be learned from studies not only in this region (Chen et al., 2010; Viña et al., 2013) but also from other regions (Alix-Garcia et al., 2008; Uchida et al., 2005; Wunscher et al., 2008). Second, diversified funding would be critic al for sustainable management of the PES program. Theoretically, there are five basic types of service buyers, who respond to different motivations (Scherr et al., 2006), including: (1) philanthropic buyers motivated by non-use values; (2) public sector buyers (e.g., governmental administrati ons at different scales); (3) private businesses, organizations or communities who engage in private deals to secure use values or other business benefits; (4) private buyers who are under regulatory obligation to offset environmental impacts, and (5) consumers of eco -certified products and services. The current NFCP implementation is dominated by the form of public sector buyers (i.e., government compensation mechanism), which currently exhibit the greatest expectations for many sectors in China (Task Force for Eco-Compensation Mechanisms and Polices in China, 2007). However, as it is widely accepted, government compensation mech anisms have limitations that are difficult to overcome, including the lack of el asticity, difficulty in defining payment rates, high operational 143 costs, and excessive governmental fiscal pressu re, among others (Liu et al., 2008; Task Force for Eco-Compensation Mechanisms and Polices in China, 2007). Currently, except through PES programs, many non-market ecosystem services (e.g., water purification, carbon sequestration) are being consumed but are not being paid for (Jack et al., 2008; Yang et al., 2008). Thus, to secure financial sustainability of current and future PES programs, both public and private funds should be engaged. In the reserve, for exampl e, hydropower companies in the basin provide a small compensation amount for th e electricity they produce but they should provide more funds since they directly benefit from the serv ice of water conservation provided through the conservation of forests. Tourism companies should also pay for the direct benefits they receive through ecotourism and recreation activities in the reserve. Third, there is a continuo us challenge to improve th e human well-being of local households (Yang et al., 2013a) with minimal interf erence to their local traditions and customs. The economic conditions of local households are still at a low level and many are still struggling to maintain basic livelihoods. Th erefore, on the one hand, local households should be guided to find alternative income sources and increase their total income. Currently, agricultural income is the major economic source for most of the local hous eholds and is vulnerable to natural disasters (e.g., rainstorms, mountain torrents, landslides). The relatively small amount of subsidies received from various PES programs is not, neither now nor in the foreseeable future, the main income source for local households . But since the reserve is fam ous worldwide as the home to the endangered giant panda, it provides great opportunities to engage academic institutions, non- governmental organizations and indust rial enterprises at different scales to meet local needs. For instance, current ecotourism and recreation activ ities in the reserve are largely directed by tourism development companies while local communities obtain few or no benefits (He et al., 144 2008; Liu et al., 2012). The spontaneously emerge d fiHappy Farmer™s Housefl tourism and other potential solutions that are adaptive to local conditions (e.g., cultivation of traditional Chinese Medicine, eco-certified products, and multiple us es of forests, especially non-timber forest products) and that bring direct benefits to local households could also be important alternative solutions to consider (Liu et al., 2012). On the other hand, it is important to improve the quality of local elementary and high-school education (L iu et al., 2003b; Liu et al., 1999a; Liu et al., 1999b), the quality of local medical care, and social relations among households and between households and the local government (Yang et al., 2013a), all of which are essential components of human well-being (MA, 2005). Finally, both short- and long-term socioeconom ic effects related to the PES program and interaction effects among various policies (Liu et al., 2008) shoul d be addressed. Whereas some studies in other areas (Daniels et al., 2010; Uchi da et al., 2009) have discussed environmental and socioeconomic effects of PES programs, they are mostly scattered, fragmented, short term and opportunistic (Liu et al., 2008). In the reserve, although this study captured some short-term effects, many consequences are neither well- understood nor seriously tackled during the decision making processes. For instance, wildlife-i nduced losses (e.g., crop raids) are becoming a prevalent phenomenon throughout many protected areas in China due to the recovery of wildlife habitat and increase of wildlife population after the implementati on of conservation policies (Cai et al., 2008; Liu et al., 2013; Zhang and Wang, 2003). However, to date, there are no policies or regulations to address such human-wildlife conflicts, since conservation needs mostly overwhelm personal and property ri ghts in protected areas. It should be then emphasized that taking the socioeconomic effects related to the PES program into consideration is not only an ethical issue but also it is critical for the sust ainability of current and future PES programs. 145 7.5. Conclusions At the local scale, the PES program seems to have been responsible for most of the observed increase in forest cover in the reserve. Nevertheless, it had also induced some socioeconomic effects to local households su ch as impacts on local livelihoods, increasing wildlife-induced losses, and threats to local cultu re, traditions and customs. Many other short- and long-term effects as well as interaction effects with other conservation and development policies are largely unknown at this moment. Lessons learned from this case study empha size the importance of integrating local conditions as well as the need for understanding underlying mechanisms into the design and implementation of PES programs for their successful and sustainabl e performance. In the reserve, the successful performance of the NFCP include d a combination of dece ntralization of forest monitoring activities to households, enforcement effo rts, integration of PES with other auxiliary measures, and the emphasis on performance- based measures complementing command-and- control measures. Key issues regarding the effectiv eness, efficiency, ethics, and sustainability of current and future PES programs include adaptati on of the policy cycle, identification, selection and engagement of potential beneficiaries and be nefactors, appropriate and effective incentive and punishment mechanisms, diversity and security of financial sources, alternative income sources for local residents, and solutions for shor t- and long-term negative socioeconomic effects. Conservation policy planning and implement ation are complex processes. Future interdisciplinary studies are needed to disenta ngle underlying complexities such as heterogeneity across space, time and PES targeting agents, comple x interactions of driving forces, contextual 146 effects, reciprocal effects and feedback loops between human a nd natural systems. As a global conservation hotspot, the Wolong Na ture Reserve for giant pandas acts as a flagship both for China™s and global conservation practice. Lessons learned in the Wolong Nature Reserve may also guide policy design and implementation in many other places across China and around the world. 147 Table 7.1. General information about the PES programs in Wolong Nature Reserve. Item Natural Forest Conservation Program (NFCP) Grain-to-Green Program (GTGP) Grain-to-Bamboo Program Beginning date 2001 2000 2002 Duration 10 years (renewed in 2010 for another 10 years till 2020) 8 years (renewed in 2008 for another 8 years) 8 years (ended in 2010) Legal mechanism National conservation policy National conservation policy Local conservation policy complementing the GTGP Targeted area Forest parcels near household locations along the main road Cropland with slopes larger than 25 degree, which are around household locations Cropland within the 15 meters buffer zone at each side of the main road Targeted households 1130 rural households ƒ All 1200 rural households All 1200 rural households Implemented area All land below treeline (i.e., 120,500 ha and 3600 m above sea level), of which 40,100 ha were assigned for household monitoring 367.3 ha from 969 households 81.9 ha from ~530 households Average payment rate* ~143 USD per household per year ~571 USD per ha per year 2143 to 2857 USD per ha per year Source: Wolong Administrative Bureau; * household interviews, 1 USD = 6.3 Chinese yuan as of May 2012. ƒ About 70 households were excluded because th eir household heads or their spouses have non- rural household registration and thus they were not regarded as local rural households. 148 Table 7.2. Paired t-test for indicators of energy use before and after the NFCP implementation. Pre-NFCP (Mean ± S.D.) Post-NFCP (Mean ± S.D.) Paired t test (two-tailed) Electricity consumption amount (kilowatt- hour, N = 169) 1165±1301 2562±1906 -9.57*** The amount of labor force for fuelwood collection per year (laborer days, N = 151) 64±30 33±29 10.02 *** Duration of cooking pig fodder with fuelwood (months, N = 149) 11.8±1.4 10.2±3.4 6.09*** Energy source for heating (N = 148) 4.9±0.4 3.7±1.6 8.70*** Energy source for cooking human food (N = 149) 4.5±0.9 2.3±1.4 19.10 *** Notes: *P < 0.05, **P < 0.01, *** P < 0.001. Pre-NFCP and post-NFCP refer to the year 1998 (i.e., reference year before NFCP) and 2001 (i.e., the first year after NFCP implementation), respectively. Energy source was classified into a 5-score scal e on percentage of fuelwood use, with 1 for no use of fuelwood, 2 for 25% of fuelwood, 3 for 50% of fuelwood, 4 for 75% of fuelwood, and 5 of 100% use of fuelwood, respectively. 149 Figure 7.1. Map of the Wolong Nature Reserve. 150 Figure 7.2. Natural Forest Conservation Program (NFCP) implementation in the Wolong Nature Reserve (WNR). 151 Figure 7.3. Forest cover area before and after the Natural Forest Conservation Program (NFCP) implementation in 2001. Notes: The historical forest cover trend from 1965 to 2001 was used to estimate the counter- factual without PES baseline in 2007. Additiona lity is the difference between the with-PES outcome and the without-PES baseline. 152 Figure 7.4. Perceptions of local households toward the Natural Forest Conservation Program (NFCP). Notes: a-d represent households™ perceptions toward NFCP benefits (multiple response question), NFCP disadvantages (multiple response question), wildlife induced losses before and after NFCP implementation (single response que stion), and administration of w ildlife induced losses (single response question), respectively. The total number of observations for each figure panel was 183. 153 Fig. 7.4 (cont™d) 154 CHAPTER 8 DYNAMICS OF HOUSEHOLD CAPITAL IN PROTECTED AREAS DURING POST- DISASTER RECONSTRUCTION In collaboration with Zhiyun Ouyang, Thomas Dietz, Wei Liu, Jianguo Liu 155 Abstract Post-disaster reconstruction is often assumed to recover natural, human, manufactured, and social capitals. However, due to the lack of pre-disaster data, there have been relatively few systematic evaluations of this proposition. Using long-term household survey data from Wolong Nature Reserve (an area of global importance in China) and surrounding areas, our results suggest that the effects of post-di saster reconstruction efforts can differ from one scale to another. After five years™ reconstruction, the overall capital at the reserve level is improving with the huge amount of reconstruction inve stment, but there was an overall deterioration of household capital both inside and outside the Reserve. Such deterioration is partly due to associated disasters after the earthquake and partly due to inappropriate management. Our findings suggest that capacity building and recovery should not be taken for granted in post-disaster reconstruction and require integrated planning an d implementation targeting each form of capital at multiple scales. 156 8.1 Introduction Many conservation policies have been impl emented throughout the globe to mitigate escalating ecosystem degradation. However, failur es are common and are often due to a lack of understanding of complex human-nat ure interactions (Ostrom et al., 2007; Tallis et al., 2008). Unfortunately, natural disasters only exacerbate such problems. It is projected that the number and scale of disasters will increase because of global environmental change and the increase of human population and human activ ities in areas at risk of disasters (IPCC, 2012). Most biodiversity hotspots are also areas with frequent natural disa sters such as earthquakes and volcanoes (Myers et al., 2000; Willis et al., 2007 ), so the problem is especially acute around efforts to protect biodiversity. Previous literature emphasizes the importa nce of diversity, cross-scale interactions, incremental and transformational adaptations, and different forms of capital in both how a region is affected by a disaster and in the success of post-disaster recovery (Gunderson, 2010; Kates et al., 2012). For example, Tilman et al. (2001) sh owed that high biodiversity helped the post-disturbance recovery of ecosystem functions. Berke and Campanella (2006) observed how a diverse economy facilitated the post-disturbance recovery of human communities. Nystrom and Folke (2001) showed how processes that cross spatial scales affected coral reef recovery after hurricanes, while Adger et al. (2005) demonstrated how aid from st ate, federal and international organizations helped local human communities recove r from disasters. A number of studies also have demonstrated how different flood prevention strategies, insurance programs and regulatory policies have helped some communities to recover from floods while making some other communities more vulnerable (Houck, 1985; Kl ein and Zellmer, 2007). Some incremental adaptations (e.g., building levees along a river) are adaptive in the short run but may turn out to 157 be maladaptive in the long run, thus requiring tr ansformational adaptations such as evacuation and relocation of human settle ments (Kates et al., 2012). The four forms of capital typically defined in discussions of sustainable development Š natural, human, manufactured, and social capital, are each important in disaster response and recovery. Their roles in helping human communitie s to build adaptive capacity, prevent disasters, reduce losses to disasters, and facilitate post-disaster recovery have been well documented in studies from floods, droughts, Tsunamis, hurrica nes and earthquakes (Berke and Campanella, 2006; Ekin, 2008; Gunderson, 2010; Lowe et al., 2010; Masten and Obradovic, 2008; Shaw and Goda, 2004). However, few studies have examined how natural disasters have affected the four forms of capital because without pre-disaster data, and it has been impossible to assess how different forms of capital are aff ected by both the disaster itself and by subsequent reconstruction efforts. There are a variety of proposed framew orks for vulnerability research (Ekin, 2008; Escobar, 1999; McLaughlin, 2011; Smit and Wande l, 2006; Turner et al., 2003), ranging across biophysical, human ecological, political economy, constructiv ist, and political ecology perspectives. But none of them provides an integrated framework due to the lack of understanding of the mutually causal linkage s among different forms of capital, adaptive capacity, resilience, and vulnerability (Gallopin, 2006; McLaughlin, 2011; McLaughlin and Dietz, 2008). Empirical evidence is needed to su pport the development of integrated frameworks and theories for vulnerability research. A long-term study in a globally important area affected by the Wenchuan Earthquake provided an opportunity to fill some of these knowledge gaps in understanding the effects of disasters and disaster recovery efforts. The earthquake (Ms 8.0 or Mw 7.9), which occurred on May 12, 2008, was the most devastating earthquake in China since 1950s. In total the earthquake 158 killed 69,227 people, injured 374,643, resettled 1.5 million, and led to 17,923 missing persons (Xinhua News Agency, 2008a). The earthquake also severely damaged ecosystem in the region, including critical panda habitat, as well as destroying manufactured infrastr ucture such as roads, houses, schools, hospitals, and tourism facilities. It is estimated that th e direct and indirect economic losses was more than one trillion yua n (1 USD = 6.2 yuan as of 2013) (Xinhua News Agency, 2008b). After the earthquake, the State Council of China developed the State Overall Planning for Post-Wenchuan Earthquake Restorati on and Reconstruction, an effort to restore conditions in the quake-hit areas to a level no worse than they were before the earthquake (SPGPWERR, 2008). This effort was initially designed to last for three years but later was shortened to two years. Since 1995 our research team has been conducting a long-term Coupled Human and Natural Systems project at Wolong Nature Rese rve (WNR), Wenchuan County in southwestern China (An et al., 2001; An et al., 2003; Bearer et al., 2008; Chen et al., 2010; He et al., 2008; Linderman et al., 2005a; Liu et al., 1999a; Viña et al., 2008; Yang et al., 2013c; Yang et al., 2013f). The focus area of our research is thus near the epicenter of the Wenchuan Earthquake. Our previous studies have evaluated how and to what extent the earthquake had affected subjective well-being of local households (Yang et al., 2013a). At the household level, we confirmed that high access to natural, human, ma nufactured, and/or social capitals before the earthquake had helped households to prevent or reduce the imp acts on their well-being (Yang et al., 2013b). Here we systematical ly evaluate how natural, human, manufactured, and social capitals of local households were affected by post-disaster restora tion and reconstruction efforts. Based on our results and extensive local knowledge, we further discuss the causes and effects of observed changes in household capital. We then provide recommendations for policy revision 159 and design to improve the environmental a nd socioeconomic outcomes of post-disaster reconstruction and achieve long-term sustainability in our study area and beyond. 8.2 Materials and methods 8.2.1 Study area We chose the WNR and the adjacent Sanjiang Township (SJT) as our study sites (Fig. 8.1). Both sites belong to the Sichuan Giant Panda Sanctuaries of the UNESCO World Heritage system. As a flagship reserve both in Chin a and globally, WNR is of special ecological importance because it is the home to the larges t wild population of giant pandas (~10%) and over 6000 other plant and animal speci es (Schaller et al., 1985). Both WNR and SJT are located in Wenc huan County, Sichuan Province, China. WNR was established in 1963 and expanded in 1975 to its current size of ~2000 km2 with ~4900 local residents distributed in 1200 households. SJT has a size of ~490 km2, of which 344 km2 incorporated into the Reserve in 1975. There are currently ~4000 local residents in SJT distributed in 1100 households outside the Reserve (Fig. 8.1). Before the implementation of the two major national conservation policies in early 2000 (Liu et al., 2008), both sites had economies based on subsistence agriculture. Most local residents are farmers who grow maize and vegetables, raise livestock, keep bees, and collect traditional Chinese medicinal plants, mushrooms, bamboo shoots and fulewood (Y ang et al., 2013c; Yang et al., 2013e). Both at WNR and SJT, our interviews with local government officials who were in charge of post-disaster reconstruction suggest that local governments have given top strategic 160 priority to tourism development and hope it will rapidly recover and thus dramatically improve local socioeconomic conditions. However, o fficials in WNR emphasize luxury ecotourism development while SJT focuses on rural tourism. Both local governments give priority to and compensate large-scale farms (e.g., pig, chicke n and cattle farms). At SJT, post-disaster reconstruction was generally completed by May 2010 in accord with the state planning goal. The infrastructure and houses damaged by the earthquak e have been repaired or rebuilt. However, the reconstruction process at WNR moves slowly and is not planned to be completed until 2015. This provides a basis for assessi ng the effects of recovery efforts on the four forms of capital. 8.2.2 Data collection and analyses Since 1999, our research team has been conducting household surveys to obtain demographic, socioeconomic, and policy impl ementation data for the WNR (An et al., 2001). The first household survey in 1999 interviewed 220 households using random sampling with strata based on administrative groups (an administra tive unit smaller than the village). Just before the earthquake, at the end of 2007 and in January of 2008, we revisited the 220 households and successfully surveyed 192 of them. In 2010 su mmer, we expanded the random sample from 220 households to 305 and surveyed 287 households, 169 of which were selected for an additional human well-being survey. Data on housing conditions (i.e., house type and total constructed area) and the long-term trend (1970s Œ 2010s) of total household income are from this human well- being survey (Yang et al., 2013a). Previous basi c household survey data and instruments (1999 Œ 2010) are available at Interuniversity Consor tium for Political and Social Research (ICPSR; http://dx.doi.org/10.3886/ICPSR34365.v1). In 2012 summer, we revisited 282 of the previously sampled households. Besides the basic household su rvey questions, the 2012 survey instrument 161 added questions on cropland area, household loans for post-disaster housing repair and reconstruction, and social support. The instru ments used in 2010 and 2012 WNR surveys were repeated at SJT with 157 randomly sampled house holds. Overall, 159 households at WNR were always sampled across 2007, 2010 and 2012 and 142 households at SJT were sampled both in 2010 and 2012. Our definitions and indicators of natural, human, manufactured, and social capitals (see details in Table 8.1) are consistent with previous literature given c onstraints of data availability (Berke and Campanella, 2006; Ekin, 2008; G underson, 2010; Shaw and Goda, 2004; Wellman and Frank, 2001; Wellman and Wortley, 1990). All monetary values used for analysis were discounted into the constant values of 2010. We pe rformed all the statistical analyses using the software STATA 12.0 (StataCorp LP, Texas, USA). 8.3 Results For natural capital, inside the Reserve, about 12% of cropland became unusable due to the earthquake and associated disasters (e.g., landslides, mud-rock flows and mountain torrents) after the earthquake, and about 24% to post-disaster reconstruction (Table 8.2). Outside the Reserve, earthquake and associated disasters caused losses of cropland were about 20% and reconstruction caused losses were about 10% (Table 8.2). These numbers make it clear that the reconstruction effort is leading to substantial cropland loss, especi ally inside the Reserve. Figure 2 shows changes in human capital. Th ere were significant increases of household expenditures in food, education, and medical care inside the Re serve after the earthquake. In 162 particular, the mean and median expenditure for medical care have increased by 3.5 and 16.5 times from 2007 to 2011, respectively. For manufactured capital, our field work suggests that the road condition in early 2013 was as bad as it was in 2009 due to massive post- disaster mountain torrents, mud-rock flows, and landslides. Figure 3 and Figure 4a show that th e housing quality and total constructed area per household both inside and outside the Reserve have continuously increased since 1970s and this continued after the earthquake. However, such improvements were slower inside the Reserve compared to those outside the Reserve. Bo th inside and outside the Reserve, housing improvements after the earthquake were achieve d through a very substantial amount of loans (Fig. 8.4b). There are many definitions of social capital , but a consensus emerged in the late 1990s and early 2000s defining social capital in terms of the potential to access resources through social relations (Lin, 2001; Portes, 1998). Here we followed this definition and measured social capital as the maximum amount of social support can be obtained in monetary values (Wellman & Wortley 1990, Wellman & Frank 2001). Our results show that the timing, total social support, and social support per lender differed from inside to outside the Reserve (Fig. 8.5). On average, the total amount of social support and social support per lender inside the Reserve dramatically decreased from 2007 to 2009 and increased from 2009 to 2012, while the opposite was the case outside the Reserve. The differe nt temporal patterns of social support were probably because households inside the Reserve used their social capital more intensively to acquire more loans and a higher percentage of them from relatives and friends between 2009 and 2011 (Fig. 8.4b) than did those outside the Reserve. For households inside the Reserve, more than 60% of the amount of social support came from other househol ds inside the Reserve. But for households 163 outside the Reserve, approximately half of the amount of their social support came from other households in the same township but the other half was from others outside their township (Fig. 8.6). 8.4 Discussion While the State Overall Planning (SPGPWER R, 2008) effort aimed for basic living and socioeconomic conditions to at least equal to th ose before the earthquake within two years, the actual situation was quite different. Overall, with the huge investment (2 thousand million yuan for the entire reconstruction plan inside the reserve), new infrastructure (e.g., road, tourism facilities, hospitals, and schools) and human settlements have been built, and vegetation recovery projects have been implemented. The overall cap ital at the reserve level is improving. But our results suggest that there was an overall deterioration of household capital both inside and outside the Reserve. We observed some impr ovements of housing conditions. But this was largely due to loans undertaken by households. There was a slow recovery of social capital inside the Reserve. But there were large shrinkage of natural capital and human capital inside the Reserve. The observed differences between inside and outside the Reserve might be partly due to the massive and frequent associated disasters afte r the earthquake and other administrative issues (e.g., delay of funding and the setting of the Reserve) inside the Reserve. Below we suggest other causes for the observed changes in the four forms of household ca pital, discuss socioeconomic and environmental impacts, and provide policy recommendations. 164 8.4.1 Causes and effects for observed changes in household capital Whereas ~60% of cropland inside the Reserve had been converted into forests before the earthquake (Yang et al., 2013e), the earthquake, associated disasters after the earthquake and especially post-disaster reconstruction have further destroyed ~36% of the remaining cropland. Most of the occupied cropland during post-disaster reconstruction was used to build the new breeding and reintroduction center for giant pandas, new human settlements, and tourism facilities. On the surface, this may mitigate human disturbance on forests and giant panda habitat by reducing agricultural activities. But for a subsistence-ba sed economy, agricultu ral income is the largest part of total hous ehold income. If there are no a lternative income sources to compensate the reduction of agricultural income , the livelihood of local households would be substantially adversely affected. In some cases they may be forced to conduct illegal activities (e.g., poaching and logging) to maintain their livelihood (Yang et al., 2013a). Our data show that before the earthquake households inside the Reserve had significantly higher gross income than those outside the Rese rve (Fig. 8.7). But after the earthquake, there was a significant decline of gross household income inside the Reserve which contrasts with a significant increase outside the Reserve. Meanwhile, the dramatic increases in expenditures on food, education and medical care do suggest stresses in generating human capital. This is in contrast to other studies that have found post-recovery increased th e living standards or education quality in other (Ekin, 2008; Evenson and Mwabu, 1996; Ranis et al., 2000). Rather, these results indicate a decline of human capital in our case. Households need to buy more food due to the loss of cropland. They spend more money to maintain their children™s educational opportunities as transportation fees have increased and many students are sent to boarding schools outside the Reserve due to losses of good teachers and overlook of education quality 165 inside the Reserve. After the earthquake, they pa y more to go outside for medical care due to the declined local medical facilities as well as phys ical and mental health problems generated by the earthquake (Yang et al., 2013a). All of these factors place stresses in generating human capital. In addition, it should be noted that many households received a relatively large amount of government compensation from cropland acqui sition (874,500 yuan per ha permanently or 54,000 yuan per ha every year), which allows them to have extra money to lend to other households. Without the government compensati on, the magnitude of social support would probably be much lower. But such post-disaster government compensation was not likely to lead to the observed difference in terms of sources for social support (Fig. 8.6). Both before and after the earthquake, those living inside the Reserve were more likely to obtain loans locally than was the case for those in SJT. The possible reason, ba sed on our investigation and local knowledge, is that social networks of households inside the Reserve are denser than those outside the Reserve. While it is difficult to accurately estimate th e poaching and illegal logging activities, our field investigations before and after the earthquake suggested an increase in poaching inside the Reserve and an increase in both poaching and illegal logging outside the Reserve (Yang et al., 2013a). This view was shared by the forest patrol who witnessed an increase in animal traps and in signs of illegal logging. The deterioration of household capital may not only lead to negative socioeconomic and environmental impacts discussed above, but al so reduce the adaptive capacity and increase the vulnerability of local households for future disasters as suggested in many previous studies (Berke and Campanella, 2006; Gunderson, 2010; Houck, 1985; Kates et al., 2012; Klein and Zellmer, 2007; Lowe et al., 2010; Masten and Obra dovic, 2008). The increased vulnerability in 166 turn can further aggravate the deterioration of household capital and form positive feedback loops, which would pose serious threats to long-term local livelihoods and biodiversity conservation. 8.4.2 Policy recommendations To improve the environmental and socioeconomic outcomes of post-disaster reconstruction and achieve long-term sustainabi lity, we offer the following recommendations. First, many critical habitats are remote a nd local community depends on transportation infrastructure to facilitate ecotourism or trade of agricultural products that provide critical income. If natural disasters damage this infras tructure it can have a very adverse effect on income and lead to spillover effects such as poaching and illegal logging. Thus a high priority must be given to rebuilding transportation in frastructure and maintaining accessibility post- disaster. For example, in WNR, since there is on ly one main road crossing the Reserve, it is essential to maintain the main road to allow ti mely response to associated disasters after the earthquake. It is particularly important that road accessibility is guaranteed in some crucial periods such as tourism peak s easons and agricultural harvest seasons. But the improvement of transportation infrastructure should be restri cted to the main road and human settlements concentrated areas limiting the impacts on wildlife habitat. It may also be necessary to communicate to potential tourists and to those buying agricultural products from a disaster impacted area that things are returning to normal. Second, besides the rebuilding and maintenan ce of transportation infrastructure, more attention should be paid to encouraging small-scale tourism bus inesses, which provide more 167 benefits to local households than luxury ecotourism does. While the local government prioritizes and expects that luxury ecotourism developmen t would improve local socioeconomic conditions, evidence from our previous studies in this area (He et al., 2008; Liu et al., 2012) and other studies around the world (Kiss, 2004) suggest that the large proportion of luxury ecotourism revenue often leaks into tourism developmen t companies and the governments, while local communities receive a very small portion of benef its (<4% at WNR). The distribution of revenue from luxury ecotourism needs to be adjusted to provide more benefits to local households (e.g., through the provision of subsidies a nd job opportunities preferably to local residents). In addition, policy revisions could enhance the capacities of local households to participate in luxury ecotourism businesses through the pr ovision of trainings and loans. Our previous study indicates that only the very few households with better education, higher social capital and more substantial financial resources were able to part icipate in luxury ecotourism businesses (Liu et al., 2012). Third, it is necessary to r ecover and maintain productive cr opland to avoid stresses on and enhance stability of household income. In WNR, some flat la nds near the main road with high agricultural productivity might be reconvert ed to cropland. These lands are not suitable panda habitat. They are below the elevation range of panda habitat, lack the bamboo species favored by pandas, and have intense human dist urbances that pandas avoid. The unpredictable nature of tourism emphasizes the importance of agriculture. Tourism is an unstable and often vulnerable industry sector full of uncertainties compared to the relative st ability of subsistence agriculture (Wahab and Pigram, 1997). For ex ample, WNR is in a mountainous area with frequent natural disasters such as landslides even in normal years. Such disasters are common during one of the annual peak tourism seasons (i n May, because of rains). This situation became 168 even worse after the earthquake with more frequent associated di sasters after the earthquake. So it is prudent to redevelop agriculture to provide an economic base that has diverse sources of income for local people. Fourthly, some other measures should be taken to expand income sources, reduce expenditures, compensate disadvantaged househ olds, improve social relationships among households, and build mutual trust between hous eholds and local governments. For instance, large-scale farms inside the Reserve may not only cause pollution but also will not be as competitive large scale farming outside the Reserv e due to the uncertain road accessibility and higher transportation costs. Therefore, local governments should prioritize high value local or eco-certified products. Products from the Reserve could be highly competitive in this market niche because of its high environmental qua lity. Policy should also compensate those disadvantaged households who suffer most from the earthquake, associated disasters after the earthquake, and post-disaster r econstruction (Yang et al., 2013a). Useful measures include providing compensation for their food, education and health care expenditures and reducing or waiving their loans or interests from government controlled banks. In addition, since millions of yuan has been spent for the post-disaster reconstruction, it is also crucial to audit how that money has been spent, which would help build mutu al trust between households and governments and improve the performance of reconstr uction efforts in the long run. Finally, policy revisions should enhance th e monitoring of and sanctions on poaching and illegal logging. However, it should be noted that there is no single solution to improve the policy performance due to the complex human-nature intera ctions (Liu et al., 2007a). For example, only enhancing the enforcement for preventing poaching and illegal logging without improving economic conditions may aggravate the deterior ation of household capacities and cause more 169 conflicts between households and local governments. In turn, such deterioration may form positive feedback loops such as the biodiversit y conservation and poverty traps that occur in other places (Barrett et al., 2011). This could lead to negative socioeconomic and environmental outcomes. Thus, to improve the policy performan ce in the long run, we emphasize an integrated approach adopting multiple measures as mentioned above simultaneously. 8.5 Conclusion To confront the escalating threat of natural disasters on human communities and biodiversity conservation, it is crucial not only to build capacity to prevent or reduce disaster impacts but also to facilitate ra pid recovery of capacity post-disaster. Using long-term household survey data from Wolong Nature Reserve and su rrounding areas, we find that the effects of post-disaster reconstruction efforts can differ from one scale to another. After five years™ reconstruction, the overall capital at the reserve level is improving with the huge amount of reconstruction investment, but ther e was an overall deteriorating trend of household capital in the flagship giant panda reserve. Such deterioration of household capital is partly due to associated disasters after the earthquake, administrative is sues (e.g., the delay of funding and the setting of the Reserve), and partly due to inappropriate or inadequate policy design and implementation (i.e., lack of attention to improving each form of household ca pital).We discussed the negative socioeconomic and environmental impacts of such deterioration and warned of the risks of the reduction of adaptive capacity, incr ease of the vulnerability of local households to future disasters, and the formation of vicious hum an-nature feedback loops (e.g., biodiversity conservation and poverty loops). Our findings suggest that capacity building and recovery should not be taken for granted in post-disaster reconstruction and require integrated planning and 170 implementation targeting each form of capital. The experience and lessons learned from our study may help to revise the current policy and design future policies not only in our study area but also other disaster affected areas in China (e.g., the recent 2013 Ya™an Earthquake) and around the world. 171 Table 8.1. The definition and selected i ndicators of four forms of capital. Definition Selected indicators Natural capital The stock of natural or human managed ecosystems that yields a flow of ecosystem goods and services for human well-being. Cropland area Human capital The stock of competencies, knowledge, social psychological attributes of individual people, such as mental and physical health, education, motivation and work skills. Food expenditure * Educational expenditure * Health expenditure* Manufactured capital The stock of human-made assets that used to produce other goods and services, such as infrastructure, buildings , machines, and tools. House type ƒ Total constructed area ƒ Social capital The stock of so cial norms and networks that support social and economic interactions among its individual and organizational members. Total social support ⁄ Social support per lender ⁄ Note: All the indicators are measured at the household level. * In normal situations, increases of these indicators indicate improvements of human capital; however, in our context of disasters-induced increases of these indicators suggest extra burdens for enhancing human capital and thus impl y deteriorations of human capital. ƒ If a household has more than two houses, th e average score of house type was used and rounded to the nearest integer. Total constructed area calculated the constructed area of all the houses a household owned. ⁄ The question used to assess social support (Wellman & Wortley 1990, Wellman & Frank 2001) is: Suppose you want to build or buy a new house or start a business, who will lend you money and how much is the MAXIMUM amount? (List th e five most possible households: the location and name of the household, relationship between you two, and the maximum amount of money)fl. 172 Table 8.2. Descriptive statistics of cropland per household inside and outside the Reserve. Inside reserve (N = 282) Outside reserve (N = 142) Cropland area before the earthquake in 2007 (mu) 4.604 (3.457) 4.502 (3.701) Cropland losses due to the earthquake and associated disasters after the earthquake (mu) 0.551 (1.096) 0.918 (1.420) Cropland losses due to post-disaster reconstruction (mu) 1.110 (2.215) 0.455 (0.892) Cropland area after the earthquake in 2012 (mu) 2.943 (3.255) 3.129 (3.001) Note: The numbers outside and inside the parenthe ses are mean values and standard deviations, respectively. mu is a Chines e area unit, 1 mu = 1/15 ha. 173 Figure 8.1. Map of Wolong Nature Reserve and the adjacent Sanjiang Township. 174 Figure 8.2. Box plots of household expenditures on food (a), education (b), and medical care (c) before and after the earthquake at Wolong Nature Reserve. Notes: The same 159 households were tracked over years. Paired-t tests were used to compare the difference of mean values before and afte r the earthquake. NS: non-significant (p > 0.05); ***: highly significant (p < 0.001). 175 Figure 8.2 (cont™d) 176 Figure 8.3. Dynamics of the house type inside (a ) and outside the Reserve (b) from 1970s to 2010s. Notes: The numbers of observations are 169 and 157 inside and outside the Reserve, respectively. The numbers from 0 to 5 represent the quality of house from the lowest to the highest, respectively (i.e., 0: no house or rented house; 1: stone house; 2: stone and wood mixed house; 3: wooden house; 4: wood and concrete mixed house; 5: concrete house). 177 Figure 8.4. Dynamics of the total constructed area per household (a) and loans per household for post-disaster housing repair or reconstruction (b). Notes: a) The error bar represents standard error. The numbers of observations are 169 and 157 inside and outside the Reserve, respectively; b) The numbers of observations are 282 and 142 inside and outside the Reserve, respectively. 178 Figure 8.5. Total amount of social support and social support per lender before and after the earthquake both inside and outside the Reserve. Notes: The error bar represents standard error. The numbers of observations are 282 and 142 inside and outside the Reserve, respectively. 179 Figure 8.6. Proportions of external and internal sources of social support for households before and after the earthquake both inside and outside the Reserve. Note: The numbers of observations are 282 and 142 inside and outside the Reserve, respectively. 180 Figure 8.7. Dynamics of the gross household inco me inside and outside the Reserve from 1970s to 2010s. Notes: The numbers of observations are 169 and 157 inside and outside the Reserve, respectively. All monetary values were discounted into the constant values of 2010. 181 CHAPTER 9 CONCLUSIONS 182 The completion of this dissertation not only advances scientific understanding but also provides information I hope will be useful fo r real-world management of human-nature interactions in a variety of ways. Below I summarized the main advances in methodology and scientific knowledge, and implications for decision making. From a methodological perspective, the di ssertation provided new methods to quantify human dependence on ecosystem services (ES; Ch apter 2), human well-being (HWB; Chapter 3), and their linkages (Chapter 4). Using long-term datasets (e.g., survey data, satellite imagery, field plots, government documents) and various m odels (e.g., Tobit model, spatial autoregressive model, structural equation mode l), it offered strong evidence about the group-size effect that has been debated for many decades (Chapter 5). Fo llowing Chapter 5, the multiple datasets and mixed methods were also applied to address research questions in Chapters 6, 7 and 8. Such integration of multiple datasets and mixed me thods provide an excellent example illustrating why and how to conduct interdisciplinary studies to address complex problems in Coupled Human and Natural Systems (CHANS). The quantitative and integrated approach to understand the linkages between ES and HWB (Chapters 2, 3 and 4) suggests that the construction of quantitative indicators for ES and HWB as well as integrated models is a viable approach for enhancing the understanding of linkages between ES and HWB. The findings show that disadvantaged groups who possess low levels of any form of capital except for natu ral capital are more dependent on ES than the affluent that have greater control of capital. The findings also suggest that external drivers may significantly affect both human dependence on ES and HWB with influence varies across different components of ES or HWB. Those who are more vulnerable to disasters are disadvantaged households with lowe r access to multiple forms of capital, more property damages, 183 or larger revenue reductions. Diversifying human dependence on ES helps to alleviate disaster impacts on HWB. Since the approach is based on the generalizable Millennium Ecosystem Assessment (MA) framework, it can be applied to di fferent scales from local, regional to global and different units of analysis (e.g., individual, household, enterprise, and nation). The substantiation of nonlinear group size effects on collective action and resource outcomes as well as the underlying mechanisms in Wolong Nature Reserve helps to guide future research and governance of the commons (Chapter 5). The results show that intermediate group size contributes the most to collective action a nd leads to the optimum resource management outcomes. Such findings thus suggest that alteri ng factors (e.g., punishing free riding, enhancing enforcement, improving social capital, and allowing self-selection for group formation) that lead to the nonlinear group-size effect can improve collective actio n and resource outcomes. The findings also explain why previ ous studies have observed contradictory group-size effects by focusing on a portion of the nonlinear curve and why a holistic approach should be taken to address complex problems not only in common-pool resource management but also in all other CHANS research. Inspired by the findings, methods, and resear ch process of the nonlinear group-size effect, Chapter 6 examined different forms of hypothetic al interaction effects of policies, which generated some additional important findings. The results suggest that there are synergistic and antagonistic effects among conser vation and/or development policies, which can even lead to unanticipated consequences. In our case, the payments from two conservation policies separately had a negative effect on household income, but jo intly led to a positive effect. Such findings challenge the traditional realm of policy studies that often treat multiple policies separately and 184 again advocate a holistic approach by considering policy interactions, jointly designing, monitoring, and evaluating policies. Learned from previous chapters, Chapter 7 evaluated the effects of the Natural Forest Conservation Program (NFCP) on both ES (using forest cover as an indicator) and HWB (using income, indicators of energy use, and attitude me trics), with consideration of the interactions between NFCP and other concurrent policies (e.g., Grain-to-Green Program, Tourism Development Program, and Electricity Subsidy Progr am). The results suggest that NFCP had an overall positive effect on ES, and mixed effects on local hou seholds™ well-being. Based on the results, field observations, and local knowledge, the underlying mech anisms of the effects on ES and HWB were discussed, which provide insi ghts (e.g., integrating local conditions and underlying mechanisms) for improving the design and implementation of policies not only in Wolong Nature Reserve but also other places around the world. Given the human-nature interactions, previous chapters have illustrated how complex the effects and underlying mechanisms of policies on both ES and HWB could be. Chapter 8 further confirmed this point and argued why and how scie ntific knowledge should be seriously taken into policy design and implementation so as to achieve positive outcomes both for ES and HWB. Chapter 4 confirmed that high access to natural, human, manufactured, and/or social capitals before the earthquake had helped households to prevent or mitigate negative impacts. As a further step, Chapter 8 evaluated how post-disaster reconstruction efforts affected the four forms of capital, which often are assumed to be positiv e. Unfortunately, the results actually show the opposite with an overall deteriorating trend of household capital after five years™ reconstruction, despite two billion yuan (1 US dollar = 6.2 yuan as of July 2013) had been investing for reconstruction. While the overall capital at the reserve level seems to be improving with the huge 185 amount of reconstruction investment, the findings suggest that the effects of post-disaster reconstruction efforts can differ from one scale to another. Therefore, capacity building and recovery should not be taken for granted and require integrated planning and implementation targeting each form of capital at multiple scales. It should also be noted that much of my wo rk in this dissertation benefits from the long- term effort to collect multiple kinds of data in WNR and surrounding areas. The lack of long- term high quality data perhaps is the biggest pr oblem in environmental science research and in our ability to manage CHANS. While this dissertation made a number of contributions to methodology and scientific knowledge and provides a number of policy implications as mentioned above, it also opens the door for further scientific inquiries. For instance, further research can apply our quantitative and integrated approach in other studies both in Wolong Nature Reserv e and other areas in China and the rest of the world to test interesting hypotheses, answer important questions, and addressing pressing problems for sustainabi lity (see example hypotheses and que stions in Table 4.2). Future research can also follow our multiple datasets and mixed methods to investigate what contextual factors determine the optimum group size either fo r the provision of collective action or resource outcomes. The accumulated knowledge from a set of literature will hopefully lead to coherent theories (e.g., human-nature feedbacks theory, vulnerability/resilience theory of CHANS) to guide the management of human-nature interactions. 186 APPENDICES 187 APPENDIX A SUPPORTING INFORMATION FOR CHAPTER 2 Table S2.1. Detailed classification of household net income and a voided costs by type of related ecosystem services. Category Sub-category Item Type of related ecosystem services* Operating Income Crop income INC101: Cabbage P0 INC102: Radish P0 INC103: Potato P0 INC104: Corn P0 INC105: Other crops P0 Animal husbandry income INC106: Bacon P1 INC107: Pig P0 INC108: Goat P0 INC109: Cattle P0 INC110: Yak P0 INC111: Horse P0 INC112: Poultry and eggs P0 INC113: Honey bee P0 INC114: Other husbandry P0 NTFPs income INC115: Non-timber Forest Products (NTFPs) P0 Other agricultural operating income INC116: Other agricultural operating income P0 Non-agricultural operating income INC117: Restaurants and hotels C1 or NAƒ INC118: Ecotourism C1 or NAƒ INC119: Transportation C1 or NAƒ INC120: Contract work NA INC121: Other small businesses C1 or NAƒ Wage Income INC201: Wage and bonus NA INC202: Local labor income NA INC203: Migrant labor income NA 188 Table S2.1 (cont™d) Category Sub-category Item Type of related ecosystem services* Property Income Land and housing rents INC301: Land and housing rents C1 or NAƒ Other property income INC302: Interest income NA INC303: Land acquisition compensation NA INC304: Other rents NA Transfer Income Gift income from relatives and friends INC401: Gift income from relatives and friends NA Payments for ecosystem services (PES) income INC402: Natural forest conservation program (NFCP) R0 INC403: Grain-to-Green program (GTGP) R0 INC404: Grain-to-Bamboo program (GTBP) R0 Social security Benefits INC405: Low income subsidy NA INC406: Pension NA INC407: Other subsidies NA Other Income INC501: Remaining other socioeconomic income NA Avoided costs Fuelwood for energy use P0 Subsidized electricity fees due to watershed conservation R1 Notes: *: Letters P, R, C, and NA represent provisioning services, regulating services, cultural services, and benefits unrelated to ecosystem services respectively. The digits fi0fl and fi1fl after fiP, R or Cfl represent direct and first-order indirect ecosystem services, respectively. In our case, PES programs were designed mainly for regulat ing services (e.g., water conservation, soil erosion control, carbon sequestration, and air purification). Thus, we classified PES as benefits related to regulating services. Dis-services and costs of delivering ecosystem services have already been included in the net income data, which were best available approximations of net benefits here. The data and detailed descripti on of each variable are provided in the Supporting Information file. ƒ: For each household, if the benefit is related to ecotourism, it is included as a benefit related to cultural services; or else, it is regarded as a benefit unrelated to ecosystem services. 189 APPENDIX B SUPPORTING INFORMATION FOR CHAPTER 3 Table S3.1. The index system for assessing human well-being based on the Millennium Ecosystem Assessment conceptual framework. Indicator layer Indicator code and content ri-t before earthquake ri-t after earthquake Affordability to necessary food Q1.2: Your household can afford enough food with nutrition to keep alive and healthy 0.64 0.66 Affordability to basic facilities and services Q1.3: Your household can afford to access basic facilities (e.g., television, washer) and services (e.g., transportation) 0.62 0.64 Satisfaction with housing condition Q1.4: You are satisfied with your housing condition (including size and quality) 0.50 0.40 Overall satisfaction with access to basic goods and services Q1.5: Overall, you are satisfied with your household™s basic goods and services (e.g., food, clothe, living conditions, transportation) for life 0.68 0.64 Life safety Q2.1: Your household™s life safety in daily life is secure 0.59 0.46 Property safety Q2.2: Your household™s property safety in daily life is secure 0.54 0.49 Local crime incidence Q2.3: The local crime in cidence (e.g., theft, robbery, murder, other violent incidents) is low 0.51 0.41 Access to government protection Q2.4: The police and judicial system is always ready to help 0.40 0.38 Reliability of government protection Q2.5: The police and judicial system can be trusted 0.38 0.33 Security for resource access Q2.6: It is safe to access basic goods and services such as food, water, and medicine etc. for life 0.56 0.49 Overall satisfaction with security Q2.7: Overall, you are satisfied with your household security (e.g., life and property) 0.66 0.59 190 Table S3.1. (Cont™d) Indicator layer Indicator code and content ri-t before earthquake ri-t after earthquake Physical health Q3.1: You are satisfied with your household™s physical health (including illness and injury)? 0.67 0.66 Mental health Q3.2: You are satisfied with your household™s mental health (includi ng stress, depression, and problems with emotions)? 0.66 0.68 Rest Q3.3: How often your household members do not get enough rest or sleep? (Options: 1. Always; 2. Often; 3. Sometimes; 4. Seldom; 5. Never) 0.50 0.60 Energy for daily life Q3.4: How often your household members are not healthy or do not have enough energy for everyday life? (Options: 1. Always; 2. Often; 3. Sometimes; 4. Seldom; 5. Never) 0.59 0.63 Emotion Q3.5: How often do your household members have negative feelings such as blue mood, despair, anxiety, depression? (Options: 1. Always; 2. Often; 3. Sometimes; 4. Seldom; 5. Never) 0.55 0.61 Leisure activities Q3.6: How often do your household members have the opportunity for leisure activities? (Options: 1. Never; 2. Seldom; 3. Sometimes; 4. Often; 5. Always) 0.45 0.52 Overall satisfaction with health status Q3.7: Overall, you are satisfied with your household™s health status 0.66 0.70 Close neighborhood Q4.1: This is a close-knit neighborhood 0.53 0.43 Opportunities of neighborhood interactions Q4.3: There are many opportunities to meet neighbors and work on solving community problems 0.31 0.37 Cohesion Q4.6: Suppose someone in your village/neighborhood had something unfortunate happen to them, such as a family member's sudden death, there are always some others would be ready to help 0.37 0.39 Overall satisfaction with social relationship Q4.7: Overall, you are satisfied with your household™s social relationships with others 0.38 0.36 191 Table S3.1. (Cont™d) Indicator layer Indicator code and content ri-t before earthquake ri-t after earthquake Affordability to quality and nutritious food Q5.2: Your household has affordable access to quality and nutritious food for an enjoyable life 0.73 0.74 Affordability to quality healthcare Q5.3: Your household has affordable access to quality medical care 0.74 0.75 Affordability to quality education Q5.4: Your household has affordable access to quality education 0.72 0.71 Affordability to quality housing Q5.5: Your household has affordable access to spacious and quality house 0.65 0.57 Free choice of employment Q5.6: It is difficult to find a satisfied job 0.30 0.33 Overall satisfaction with freedom of choice and action Q5.8: Overall, you are satisfied with your freedom of choice and actions 0.63 0.61 Notes: ri-t : item-test correlation. Cronbach™s values for standardized items are 0.92 and 0.91 before and after the earthquake, respectively. Except for response options specified after indicator contents, options for all other indicator contents are designed in the five-category Likert scale (i.e., strongly disagree, mild ly disagree, unsure, mildly agree, and strongly agree). All the responses are coded in the order from the lowest score of 1 to the highest score of 5. A higher score represents a higher level of well-being. Actual surveys are conduc ted through face-to-face interviews using the local language that is easily understandable to interviewees. 192 Table S3.2. Standardized coefficients of the confirmatory factor analys is for Human Well-Being Index (HWBI). Dependent variable Independent variable Standardized coefficients Robust S.E. Overall HWBI Q1 0.613*** 0.034 Q2 0.662*** 0.045 Q3 0.914*** 0.042 Q4 0.638*** 0.054 Q5 0.584*** 0.034 Q1Basic material for good life Q1.2 0.826*** 0.018 Q1.3 0.829*** 0.018 Q1.4 0.235*** 0.043 Q1.5 0.427*** 0.035 Q3.6 0.218*** 0.044 Q5.2 0.657*** 0.080 Q5.3 0.374*** 0.079 Q2Security Q2.1 0.617*** 0.035 Q2.2 0.602*** 0.035 Q2.4 0.336*** 0.042 Q2.5 0.314*** 0.042 Q2.6 0.381*** 0.044 Q2.7 0.818*** 0.027 Q1.4 0.262*** 0.044 Q1.5 0.368*** 0.040 Q3.2 0.199*** 0.046 Q3Health Q3.1 0.787*** 0.026 Q3.2 0.642*** 0.042 Q3.3 0.618*** 0.033 Q3.4 0.700*** 0.027 Q3.5 0.597*** 0.032 Q3.7 0.826*** 0.021 193 Table S3.2. (Cont™d) Dependent variable Independent variable Standardized coefficients Robust S.E. Q4Good social relations Q4.1 0.548*** 0.039 Q4.3 0.300*** 0.041 Q4.6 0.438*** 0.037 Q4.7 0.417*** 0.043 Q2.3 0.533*** 0.035 Q3.6 0.367*** 0.048 Q5Freedom of choice and action Q5.2 0.263*** 0.075 Q5.3 0.553*** 0.080 Q5.4 0.909*** 0.017 Q5.5 0.784*** 0.024 Q5.6 0.274*** 0.041 Q5.8 0.808*** 0.023 Q2.6 0.226*** 0.043 Q1 Q5 0.738*** 0.033 Q2 Q4 0.686*** 0.075 Notes: *** p < 0.001. For description of each code, please refer to Appendix A. Only paths that are theoretically meaningful are added. Only path s with coefficients that are tested to be significant (p < 0.05) are included in the mode l. Paths between observed indicators are not shown here. The confirmatory factor analysis is constructed using the MLR estimator in Mplus. The number of total observations is 326. 194 APPENDIX C SUPPORTING INFORMATION FOR CHAPTER 4 Table S4.1. Descriptive statistics of variables used in the model. Variable Mean (S.D.) Change of HWBI Œ 0.080 (0.010) HWBI in 2007 0.638 (0.118) IDES in 2007 0.634 (0.295) Change of IDES Œ 0.303 (0.423) Non-agricultural income share in 2007 0.581 (0.283) Household income in 2007 (thousand yuan) 28.651 (304.954) Household income in 2009 (thousand yuan) 68.105 (104.715) Change of household income per capita (thousand yuan) 11.160 (35.306) Household size in 2007 3.317 (1.319) Number of seniors (age >= 60 years) 0.653 (0.805) Average education of adults in 2007 (year) 5.052 (2.760) Female adult share in 2007 0.469 (0.165) House damage (0: low, 1: high) 0.624 (0.487) Social ties to local leaders (0: weak, 1: strong) 0.109 (0.313) Gender of interviewee (0: female, 1: male) 0.584 (0.495) Age of interviewee (year) 53.604 (11.333) Notes: Change of HWBI, change of IDES, or change of household income per capita are calculated as the corresponding values in 2009 subtracting the corresponding values in 2007, respectively. A household with strong social ties to local leaders is defined as a household with at least one household member or one immedi ate relative (e.g., parents, children, and blood brothers) as local leaders; otherwise it is defined as one with weak social ties to local leaders. The number of observations is 101. 195 Table S4.2. Impacts of the earthquake on net be nefits from socioeconomic activities and ecosystem services. Overall index/sub-index Before earthquake (2007) After earthquake (2009) t value Net socioeconomic benefit (thousand yuan) 13.899 (2.612) 57.681 (9.964) Œ 4.434*** Net benefit from provisioning services (thousand yuan) 10.279 (1.842) 3.063 (0.959) 4.821*** Net benefit from regulating services (thousand yuan) 3.519 (0.226) 3.468 (0.250) Œ 0.730 Net benefit from cultural services (thousand yuan) 0.992 (0.390) 0.424 (0.251) 1.686* Notes: Numbers outside and inside parentheses are means and standard errors, respectively. Monetary values for net benefits in 2009 were discounted into present values of 2007 for comparison. The sampled households in 2007 and 2009 are the same 101 households. *p < 0.05, *** p < 0.001. 196 Table S4.3. Associations between HWBI and affluence before and after the earthquake. HWBI in 2007 before earthquake HWBI in 2009 after earthquake Household income in 2007 (yuan) 0.309** 0.291 ** Household income in 2009 (yuan) 0.254* 0.309 ** Notes: Numbers are Spearman™s rho. The number of observations is 101. *p < 0.05, **p < 0.01. 197 Table S4.4. Supplementary regressions for factors associated with the change of HWBI before and after the earthquake. Dependent variable: Change of HWBI Model (1) Model (2) Model (3) HWBI in 2007 Œ 0.116 (0.082) Œ 0.119 (0.082) Œ 0.117 (0.081) Net benefit from provisioning services in 2007 (thousand yuan) 2.13e-04 (5.29e-04) Œ Œ Net benefit from regulating services in 2007 (thousand yuan) Œ Œ 0.003 (0.005) Œ Net benefit from cultural services in 2007 (thousand yuan) Œ Œ 0.001 (0.002) Change of IDES Œ 0.033 (0.029) Œ 0.027 (0.027) Œ 0.030 (0.028) Agricultural income share in 2007 Œ 0.065 (0.052) Œ 0.064 (0.044) Œ 0.053 (0.044) Household income in 2009 (thousand yuan) 3.27e-04* (1.50e-04) 3.48e-04* (1.33e-04) 3.36e-04* (1.41e-04) Change of per capita income (thousand yuan) Œ 8.38e-04 (5.42e-04) Œ 9.38e-04* (4.28e-04) Œ 8.94e-04* (4.35e-04) Household size in 2007 0.012* (0.006) 0.014 * (0.006) 0.012 * (0.006) Number of seniors (age >= 60 year s) Œ 0.021 (0.017) Œ 0.023 (0.017) Œ 0.022 (0.017) Average education of adults in 2007 (year) 0.006 (0.004) 0.006 (0.005) 0.006 (0.005) Female adult share in 2007 Œ 0.090 (0.071) Œ 0.087 (0.070) Œ 0.094 (0.072) House damage (0: low, 1: high) Œ 0.043* (0.018) Œ 0.041* (0.018) Œ 0.044* (0.018) Social ties to local leaders (0: weak, 1: strong) 0.040 (0.025) 0.038 (0.026) 0.040 (0.026) Gender of interviewee (0: female, 1: male) 0.033 (0.019) 0.032 (0.019) 0.033 (0.019) Age of interviewee (year) 1.60e-05 (0.001) 5.36e-05 (0.001) 6.39e-05 (0.001) Constant Œ 0.018 (0.077) Œ 0.008 (0.081) Œ 0.018 (0.077) F-statistic 2.01* 2.13 * 2.08 * R-squared 0.2485 0.2514 0.2490 N 101 101 101 Notes: Dependent variable is the change of HWBI before and after the earthquake, which is HWBI in 2009 subtracting HWBI in 2007. Numbers inside and outside parentheses are coefficients and robust standard errors, respectively. *p < 0.05. Variance inflation factors are tested to be less than 5. 198 APPENDIX D SUPPORTING INFORMATION FOR CHAPTER 5 1. Supporting methods 1.1. Data 1.1.1. Remotely sensed and geographic information system data We acquired a map and associated docume ntation (e.g., the number of households for monitoring each forest parcel) of the Natural Forest Conservation Program (NFCP) monitored parcels from the administrative bureau of Wolong Nature Rese rve (WNR). Of ~40,100 ha in household monitoring areas, ~24,300 ha in 152 parcels involving 812 households had explicit household monitoring boundaries delineated on the map. Of these 152 parcels, 23 were monitored by single households. The remaini ng ~15,800 ha have boundaries described in the documentation but not on the map. These non-mapped areas were all monitored by single households. We verified that even excluding the 23 parcels, the nonlinear group-size effect on forest outcomes (Table 5.2) still holds (see deta ils in Section 2.5.2). In addition, one parcel co- monitored by nine households and located on the boundary between WNR and the adjacent Sanjiang Township is known to experience intensive land conversion due to tourism development. A Bonferonni test for most extr eme observations (Fox, 1997) also indicates this parcel is the only outlier in our dataset (for this parcel, studentized residuals = Œ 3.786, Bonferonni p-value < 0.05; all other parcels had Bonferonni p-value > 0.05). Therefore, we excluded this parcel and used the remaining 151 spatially explicit parcels (Fig. 5.2) for our study. We used previously published forest-cover maps derived from Landsat imagery in 2001 and 2007 to estimate forest-cover change (Viña et al., 2007; Viña et al., 2011). We generated 199 forest-cover maps via an unsupervised clas sification algorithm using the ISODATA technique, an iterative process for non-hierarchical pixel classification (Jensen, 1996). We performed accuracy assessments using ground-truth points ob tained in the summers of 1998 (209 points), 2000 (83 points), 2001 (83 points), and 2007 (593 points). We measured these points using GPS receivers with high accuracy (error less than 1 m). Overall accuracies of these maps were between 80% and 83%. Major disagreements occurre d primarily in high-ele vation areas and with complex topography (i.e., northwester n part of WNR, Fig. 5.2) rather than in the areas monitored by households that are the focus of our study. T hus, classification accuracies of the selected household monitoring areas should be higher than those of the overall maps. In addition, changes in land cover between field and remotely sensed data collection dates partially accounted for the disagreement between the image classification and ground-truth data. For a detailed description of classification procedures and assessments of map accuracy, please refer to the citied studies (Viña et al., 2007; Viña et al., 2011). We obtained data on topographic characteristics such as elevation, slope, and the Compound Topographic Index, a relative measure of wetness (Gessler et al., 1995), from a digital elevation model at a spatial resolution of 90 m/pixel, acquired by the Shuttle Radar Topography Mission (Berry et al., 2007). We meas ured all household locations inside and surrounding WNR using GPS receivers. Based on th e digitized NFCP parcel map and forest- cover maps, we calculated the size of each parcel , parcel size per household, average elevation, average slope, average wetness, distance be tween each parcel and the nearest household, distance between each parcel and the main road, distance between each household and its monitored parcel, distance between each household and the main road, initial forest cover in 2001, and the percent of forest cover change from 2001 to 2007. 200 1.1.2. Focus group, personal, and household interviews Our research team has been investigati ng our study area since 1995, starting six years before the NFCP implementation. We drew upon bot h our experience in the area and the best available data collected by our research team in studying the underlying mechanisms of nonlinear group-size effects, supplementing quantitative analysis with information obtained from qualitative interviews. We conducted focus group and personal interviews on NFCP in 2007. For retrospective information, we used the standard practice of life-history calendars to enhance respondents™ recall accuracy (Axinn et al., 1999; Freeman et al., 1988). These interviews were collected to understand the NFCP planning, implementation, ev aluation, and decision-making processes and to prepare for the household interview. For focus group interviews, we interviewed eight members of the WNR administrative staff. For pe rsonal interviews, we he ld discussions with five officials who were and/or are in charge of the NFCP planning, implementation, evaluation, and decision making. Our understa nding of how the monitoring groups were formed was based on focus groups and personal interviews, and we further examined the criteria of group assignment with household survey data (see Section 1.2). From household interviews, we used data acqui red from all households that participated in the NFCP in 2007 (surveyed at the end of 2007) and 2009 (surveyed in the summer of 2010). Usually the household heads or their spouses were chosen as interviewees because they are the decision makers and are familiar with household a ffairs. We tracked the same randomly sampled households across the years, but some households were missing in one year or another. For instance, some households were merged, some naturally died out, and some were away for 201 migrant work during our entire investigation period in a given year. A total of 156 NFCP households covering the full range of group size (i.e., from 1 to 16) were interviewed in 2007 to examine how group size affects collective action (i.e., household forest monitoring), and 113 households for group size larger than one were interviewed in 2009 to examine the mechanisms of nonlinear group-size effects. For each analysis, to avoid errors due to using data from different survey years, we only used factors that were m easured for all households in the same year. The information elicited included demographic factor s, household socioeconomic activities, social ties to local leaders, NFCP payment receiv ed, NFCP monitoring e ffort, and within-group monitoring enforcement. The instrument for ho usehold socioeconomic data was based on the standard practices of the National Bureau of Statistics of China (National Bureau of Statistics of China, 2011). Please see a detailed description of variables used for analyses in Section 2. 1.2. Study area and group formation Wolong Nature Reserve (N 30 û45' Œ 31 û25', E 102 û52' Œ 103 û24') is located in Wenchuan County, Sichuan Province, China (Fig. 5.2). It is situ ated in the transition of Sichuan Basin from the east to the Qinghai-Tibet Pl ateau on the west, with elevations ranging from 1,200 m to 6,250 m. WNR™s size was ~20,000 ha in 1963 and expanded to its current size of ~200,000 ha in 1975 (Wolong Nature Reserve, 2005). The Reserve wa s established to protect regional forest ecosystems and rare plant and animal spec ies, primarily the iconic giant panda ( Ailuropoda melanoleuca) (Wolong Nature Reserve, 2005). The majority of local residents are farmers involved in activities such as cultivating maize and vegetables , raising livestock (e.g., pigs, cattle, yaks, horses), collecting traditional Chinese medicinal plants, beekeeping, and collecting 202 fuelwood for heating and cooking. Basic demogr aphic and socioeconomic descriptions are summarized with our household survey data in Table S5.1. NFCP is a nationwide conservation program th at aims to conserve and restore natural forests with logging bans, afforestation, and monitoring using a payments-for-ecosystem- services scheme to motivate conservation behavior (Liu et al., 2008). To respond to the national call, and to restrain the degrada tion of forest and panda habitat over the past three decades (Liu et al., 2001), WNR started to implement the NFCP in 2001. The central government allocated an annual fund of two million yuan for NFCP impl ementation in the Reserve. To improve enforcement of the NFCP and liv elihoods of local residents, the Wolong Administrative Bureau decided to use about half of the NFCP funds to engage local households in the forest monitoring program (Wolong Nature Reserve, 2005). The initial idea of the Wolong Administra tive Bureau was to assign each household a forest parcel to monitor, but it turned out to be too difficult to clarify the boundaries of many small parcels and would be too costly for management. Finall y, the Bureau followed natural boundaries (e.g., rivers, ridges, valleys) of forest parcels and divided and assigned them to household groups ranging from 1 to 16. Of th e total ~120,500 ha NFCP monitoring area, ~40,100 ha were assigned to ~1,100 rural households; the remaining areas were monitored directly by the Bureau™s staff (Fig. 5.2). According to our interviews with government officials who were in charge of the NFCP implementation, large parcels were assigned to large household groups to keep parcel size per household similar across monitoring areas. Parc els distant from household locations were assigned to large household groups with slightly higher payments. We compared these arrangements with our independent ly collected household survey data and found the reports from 203 household interviewees were consistent with those from the progr am officials (Table S5.2). We also conducted statistical analyses and found that the distance from each parcel to its monitoring household(s) and NFCP payment were statistica lly exogenous to the group-size effects (see Section 2.4.3). Thus, distribution of households to groups, although not completely random as in a classical experiment, is suitable to exam ine the group-size effects and the underlying mechanisms. 1.3. Tobit model 1.3.1. Model specification The dependent variable is the total labor input per year by a household for monitoring. The distribution of this variable suggests a Tobit model, in which a large fraction of the observations cluster at the minimum value (zero in this case) or maximum va lue (20 in this case). Conceptually, this is a censored value model in that it treats the minimum and maximum values as if the values of monitoring were not obser ved. The minimum monitoring effort is zero by default, which means a household does not spend any time in monitoring. According to our field investigation, the maximum annual payment for monitoring is ~1,000 yuan, and the local labor price in 2007 was ~50 yuan per laborer per day. This suggests the maximum monitoring effort of a household is 20 laborer days if they are strict economic humans. This theoretical estimate was consistent with empirical data collected from reports of local natural resource managers and monitoring households. In other words, households spent an amount of labor less or equal to what they received in NFCP payments. Thus, besi des zero as the minimum monitoring effort, we also set 20 as the maximum value in the Tobit model. 204 When analyzing censored data, traditiona l regression methods (e.g., ordinary least squares, OLS) may yield inconsistent estimates and provide inappropriate predictions. However censored regression (i.e., Tobit model) can produ ce consistent and efficient estimates of model parameters and partial effects, as well as appropriate predictions (Wooldridge, 2002). A Tobit model is given by (Wooldridge, 2002): biybaiyabiyaiyiy*1*1*1*11 (5.1) iuiXiy11*1 , (5.2) where iy1 is the observed monitoring effort, *1iy is a latent variable satisfying the classic linear model assumption, a is the minimum limit, b is the maximum limit, iX1 is a vector of exogenous explanatory variables, 1is a parameter vector to be estimated, i is the ith observation, and iuis an error term that has a norma l distribution with mean of zero. When the Tobit model contains endogenous va riables, Eq. (5.2) is specified as follows (Wooldridge, 2002): iuiXiyiy1122*1 (5.3) iviXiXiy22112, (5.4) where iy2 is a vector of potentially endogenous explanatory variables and the equation for iy2 is written in reduced form, iX1 is a vector of exogenous explanatory variables, iX2 is a vector of additional instruments, 2 is a vector of structural parameter, 1 and 2are matrices of 205 reduced-form parameters, and iv is an error term that has a no rmal distribution with mean of zero. 1.3.2. Model construction Descriptive statistics of vari ables used in model constructio n are shown in Table S5.1. To improve the interpretability of coefficients a nd reduce the collinearity between the linear and quadratic terms of group size, all continuous inde pendent variables and instruments were mean centered prior to their input into the regression models (Schielzeth, 2010). We first constructed our models with OLS without c onsidering the censoring effects, then constructed the Tobit models considering the censoring effects. Whet her or not we took the censoring effects into account, the nonlinear group-size effects were consistent. Thus, we reported the final results from the Tobit models. Parameter and marginal e ffect estimations were conducted using Stata 12 (StataCorp LP, USA). 1.4. Spatial autoregressive model 1.4.1. Model specification Previous studies on group-size effects have not considered the spatial autocorrelation of measurements across resource units (e.g., forest parc els) of key variables (e.g., percent of forest-cover change). Ignoring spatial autocorrelation may violate the assumption of independently distributed errors of classical statistical tests and may lead to incorrect conclusions (Lichstein et al., 2002). In brief, the ecological reason for considering spatial autocorrelation here is that a parcel is more likely to regenerate or recove r faster when its surrounding parcels are forested 206 (Lichstein et al., 2002). We constructed spatial autoregressive models to take spatial autocorrelation into consideration. The general mixed form of the spatial autoregressive model in our study is given by (Anselin, 1988; LeSage and Pace, 2009): y = Wy + X + (5.5) = W + , (5.6) where y is the n (number of observations) × 1 vector of the dependent variable (i.e., percent of forest-cover change from 2001 to 2007), is the coefficient of th e spatially lagged dependent variable, W is a given n × n spatial weighting matrix, X is the n × k (number of independent variables plus intercept) matrix of th e independent variables plus intercept, is the k × 1 vector of coefficients, is the spatial error correlation coefficient, and is the n × 1 error term assumed to be independent and identically distributed. The mixed model is reduced to a spatial lag model when = 0, to a spatial error model when = 0, and to a traditional regression model when both are zero. 1.4.2. Model construction Descriptive statistics of va riables used in the model cons truction are shown in Table S5.3. A spatial weighting matrix of forest parcels was created using the GeoDa software (version 0.9.9.1), defining a neighbor based on the Queen contiguity approach (i.e., common borders and corners) (Anselin et al., 2006). We compared different spatial weighting matrices (i.e., Queen contiguity of order 1, Queen contiguity of order 2, Rook contiguity of order 1, and Rook contiguity of order 2), and the results are similar. Thus, we reported the results using Queen contiguity of order 1. Model construction and all statistical analys es were performed using the R 207 software (version 2.12.2) (R Development Core Team, 2011). Spatial simultaneous autoregressive models were construc ted with the package fispdepfl in R. 1.5. Structural equation model 1.5.1. Model specification A structural equation model (SEM) is a statistical technique for testing and estimating causal relationships (Pearl, 2000; Simon, 1953; Wr ight, 1921). It allows analysis of multiple simultaneous causal relations among endogenous variables, and between endogenous and exogenous variables. A typical SEM contains two main compone nts: the structural model representing potential causal dependencies between endogenous and exogenous variables, and the measurement model representing the relations between latent variables and their indicators. The general form of SEM is gi ven by (Bollen and Noble, 2011): Structural model: jjjBj (5.7) Measurement model: jjyyjy (5.8) jjxxjX , (5.9) where j is the vector of latent endogenous variables for unit j; , y, and x are intercept vectors; j is the vector of latent exogenous variables; jy and jX are vectors of the observed indicators of j and j, respectively; B is the matrix of coefficien ts giving the expected effects of the latent endogenous variable () on each other; is the coefficient matrix giving the expected effects of the la tent exogenous variables () on the latent endogenous variables (); 208 yand xare the matrices of coefficients giving the effects of the latent j and j on jy and jX, respectively; j, j, and jare the vectors of error terms; and j is the jth observation. The commonly used factor analysis, regression analysis, and path analysis methods are all special cases of SEM. Specifically, path analysis is SEM with a structural model but no measurement model. In this study, all variables can be reasonably treated as observable. So we used SEM for path analysis of the nonlinear group-size effects. Guided by the two hypothetical, opposing forces through which group size affects collective action and then resource outcomes, based on previous literature, we hypothesized that some factors may affect group size, group size may directly affect free riding, and group size ma y also indirectly affect free riding through within-group enforcement (Fig. 5.1). 1.5.2. Model construction Descriptive statistics of variables used in the model are shown in Table S5.4. Three structural models of increasing complexity were constructed. First, whether or not a household would be a free rider was estimated as a func tion of the size of its group and within-group enforcement. Second, within-group enforcement was estimated as a function of monitoring group size, whether or not there was within-group division (see description in Section 2.3), and interaction term of group size and within-group division. Third, other factors that may affect group size were controlled as exogenous variables acting on group size. Because some dependent variables are binary (i.e., free rider and within-group enforcement), we conducted the path analysis using Mplus (Muthén and Muthén, 1998-2010), which handles path analysis with categorical outcomes. We used the default robu st weighted least squares (WLSMV) estimator, which uses a diagonal weight matrix with stan dard errors adjusted, and mean- and variance- 209 adjusted chi-square test statistics (Muthén and Muthén, 1998-2010). The final path model retained some variables that were not signifi cant because the goodness of fit of the path model was high (Table S5.5), and because those variables we re theoretically interesting as controls (see Section 2.3). 2. Supporting text and analysis 2.1. Forest monitoring and within-group enforcement 2.1.1. Monitoring efforts: household labor input The forest monitoring efforts of each house hold were measured by the total amount of labor input. One unit of labor input was defined as one work day (i.e., eight work hours) of a laborer spent on monitoring activi ties, including the time travelling to and from the monitored parcel. In our data, each household either does no t send any laborer or se nds only one laborer for each monitoring activity to join with other laborers from the assigned monitoring group. Therefore, for each household, the total annual amount of labor input equals the total work days that one laborer spent on monitoring (Eq. 5.10). Household labor input (laborer days) = Monitoring frequency (times) × Time per monitoring (hours) / 8 (hours day-1 laborer-1) (5.10) Illegal logging activities can be detected in several ways. First, because many households live near the forest and conduct agricultural activities in and around the forest, they can hear the sounds of cutting and falling trees. Local househol ds can also see illegal loggers when they transport the timber to their homes or along the main road to outside areas (because it is a mountainous area with complex topo graphy, it is very difficult to transport wood except via the main road). Second, for forests far from local ho useholds, illegal activities may be detected by 210 households who are monitoring the forests or othe r people who happen to pass by as they collect fulewood or conduct other legal activities. Local households are motivated to report illegal activities because the local government rewards reporters in cash. Finally, even if these ways fail, the timber checkpoints at the two ends of the main road across the reserve also can detect illegally logged timber. 2.1.2. Free riders Free riders are defined as people who rece ive the NFCP payment but do not spend time and labor on forest monitoring. Here, we classifi ed those households who self-reported that they did not conduct monitoring activit ies as free riders. Our team ha s been conducting research in this study area since 1995 and has established good social relationships with local households. Social desirability would inclin e respondents toward overreporting their monitoring efforts. Thus, we could be reasonably certain that those households reporting zero monitoring effort did in fact free ride. 2.2. Forest monitoring outcomes: changes in forest cover The main aim of assigning forest parcels to households for monitoring is to prevent logging. Thus, to assess the outcomes of househol d monitoring efforts, the most important indicator is the number of trees in each parcel. However, it is difficult and costly, if not impossible, to count all trees. An alternative appr oach is to assess the forest cover. We adopted the forest definition of the Food and Agriculture Organization of the United Nations as fiLand spanning more than 0.5 hectares with trees higher than five meters and a canopy cover of more than 10 percent, or trees able to reach these thresholds in situ fl (Forestry Department of Food and 211 Agriculture Organization of the United Nations, 2010). Detailed descriptions of the forest-cover data are summarized in Section 1.1 and our previ ous publications (Viña et al., 2007; Viña et al., 2011). The forest monitoring outcome was therefor e measured on a per-parcel basis by percent forest-cover change (i.e., percent forest cover in 2007 minus percent forest cover in 2001). 2.3. Structural factors Two main barriers for examining group-size e ffects are the heterogeneities of groups and resource units (Gautam, 2007; Ostrom, 1990; Po teete and Ostrom, 2004). In this section, we provide a review of structural factors used in our analyses. We include factors commonly found to be relevant to collective action and commons management (A grawal and Goyal, 2001; Baland and Platteau, 1996; Ostrom, 1990; Rustagi et al., 2010): characteristics of household groups, households, and forest parcels. 2.3.1. Characteristics of household groups Group size. Group size refers to the numb er of households for monitoring a single forest parcel. Within-group enforcement. We regarded a group as having strong monitoring enforcement if punishment measures (e.g., payment deduction, ve rbal condemnation) existed within the group for members who did not participate in monitoring (i.e., free riding). Otherwise, we regarded a group as having weak monitoring enforcement. Within-group division. Groups with two or more households could divide laborers to improve monitoring efficiency. In our case, if groups were divided into small subgroups to conduct monitoring in turns, we coded them as groups ha ving within-group division of labor; otherwise, groups did not have within-group division of labor. 212 2.3.2. Characteristics of households Household size. Household size refers to the number of household members. Education of adults. Education affects an individual™s atti tude and behavior (Rustagi et al., 2010). Since adult household members are the main decision makers of hous ehold activities and actually participate in forest monitoring, we us ed the average education level of all adult household members. Number of laborers. The number of laborers in a household is a measure of available household labor resources for the forest monitoring activity. A laborer is defined as an individual between the ages of 15 and 59. Social ties to local leaders. Social organization in rural areas in China (such as our study area) is largely based on kinship and leadership. Local leaders are well doc umented to be influential on the behavior of group members (Rustagi et al., 2010). Therefore, we expected connections to local leaders to affect a household™s contribution to collective action. In our study area, individuals who work as leaders in villages, administrative groups, or local government-owned enterprises are widely regarded as local leaders. We defined a household with strong social ties to local leaders as a household that had at least one household member or one immediate relative (e.g., parent, child, brother) who was a local leader; otherwise, a household had weak social ties to local leaders. Age of adults. As individuals get older, their household structures and social ties also change because their relatives and friends die, and/or their children leave home. Thus, the age structure of a household may be an important factor affect ing social ties to local leaders. Since adult 213 household members are the actors fo r social ties connecting to local leaders, we used the average age of all adult members of a household as the age structure measurement for each household. Percentage of adult females. Gender plays an important role in developing different social ties (van Emmerik, 2006). Thus, differences in th e proportion of women among adult household members may contribute to the differences in ho useholds™ social ties to local leaders. The percentage of adult females among adults refers to the ratio of the number of female adults to total adults in a household. Household income. We acquired gross household income data from face-to-face interviews following the standard protocol of the National Bureau of Statistics of China (National Bureau of Statistics of China, 2011). In our study area, household income covers a wi de range of categories such as agricultural income (e.g., from animal husbandry, sales of crops and/or nontimber forest products), wage income, small businesses income (e.g., operating restaurants, hotels, and other tourism-based businesses), property income (e.g., land and housing rents), gift income from relatives and friends, government payments for ecosy stem services, and social security benefits (e.g., low-income subsidy, pension). Per capita household income. Per capita household income is the total household income divided by household size. Agricultural income. Agricultural income refers to income related to agricultural practices such as cultivating cropland, raising livestock, and collecting nontimber forest products. Percentage of agricultural income. The percentage of agricultural income is the ratio of household agricultural income to its total household income. Area of cropland. The area of cropland refers to the total area of cropland owned by a household. 214 NFCP payment. NFCP payment refers to the amount of cash subsidy a household received for participating in the NFCP forest monitoring program. Distance between each household and the main road. In our study area, households farther from the main road are likely to cultivate more cropland (Spearman™s = 0.436, p < 0.001), rely more on agricultural income (Spearman™s = 0.249, p < 0.01), and are closer to their monitored parcels (Spearman™s = Œ 0.201, p < 0.05). Such reflected heterogeneity of households may also affect their participation and c ontribution to forest monitoring. Distance between each household and its monitored parcel. The further a household is away from its monitored parcel, the longer distance the household must travel and the more time it takes for a single monitoring activity. Therefor e, the distance from each household to its monitored parcel is a surrogate measure of mon itoring cost. Because this factor was correlated with group size and might cause an endogeneity problem (Table S5.2), we did not directly include it in the models (Table 5.1 and 5.2). Instead, we treated it as a hypothesized instrumental variable and examined whether the nonlinear gr oup-size effects were caused by this factor or indeed by the group size (see Section 2.4.3). 2.3.3. Characteristics of forest parcels Elevation. Because elevation is correlated to climatic factors such as temperature, it is an important factor influencing forest growth. Furthermore, elevation is also a natural barrier that reduces human access. However, since household- monitored parcels are mainly located along the main road and at relatively lower elevati ons, the variation of average elevation for each parcel is not large (Table S5.3). 215 Slope. Similar to elevation, steeper slope reduces human access and thus reduces logging and forest degradation. Wetness. Wetness, as measured by the Compound Topographic Index, is a measurement of relative soil moisture (Gessler et al., 1995). Soil moisture is an important factor affecting forest growth (Viña et al., 2011). We expected forests to recover faster in relatively wetter regions. Parcel size. Larger parcels are more exposed to loggi ng and require more monitoring efforts to prevent illegal access (Agrawal and Goyal, 2001). Therefore, the parcel size is an important factor that may affect forest-cover changes. Parcel size per household. Parcel size per household is the total size of a parcel divided by the number of households assigne d to monitor the parcel. Initial forest cover in 2001. Initial forest cover is a key factor to determine the potential of forest growth. A region with high initial forest cover doe s not have much room to grow and thus forest regeneration is more likely to occur in pl aces with relatively lower initial forest cover. Distance between each parcel and the main road. Since most households and household- monitored parcels are located along the main road (Fig. 5.2), illegal harvests closer to the main road should be easier to catch. Therefore, the distance between each parcel and the main road is a measurement of the difficulty of de tecting illegal harvest. We used the distance from the centroid of each parcel to the main road as an average estimate of di stance for each parcel. The same approximation of using the centroid of each parcel was adopted for measuring distances between each parcel and other locations (e.g., the ne arest household, each parcel™s corresponding monitoring households). Distance between each parcel and the nearest household. Distance between each parcel and the nearest household is a factor measuring re source vulnerability to illegal harvest. Since 216 households tend to collect forest products (e.g., fuelwood) closer to their households, forests closer to households may be more likely to suffer from illegal harvest. 2.4. Causality Experiments with randomization are usually considered the best method for establishing causal relationships. But such experiments are hard to conduct in the field and around policy implementation, and laboratory experiments often suffer from a lack of external validity. However, our 17-year investigation in our study area both before and after NFCP implementation and evidence from the literature and supplemental analyses of our data suggest that we have established plausible evidence for causal effects of group size on both collective action (i.e., forest monitoring by household group s in our case) and resource outcomes (i.e., changes in forest cover in our case). 2.4.1. Why does collective action contribute to resource outcomes? Evidence from literature. The fiexternalityfl characteristic of common-pool resources explains why collective action could lead to their destru ction (i.e., the tragedy of the commons) (Hardin, 1968). This conventional wisdom was challenged by the landmark work of Elinor Ostrom and her colleagues, suggesting that collective action can guard the provision of common-pool resources by reducing free riders through means such as clarif ication of resource boundaries, designing adaptive access rules, and monitoring (O strom, 1990; Ostrom, 2009; Ostrom et al., 1999; Ostrom et al., 1994; Ostrom et al., 2007) . Laboratory experiment s (Boyd et al., 2010; Carpenter, 2007; Fehr and Gachter, 2002) and field observations (Agrawal and Chhatre, 2006; Chhatre and Agrawal, 2008; Gibson et al., 2005; Rustagi et al., 2010) also provided evidence that 217 monitoring and sanctions could reduce free riders and enhance cooperation and thus improve resource outcomes. Evidence from our analyses. Ideally, we should use forest outcome as a dependent variable and include group monitoring efforts as one independent variable with other control variables in a regression model at the parcel level (i.e., at the group level) to test the association between group monitoring efforts and group outcome. Unfortunatel y, we do not have household survey data for all households participating in forest monitoring activities. For instance, we may have information for two of the ten households that monitor a parcel. Thus, we could not measure group monitoring efforts at the parcel level. The alternative approach is to conduct analyses for collective action at the household level and for outcomes at the parcel level (Tables 1 and 2). Here, we provided additional analyses to explain why the monitoring efforts contributed to the forest outcomes. First, illegal logging by people from outsid e WNR has not been a problem since the NFCP implementation. The complex topography provi des a natural barrier to prevent illegal logging from outsiders. There is only one main road through the Reserve (Fig. 5.2), with a timber checkpoint located at each end. Based on our field investigation, before the NFCP implementation, some employees at the two timber checkpoints were involved in illegal log transportation with outsiders, but after the NFCP implementation, this problem was solved and the forest laws and regulations have been strenuously enforced. In addition, all our interviewed households and government officials shared a cons ensus view that illegal logging from outsiders had almost disappeared in WNR. Therefore, logging would be largely the action of local residents, and household monitoring efforts could effectively enhance NFCP enforcement, reduce illegal logging, and contribute to forest recovery. 218 Second, our results show similar nonlinear e ffects of group size on both collective action and resource outcomes (Tables 5.1 and 5.2). This is also indirect evidence to support the inference that collective action contributed subs tantially to the outcomes. Since there are no forest parcels without NFCP m onitoring in our study area and surrounding regions, we could not compare the outcomes between parcels with and without monitoring. However, our study (Tuanmu, 2012) compared panda habitat recove ry rates between household-monitored parcels and government-monitored parcels. We found that panda habitat recovered faster in household-monitored parcels than in government-monitored parc els. Because forest is the essential part of panda habitat, the results supported that household monitoring directly improved forest and panda habitat recovery. Meanwhile, all our evidence indirectly indicates that, along with intangible social norms and networks, house hold monitoring prevents illegal logging and contributes to forest recovery. In the following sections, we further elucidat e how tangible actions and intangible social norms and netw orks affected collective action and resource outcomes. 2.4.2. How do social ties to local leaders affect collective action? Evidence from literature. Social learning theories suggest that individuals do not imitate behaviors from others randomly, but rather that leaders will be disproportionally imitated, and thus have more influence on others™ behaviors (Henrich and Gil-White, 2001; Milinski et al., 2002; Rustagi et al., 2010). Compared to others, l eaders often are elders and/or wealthier, more educated, prestigious, reputable, powerful (Henrich and Gil-White, 2001; Rustagi et al., 2010). Therefore, behaviors of leaders would more likely influence ot her members and thus affect collective action. 219 Although the detailed mechanisms of how soci al capital, including social norms, ties, and networks, affect collective action have not yet been theorized, it is widely recognized that social capital plays an important role in affecting collective action and resource outcomes (Anthony and Campbell, 2011; Dietz and Henry, 2008). On one hand, social capital may encourage trust and communication, ensure rule compliance, reduce monitoring and transaction costs, and thus enhance collective action (Pretty, 2003). On the other hand, social capital may encourage political coercion and act as an obstacle to shape inappropriate social arrangements (Adhikari and Goldey, 2010; Portes and Landolt, 1996). Evidence from our analyses. To understand why and how social ties to local leaders affect household monitoring efforts in our case, we first examined what characteristics determine whether a household has strong or weak social ties to local leaders. Our results suggest that households with strong social ties to local leaders tend to have a higher average age and education level of adults (Tables S5.6 and S 5.7). Because our measurement of social ties occurred during the NFCP implementation period, and the birth years and education levels of adult household members were determined far befo re the establishment of their social ties to local leaders, the causal inference can only be that higher average age and education of adult members help a household to accumula te strong social ties to local leaders rather than vice versa. Our results (Table 5.1) also suggest that households with strong social ties to local leaders tended to spend less on monitoring efforts. The reasons are implied by the words of some household interviewees, for example, fiAlmost every household has been assigned to a forest parcel. If you go to cut trees in others™ parcels and happen to be known by them, it will harm social relationships with them. – For households with strong social ties to local leaders, we dare not and do not 220 want to offend them because they know more than us, have more social relationships and power, and we often need to turn to them for help.fl The staff members in the Wolong Admini strative Bureau who are in charge of combatting illegal logging activities are hired from outside WNR (usually college graduates) and are not related to local residents. In addition, anyone can report illegal logging and receive a cash reward from the Bureau. We are also not awar e of a single case in which staff members have turned a fiblind eyefl to illegal logging so households with strong ties to them could avoid monitoring or sanctions. Thus, these reasons cannot explain why households with strong social ties to local leaders have less need to monitor. Rather, combining interviewees™ statements and the analyses above, the reasonable explanation is that households with strong social ties to local leaders often have more social relationships, power, knowledge, and experience than households with weak social ties to local leaders, and such social ties pr ovide social capital and repuation that prevent others from illegal activities in their monitoring parcels. Thus, these households have less need to monitor. 2.4.3. Is the nonlinear effect on collective action really caused by group size? To answer this question, we used two approaches to support our argument that the nonlinear effect is indeed caused by group size and that instrumental vari ables affect collective action through group size. Since only the distance from each household to its monitored parcel and the NFCP payment were linearly associated with group size as criteria in group formation (Section 1.2, Table S5.2), we used them as hypothesized instrume ntal variables. For the first approach, we examined the association type between each hypothesized instrumental variable and colle ctive action. Our results (Table S5.8) suggest that either the 221 distance from each household to its monitored parcel or the NFCP payment is linearly associated with collective action. Since all the additional controls are the same as the ones used in the model of nonlinear group-size effect on collective action (Table 5.1 or Table S5.9), these results suggest that the two hypothesized instrume ntal variables cannot explain the nonlinear effect of group size on collective action. Rather they affect collective action through group size. These results also support that the group-size distribution in our dataset, although not completely random, is still suitable to analyze the group-size effects. For the second approach, we used a two-st ep Tobit model with endogenous variables. Using either the instrument of distance from eac h household to its monitored parcel or the NFCP payment, our results (Table S 5.10) suggest that our instrument s are powerful (F-statistic > 47, p < 0.001) and exogenous (Wald test of exogeneity of p > 0.1). The second-stage regression (Table S5.11) also suggests that group size has a nonlinear effect on household monitoring efforts, regardless of using any of the two hypothesized instruments. These results also suggest that the distance from each household to its monitored pa rcel and the NFCP payment affect collective action through group size. Again, this supports that the group formation in our study, which may not be completely random, does not constitute an impediment to examine group-size effects. 2.4.4. How does group size cause nonlinear eff ects on collective action and resource outcomes? Based on the results of path analysis (Fig. 5.1, Table S5.5), we explained the mechanisms of how group size affects collective action and resource outcomes through two opposing forces in the main text. Here we provided additional qua litative evidence from our interviews to support this conclusion. 222 The causal inference of the mechanisms of nonlinear group-size effects were also confirmed by our interviewees. One of our house hold interviewees provided us a vivid example, expressing a point also ma de by many other interviewees, of how the cost of social relationship deterioration for free riders would increase with group size: fiIf I do not go to monitor the parcel a ssigned to our group, only one group member would complain to me if the parcel is co -monitored by the two of us, but nine other households may do so if I am in a group of 10 households.fl 2.5. Supporting regression models In this section, we present a more detailed set of the control variables for the results shown in Tables 1 and 2 in the main text. We al so present regression diagnostics and results of other supporting regression models to support our results that the nonlinear group-size effects are robust even when we (i) used different combinati ons of control variables, (ii) did not consider spatial autocorrelation, and (iii) discarded some edge points (i.e., observations with group sizes of 1, 15, or 16) from the total set of observations. To display the distribution of group size and how collective action or resource outcomes change with group size, we also visualized the nonlinear relationship between group size and collective action or forest outcomes (Fig. 5.3). Ho wever, given the nature of data and methods we used (i.e., Tobit model for censored data a nd spatial autoregressive model for data with spatial autocorrelation), our models are not simple, classic regression models (e.g., OLS regression). Visually, the actual observations do not fit the predicted lines in the same way as those in OLS regression s (Daniel et al., 1999). 223 2.5.1. Supporting results and regression diagnostics of Tobit models For supporting results and diagnostics of a ll combinations of Tobit models, please see Tables S5.9, S5.12, and S5.13. 2.5.2. Supporting results and regression diagnos tics of spatial autoregressive models For the construction of the spatial autoregressive model, we compared spatial mixed, spatial lag, and spatial error models. The coeffici ent of the spatially la gged dependent variable (i.e., ) was not significant (z-value: Œ 0.504, p > 0.1) in the mixed model. The coefficient of the spatial error correlation (i.e., ) was significant (p < 0.001) in both the mixed model (z-value: 5.691, p < 0.001) and the error model (z-value: 8.553, p < 0.001). Meanwhile, the error model had the minimum Akaike Information Criteria (AIC ) value. The AIC values for the mixed, lag, error, and OLS models were Œ 312.89, Œ 298.14, Œ 314.56, and Œ 279.89, respectively. These results suggest that the error model was most appropriate. We examined both the linear and nonlinea r relationships between group size and the percent of forest-cover change from 2001 to 2007. The coefficient of group size in the linear model was nonsignificant (p > 0.1, Table S5.14), whic h is not surprising given the presence of an optimum point within the range of the data. Whet her we included spatial autocorrelation or not, the coefficients of both the quadratic and linear terms of group size we re significant (Tables S5.15 and S5.16). But Moran™s I test suggests that spatial autocorrelation should be included (Table S5.15). Thus we report the results from th e spatial autoregressive model in the main text. We added the cubic term of group size into the model in Table 5.1 (or Table S5.16). It was nonsignificant (z-value: 0.221, p > 0.1). 224 As mentioned in Section 1.1.1, even excludi ng parcels with a group size of one, the quadratic term of group size was still signifi cant (z-value = Œ 2.460, p < 0.05). Given there was only one parcel with group size of 15 and one of 16 in our dataset, we also tested the group-size effects by excluding these two parcels. The nonlinear group-size effect is still significantly present among the remaining 149 parcels (z-v alue = Œ 2.552, p < 0.05 and z-value = Œ 2.872, p < 0.01 for the quadratic and linear terms of group size, respectively.) 225 Table S5.1. Descriptive statistics of variables for 156 randomly sampled monitoring households. Variable Mean (S.D.) Total labor input for monitoring (dependent variable, laborer day) 5.23 (6.15) Group size (number of households) 3.95 (4.15) Social ties to local leaders (binary: 0 for weak social ties; 1 for strong social ties) 0.15 (0.36) Distance between each household and its monitored parcel (km) 4.40 (4.82) Distance between each household and the main road (km) 0.40 (0.60) Number of household labo rers (individual) 1.98 (1.05) Household size (number of individuals) 3.48 (1.33) Education of adults (year) 4.80 (2.60) NFCP payment (yuan) 862.60 (87.27) Household income (yuan) 27,965.42 (28,637.83) Percentage of agricultural income 41.69% (28.86%) Table S5.2. Correlation between group size and other biophysical, demographic, and socioeconomic variables. Group size Spearman™s Distance between each household and its monitored parcel (km) 0.214** NFCP payment (yuan) 0.522*** Distance between each household and the main road (km) Œ 0.142 Social ties to local leaders (binary: 0 for weak social ties; 1 for strong social ties) 0.051 Household size (number of individuals) Œ 0.117 Average age of adults (year) Œ 0.061 Education of adults (year) 0.114 Number of household la borers (individual) 0.008 Household income (yuan) Œ 0.021 Per capita household income (yuan) 0.050 Agricultural income (yuan) Œ 0.162 Percentage of agricultural income Œ 0.161 Area of cropland (mu, 1 ha = 15 mu) Œ 0.061 Notes: Tested with 156 randomly sampled households. **p < 0.01; *** p < 0.001. 226 Table S5.3. Descriptive statistics of variables for 151 household-monitored parcels. Variable Mean (S.D.) Percent of forest-cover change from 2001 to 2007 (dependent variable) 13.66% (12.50%) Group size (number of households) 5.32 (3.45) Parcel size (100 ha) 1.59 (1.37) Parcel size per household (100 ha per household) 0.33 (0.27) Elevation (1,000 m, above sea level) 2.42 (0.37) Slope (radian) 0.53 (0.07) Wetness (unitless) 10.89 (0.52) Distance between each parcel and the nearest household (km) 3.79 (4.64) Distance between each parcel and the main road (km) 2.60 (3.39) Initial forest cover in 2001 64.21% (31.75%) Table S5.4. Descriptive statistics of characteristics of 113 randomly sampled monitoring households and their assigned groups. Variable Mean (S.D.) Group characteristics Within-group enforcement (binary: 0: weak enforcement; 1: strong enforcement) 0.38 (0.49) Group size (number of households) 7.04 (3.31) Within-group division (bin ary: 0: no within-group division; 1: has within-group division) 0.45 (0.50) Household characteristics Free rider (binary: 0 for a household that does not free ride; 1 for a household that free rides) 0.24 (0.43) Social ties to local leaders (binary: 0 for weak social ties; 1 for strong social ties) 0.27 (0.45) Distance to the main road (km) 0.37 (0.62) Number of household laborers (number of individuals) 2.53 (1.23) Household size (number of individuals) 3.68 (1.57) Education of adults (year) 5.51 (2.76) Household income (yuan) 34,257.02 (36,296.61) Percentage of agricultural income 13.74% (21.88%) Notes: Because collective action within each gro up requires a group having at least two members, here only those groups with at least two members were included for analysis. 227 Table S5.5. Path analysis of the two opposing forces through which group size affects collective action. Path analysis Unstandardized coefficient (S.E.) Standardized coefficient Dependent variable: Free rider Group size 0.146** (0.051) 0.314 ** Within-group enforcement Œ 0.522** (0.184) Œ 0.476 ** Dependent variable: Within-group enforcement Group size 0.103** (0.038) 0.243 ** Within-group division 0.376 (0.266) 0.178 Group size × Within-group di vision Œ 0.050 (0.061) Œ 0.104 Dependent variable: Group size Social ties to local leaders 0.052 (0.651) 0.009 Distance between each household and the main road (log) Œ 0.067 (0.136) Œ 0.052 Number of laborers Œ 0.051 (0.350) Œ 0.025 Household size 0.027 (0.243) 0.017 Education of adults 0.016 (0.117) 0.018 Household income (log) Œ 0.093 (0.311) Œ 0.031 Percentage of agricultural income 1.839 (0.946) 0.162 Tests of model fit Chi-Square/degrees of freedom 10.854/18 Comparative Fit Index (CFI) 1.000 Tucker-Lewis Index (TLI) 2.898 Root Mean Square Error of Approximation (RMSEA) 0.000 90% Confidence Intervals of RMSEA (0.000, 0.036) Notes: Unit of analysis is the household, but characteristics of bot h households and their assigned groups are considered. Continuous independent variables are mean centered. Total number of observations is 113 households. **p < 0.01. 228 Table S5.6. Characteristics of households with strong and weak social ties to local leaders. Variable Households with weak social ties Households with strong social ties Test statistic (t value) Distance between each household and the main road (km) 390.39 (51.89) 424.40 (134.00) Œ 0.237 Household size (number of individuals) 3.53 (0.12) 3.33 (0.29) 0.638 Percentage of females in adults 0.48 (0.02) 0.47 (0.03) Œ 0.292 Education of adults (year) 4.55 (0.22) 6.21 (0.58) Œ 2.904**Average age of adults (year) 46.31 (0.92) 46.90 (2.07) Œ 0.263 Household income (yuan) 27,132.16 (2471.90) 32,209.08 (6182.76) Œ 0.763 Per capita household income (yuan) 8,309.03 (963.20) 10,858.72 (2188.14) Œ 1.067 Agricultural income (yuan) 11,337.70 (1,455.17) 13,173.63 (3,113.16) Œ 0.534 Percentage of agricultural income 0.42 (0.02) 0.44 (0.07) Œ 0.209 Area of cropland (Mu, 1 Mu = 1/15 ha) 4.03 (0.21) 4.02 (0.44) Œ 0.030 Notes: Numbers within parentheses are standa rd error of mean. The test used was unequal variance t-test. **p < 0.01 229 Table S5.7. Logit estimation of factors associat ed with social ties to local leaders. Variable Coefficients (Robust S.E.) Marginal effects Intercept Œ 8.711* (3.842) Œ Township (dummy) 0.071 (0.488) 0.008 Distance between each household and the main road 0.435(0.503) 0.052 Household size Œ 0.073 (0.215) Œ 0.009 Percentage of females in adults Œ 0.690 (1.219) Œ 0.082 Education of adults 0.321** (0.109) 0.038 Average age of adults 0.048* (0.024) 0.006 Household income (log) 0.352 (0.354) 0.042 Percentage of agricultural income 0.255 (0.981) 0.030 Area of cropland Œ 0.053 (0.121) Œ 0.006 Notes: Dependent variable is the social ties to local leaders (binary: 0 for weak social ties; 1 for strong social ties). Total number of observ ations is 156. Log pseudolikelihood is Œ 60.337. Pseudo R-squared is 0.099. Variance Inflation Factors were tested to be < 5. *p < 0.05; **p < 0.01. 230 Table S5.8. Tobit models for hypothesi zed instrumental variables. Variable Model (1) Model (2) Model (3) Model (4) Intercept 3.533* (1.568)2.783 (1.621)3.328* (1.388) 3.341 * (1.379)Distance between each household and its monitored parcel 0.351* (0.160)0.130 (0.293)Œ Œ Quadratic term of distance to monitoring parcel Œ0.020 (0.021)Œ Œ NFCP payment ŒŒ 0.019* (0.008) 0.019 * (0.008)Quadratic term of NFCP payments ŒŒŒ Œ 8.250 E-06 (1.330 E-05)Social ties to local leaders (binary) Œ 4.365*(1.857)Œ 4.206*(1.890)Œ 4.898* (1.875) Œ 4.936**(1.875)Distance between each household and the main road 4.143** (1.439)4.089 ** (1.430)3.300 * (1.350) 3.292 * (1.347)Number of laborers 0.602 (0.830)0.557 (0.814)0.386 (0.794) 0.370 (0.797) Household size Œ 1.003 (0.678)Œ 1.034 (0.689)Œ 0.999 (0.682) Œ 0.992 (0.681) Education of adults 0.313 (0.388)0.273 (0.388)0.253 (0.398) 0.249 (0.398) Household income (log) Œ 0.064 (1.101)Œ 0.210 (1.111)Œ 0.022 (1.060) Œ 0.059 (1.062) Percentage of agricultural income Œ 4.398 (3.031)Œ 4.768 (3.031)Œ 4.904 (3.045) Œ 4.875 (3.034) Sampling weight 0.638 (1.113)0.823 (1.067)0.851 (0.914) 0.896 (0.933) Log pseudolikelihood Œ 395.934Œ 395.530Œ 395.465 Œ 395.381 Pseudo R-Squared 0.0230.0240.024 0.024 Notes: Unit of analysis is the household. Dependent variable is total labor input for monitoring. Total number of observations is 156. The numbe rs of left-censored and right-censored observations are 47 and 14, respectively. Independent variables were mean centered. Numbers within parentheses are robust standard errors. For clusters of households from the same monitoring groups, the sampling weight matrix is applied and standard errors are adjusted. Variance Inflation Factors were tested to be < 5. *p < 0.05; **p < 0.01; *** p < 0.001. 231 Table S5.9. Coefficients of the Tobit model fo r the nonlinear effect of group size on collective action. Variable Description Coefficients (Robust S.E.) Marginal effects Intercept 8.921*** (2.360) Œ Quadratic term of group size The quadratic term of group size Œ 0.128** (0.041) Œ Group size The number of households for monitoring a single forest parcel 1.331** (0.408) 0.767 Social ties to local leaders Binary : 0 for weak social ties; 1 for strong social ties Œ 5.377** (1.920) Œ 3.012 Distance between each household and the main road Euclidean distance between each household and the main road 2.787* (1.216) 1.749 Laborers Number of hous ehold laborers 0.296 (0.792) 0.186 Household size Number of h ousehold members Œ 0.741 (0.630) Œ 0.465 Education Average education of adult household members 0.309 (0.369) 0.194 Household income (log) Total household income in 2007 Œ 0.011 (1.042) Œ 0.007 Percentage of agricultural income Percentage of agricultural income to total household income Œ 2.452 (2.760) Œ 1.539 Sampling weight Sampling weight adjusting households sampled from the same monitoring groups Œ 1.432 (1.126) Œ 0.899 Notes: Unit of analysis is the household. Dependent variable is total labor input for monitoring. Log pseudolikelihood is Œ 390.962. Pseudo R-Squared is 0.035.Total number of observations is 156. The numbers of left-censored and right-censored observations are 47 and 14, respectively. Standard errors are adjusted for clusters of households from the same monitoring groups. Independent variables were mean centered before entering the model. Variance Inflation Factors were tested to be < 5 (Table S5.12). *p < 0.05; **p < 0.01; ***p < 0.001. 232 Table S5.10. First-stage regression results of the two-step Tobit model. Dependent variable: Group size Model (1) Model (2) Intercept Œ 4.178*** (0.329) Œ 4.213*** (0.329) Distance between each household and its monitored parcel 0.104** (0.037) Œ NFCP payment Œ 0.007** (0.002) Quadratic term of group size 0.098*** (0.008) 0.929 *** (0.008) Social ties to local leaders (binary) 0.755 (0.479) 0.578 (0.465) Distance between each household and the main road 0.672 (0.319) 0.444 (0.299) Number of laborers 0.200 (0.213) 0.116 (0.206) Household size Œ 0.161*(0.164) Œ 0.169 * (0.159) Education of adults Œ 0.012 (0.071) 0.035 (0.069) Household income (log) Œ 0.027 (0.248) Œ 0.027 (0.241) Percentage of agricultural income Œ 1.285 (0.651) Œ 1.585ƒ (0.640) Sampling weight 1.691*** (0.228) 1.800 *** (0.220) F-statistic 47.69*** 51.34 *** Wald test of exogeneity 2(1) 1.94 1.74 R-squared 0.767 0.780 Adj. R-Squared 0.751 0.765 Notes: Unit of analysis is the household. Dependent variable is group size. Log pseudolikelihood is Œ 399.544. Total number of observations is 156. The numbers of left- censored and right-censored observations are 46 and 14, respectively. Independent variables were mean centered before entering the model. Numbers in parentheses are standard errors. For clusters of households from the same m onitoring groups, the sampling weight matrix is applied and standard errors are adjusted. Variance Inflation Factors were tested to be < 5. ƒp < 0.1; *p < 0.05; **p < 0.01; *** p < 0.001. 233 Table S5.11. Second-stage regression results of the two-step Tobit model. Dependent variable: Monitoring efforts Model (1) Model (2) Intercept 17.267* (7.075) 14.650 ** (5.061) Quadratic term of group size Œ 0.324ƒ (0.166) Œ 0.263 * (0.119) Group size 3.265* (1.602) 2.667 * (1.132) Social ties to local leaders (binary) Œ 6.648** (2.491) Œ 6.235 ** (2.263) Distance between each household and the main road 1.958 (1.570) 2.195 (1.431) Number of laborers 0.003 (1.008) 0.108 (1.052) Household size Œ 0.484 (0.783) Œ 0.556 (0.732) Education of adults 0.367 (0.336) 0.355 (0.319) Household income (log) Œ 0.027 (1.141) Œ 0.010 (1.082) Percentage of agricultural income Œ 9.372 (3.461) Œ 1.001 (3.107) Sampling weight Œ 4.850 (3.017) Œ 3.779ƒ (2.221) Notes: Models (1) and (2) used distance to monitored parcel and NFCP payment as an instrument, respectively. Unit of analysis is th e household. Dependent variable is total labor input for monitoring. Total number of observati ons is 156. The numbers of left-censored and right-censored observations are 46 and 14, respectively. Independent variables were mean centered before entering the model. Numbers within parentheses are standard errors. For clusters of households from the same monitoring groups, the sampling weight matrix is applied and standard errors are adjusted. Variance In flation Factors were tested to be < 5. ƒp < 0.1; *p < 0.05; **p < 0.01; *** p < 0.001. Table S5.12. Variance inflation factors for variables used in the Tobit model examining the nonlinear group-size effect (Table 5.1 or Table S5.9). Variable Variance Inflation Factor Group size 2.86 Quadratic term of group size 2.90 Social ties to local leaders 1.07 Distance between each household and the main road 1.24 Laborers 1.98 Household size 1.70 Education 1.21 Household income (log) 1.39 Percentage of agricultural income 1.25 Notes: VIFs should be < 5. Independent variables were mean centered before entering the model. 234 Table S5.13. Different combinations of Tobit models for the nonlinear effect of group size on collective action. Variable Model (1) Model (2) Model (3) Intercept 8.452*** (2.318)8.918 *** (2.361) 9.074 *** (2.353)Quadratic term of group size Œ 0.122** (0.041)Œ 0.128 ** (0.041) Œ 0.128**(0.042)Group size 1.283** (0.401)1.332 ** (0.408) 1.369 ** (0.414)Social ties to local leaders (binary) Œ 5.349** (1.864)Œ 5.372 ** (1.925) Œ 5.450**(1.907)Distance between each household and the main road 2.986* (1.208)2.785 * (1.218) 2.311 * (1.145)Laborers 0.227 (0.804)0.304 (0.792) 0.282 (0.795) Household size Œ 0.729 (0.637)Œ 0.744 (0.652) Œ 0.706 (0.643) Education 0.293 (0.358)0.310 (0.368) 0.326 (0.366) Percentage of agricultural income Œ 4.660 (3.993)Œ 2.449 (2.759) Œ Agricultural income (log) Œ0.283 (0.403)Œ Œ 0.006 (0.278) Income per capita (log) ŒŒ 0.015 (1.041) Œ 0.051 (1.067) Sampling weight ŒŒ 1.429 (1.123) Œ 1.542 (1.064) Log pseudolikelihood Œ 390.62Œ 390.962 Œ 391.364 Pseudo R-squared 0.0360.035 0.034 Notes: Unit of analysis is the household. Dependent variable is total labor input for monitoring. Total number of observations is 156. The numbe rs of left-censored and right-censored observations are 47 and 14, respectively. For clusters of households from the same monitoring groups, the sampling weight matrix is applied an d standard errors are adjusted. Independent variables were mean centered before entering the model. Numbers in parentheses are robust standard errors. Variance Inflati on Factors were tested to be < 5. *p < 0.05; **p < 0.01; *** p < 0.001. 235 Table S5.14. Coefficients of the spatial autoregressive error model for the linear effect of group size. Variable Coefficients (S.E.) Intercept 0.132*** (0.014) Group size 0.002 (0.003) Parcel size Œ 0.010 (0.009) Parcel size per household Œ 0.001 (0.038) Elevation 0.050 (0.037) Slope 0.340** (0.123) Wetness 0.0491*** (0.012) Distance between each parcel and the nearest household 2.828E-04 (0.004) Distance between each parcel and the main road Œ 0.002 (0.005) Initial forest cover in 2001 Œ 0.263*** (0.030) Moran™s I 0.023 Notes: Unit of analysis is the parcel. Dependent variable is the percent of forest-cover change from 2001 to 2007. Total number of observati ons is 151. Log likelihood is 167.942. Variance Inflation Factors were tested to be < 5. *p < 0.05; **p < 0.01; *** p < 0.001. 236 Table S5.15. Coefficients of the multiple linear re gression for the nonlinear effect of group size. Variable Coefficients (Robust S.E.) Intercept 0.161** (0.011) Group size 0.010* (0.005) Quadratic term of group size Œ 0.002** (0.001) Parcel size Œ 0.014 (0.010) Parcel size per household 0.038 (0.042) Elevation 0.023 (0.040) Slope 0.185 (0.110) Wetness 0.035* (0.021) Distance between each parcel and the nearest household 0.002 (0.002) Distance between each parcel and the main road Œ 0.009* (0.004) Initial forest cover in 2001 Œ 0.226** (0.032) Moran™s I 0.355** Notes: Unit of analysis is the parcel. Dependent variable is the percent of forest-cover change from 2001 to 2007. Total number of observations is 151. R-squared is 0.496. Adjusted R-squared is 0.460. Numbers in parentheses are robust sta ndard errors. Variance Inflation Factors were tested to be < 5. The Moran™s I for residuals is significant, indicating the multiple linear regression is inappropriate and the spatial autocorrelati on should be considered. *p < 0.05; **p < 0.001. 237 Table S5.16. Coefficients of the spatial autoregressive error model for the nonlinear effect of group size on resource outcomes. Variable Description Coefficients (S.E.) Intercept 0.146*** (0.015)Quadratic term of group size The quadratic term of group size (household squared) Œ1.056E-03* (4.800E-04)Group size The number of households for monitoring a single forest parcel (household) 7.205E-03* (3.643E-03)Parcel size Area of each parcel (100 ha) Œ 0.013 (0.009) Parcel size per household Ratio of parcel size to group size (100 ha per household) 0.026 (0.040) Elevation Average elevation of each parcel (1,000 m, above sea level) 0.034 (0.037) Slope Average slope of each parcel (radian) 0.339** (0.121)Wetness Compound Topographic Index as a measurement of wetness of each parcel (unitless; (Gessl er et al., 1995)) 0.048*** (0.012)Distance between each parcel and the nearest household Euclidean distance from each parcel to the nearest household location (km) 4.402E-04 (0.003) Distance between each parcel and the main road Euclidean distance from each parcel to the main road (km) Œ 0.004 (0.005) Initial forest cover in 2001 Averag e forest cover of each parcel in 2001 Œ 0.269*** (0.030) Spatial error correlation coefficient 0.561*** Moran™s I Moran™s I test of spatial autocorrelation for model residuals 0.021 Notes: Unit of analysis is the parcel. Dependent variable is the percent of forest-cover change from 2001 to 2007. Log likelihood is 170.281. Total num ber of observations is 151. Independent variables were mean centered before entering the model. Variance Inflatio n Factors were tested to be < 5 (Table S5.17). *p < 0.05; **p < 0.01; *** p < 0.001. 238 Table S5.17. Variance inflation factors (VIFs) for variables used in the spatial simultaneous autoregressive error model (T able 5.1 or Table S5.16). Variable Variance Inflation Factor Group size 4.01 Quadratic term of group size 1.89 Parcel size 4.07 Parcel size per household 3.09 Elevation 3.54 Slope 1.44 Wetness 1.18 Distance between each parcel and the nearest household 2.09 Distance between each parcel and the main road 3.51 Initial forest cover in 2001 1.42 Notes: VIFs were tested to be < 5. Independent variables were mean centered before entering the model. 239 APPENDIX E SUPPORTING INFORMATION FOR CHAPTER 6 Centered polynomial regression is a co mmon technique to reduce the collinearity between components and the product terms that repr esent interaction or nonlinear effects (Aiken and West, 1991; Dalal and Zickar, 2012). Often mean -centering is used, it does not change the fit of regression models, does not a ffect the power to detect interaction or nonlinear effects, does not alter the reliability of product terms, and doe s not change the highest order coefficients. But it does change the coefficients of lower order terms (and thus the best-fit values at peaks and valleys of the curve), standard errors, and confidence intervals, as well as improve the interpretation. Here we provided a brief proof to clarify the changes of coefficients and improvements in interpretation. For a detailed discussion, please see cited references (Aiken and West, 1991; Dalal and Zickar, 2012). The general form of uncentered polynomia l regression with first order terms and multiplicative interaction term is given as (Aiken and West , 1991; Dalal and Zickar, 2012): Y = b0 + b1X1 + b2X2 + b 3X1X2 + (6.2.1) where Y is the dependent variable; X1 and X2 are independent variables; X1X2 is the multiplicative interaction term; b0 is the intercept; b1 and b2 are coefficients to be estimated; is the error term that has a normal di stribution with mean of zero. When X1 = X2, the multiplicative interaction term become s the quadratic term. 240 After mean-centering X1 and X2 in Equation 6.2.1, the equation of centered polynomial regression is given as: 221132221110XXXXbXXbXXbbY (6.2.2) where 1Xand 2X are the means of X1 and X2, respectively; 2211XXXX is the multiplicative interaction term of X1 and X2 after mean-centering. Equation 6.2.2 can be rearranged a nd factored into Equation 6.2.3: 2132132123121322110XXbXXbbXXbbXXbXbXbbY (6.2.3) Comparing Equation 6.2.1 and 6.2.3, it is easy to see that the coefficient of the interaction term ( X1X2) does not change; however, the intercept and coefficients for first order terms ( X1 and X2) do alter in a predictable way. From Equati on 6.2.2, it is also easy to see that the interpretation of marginal e ffects [also called conditional or simple effects for polynomial regressions(Dalal and Zickar, 2012)] become easy since b1 and b2 are actually the marginal effects of X1 and X2 at the mean values of X2 and X1, respectively. 241 Table S6.1. Descriptive statistics of variables used in modeling the effects of conservation and development policies on changes in total household income and income structure. Variables Description Mean (S.D.) Dependent variable for the model in Table 6.1 Changes in total household income The difference of total household income in 2007 subtracting total household income in 1998 ( thousand yuan) 21.988 (27.286) Dependent variable for the model in Table 6.2 Changes in the percentage of agricultural income The difference of percentage of agricultural income in 2007 subtracting the percentage of agricultural income in 1998 Œ0.218 (0.389) Policy intervention NFCP payment The payment each household received annually from the NFCP ( thousand yuan) 0.948 (0.183) NFCP payment percentage NFCP payment divided by total household income in 2007 0.065 (0.060) GTGB payment The payment each household received annually from the cropland conversion policies (i.e., GTGP and GTBP) ( thousand yuan) 2.888 (2.320) GTGB payment percentage GTGB payment divided by total household income in 2007 0.160 (0.153) Tourism participation Household participation in tourism business (1: participated; 0: not participated) 0.274 (0.447) Electricity subsidy Initial subsidy receiv ed for electricity consumption (yuan) 85.982 (104.429) Electricity subsidy percentage NFCP payment divided by total household income in 1998 (›) 0.027 (0.041) 242 Table S6.1 (cont™d) Variables Description Mean (S.D.) Financial capital Initial total household income Total household income in 1998 (thousand yuan) 6.285 (4.932) Initial percentage of agricultural income The percentage of agricultural income in 1998 0.630 (0.313) Changes in total agricultural income The difference of total agricultural income in 2007 subtracting total agricultural income in 1998 (thousand yuan) 7.817 (15.393) Human capital Number of laborers The number of laborers in each household 2.820 (1.455) Changes in number of laborers The difference of number of laborers in 2007 subtracting the number of laborers in 1998 Œ0.727 (1.795) Education Education level of the most educated non-student adult in 2007 (year) 7.120 (3.432) Natural capital Cropland area The total area of cropland for each household in 2007 (Mu, 1 Mu = 1/15 ha) 10.450 (4.163) Built-up capital Distance to the main road The Euclidean distance from each household location to the main road (km) 0.431 (0.629) Social capital Social ties to local governments Whether the household has a member of immediate relative working in local governments or government enterprises: 1. Yes; 0: No. 0.120 (0.326) Notes: All monetary values in 1998 are discounted into present values in 2007. The number of observations is 179. 243 Table S6.2. Complementary resu lts of the effects of conserva tion and development policies on changes in total household income. Variables Unstandardized coefficients Standardized coefficients Robust S.E. Policy intervention NFCP payment Œ4.816 Œ0.021 19.066 GTGB payment 1.925** 0.163 ** 0.694 Initial electricity subsidy 44.577* 0.171 * 22.335 Tourism participation 11.905** 0.195 ** 3.934 NFCP payment × GTGB payment 3.985 0.035 11.219 NFCP payment × Initial electricity subsidy Œ0.050 Œ0.016 0.144 NFCP payment × Tourism participation Œ116.000* Œ0.134 * 48.657 GTGB payment × Tourism participation 1.505 0.083 2.265 GTGB payment × In itial electricity subsidy 0.005 0.038 0.009 Financial capital Initial total household income Œ0.401 Œ0.072 0.326 Changes in total agricultural income 1.253*** 0.712 *** 0.241 Human capital Number of laborers 3.295* 0.177 * 1.513 Changes in number of laborers 2.610* 0.173 * 1.116 Education 0.266 0.033 0.431 Natural capital Cropland area Œ0.313 Œ0.028 0.646 244 Table S6.2 (cont™d) Variables Unstandardized coefficients Standardized coefficients Robust S.E. Built-up capital Distance to the main road, log Œ1.234* Œ0.091 * 0.602 Social capital Social ties to local governments 2.354 0.028 4.617 Township 1: Wolong; 0: Gengda 6.240* 0.114 * 3.045 Constant 14.330*** Š 3.202 Notes: All continuous independent variables were mean-centered before entering the model. The number of observations is 179. The R-squared is 0.691. Variation Inflation Factors (VIFs) were tested to be less than 5. ƒ P < 0.1; * p < 0.05; ** P < 0.01; *** p < 0.001. 245 Table S6.3. Complementary result s of the effects of conserva tion and development policies on changes in the percentage of agricultural income. Variables Unstandardized coefficients Standardized coefficients Robust S.E. Policy intervention NFCP payment Œ0.068 Œ0.020 0.196 GTGB payment Œ0.037*** Œ0.221 *** 0.009 NFCP payment × GTGB payment Œ0.088 Œ0.054 0.098 Initial electricity subsidy Œ3.262e-04* Œ0.088 * 1.355e-04 Tourism participation Œ0.096* Œ0.110 * 0.047 NFCP payment × Initial electricity subsidy Œ8.83e-04 Œ0.020 0.002 NFCP payment × Tourism participation 1.526* 0.123 * 0.752 GTGB payment × Initial electricity subsidy 6.07e-05 0.032 4.70e-05 GTGB payment × Tourism participation 0.009 0.036 0.014 Financial capital Initial total household income 0.007* 0.088 * 0.004 Initial agricultural income percentage Œ0.862*** Œ0.687 *** 0.057 Changes in total agricultural income 0.007** 0.277 ** 0.002 Human capital Number of laborers 0.026 0.099 0.022 Changes in number of laborers 0.003 0.013 0.015 Education Œ0.007 Œ0.060 0.005 Natural capital Cropland area 0.023** 0.147 *** 0.008 246 Table S6.3 (cont™d) Variables Unstandardized coefficients Standardized coefficients Robust S.E. Built-up capital Distance to the main road, log 0.017ƒ 0.086 ƒ 0.009 Social capital Social tie s to local governments Œ0.118* Œ0.098 * 0.050 Township 1: Wolong; 0: Gengda Œ0.055 Œ0.071 0.034 Constant Œ0.116** Š 0.043 Notes: All continuous independent variables were mean-centered before entering the model. The number of observations is 179. The R-squared is 0.756. Variation Inflation Factors (VIFs) were tested to be less than 5. ƒ P < 0.1; * p < 0.05; ** P < 0.01; *** p < 0.001. 247 REFERENCES 248 REFERENCES Abdallah, S., Thompson, S., Marks, N., 2008. Estim ating worldwide life satisfaction. Ecol. Econ. 65, 35-47. Adams, W.M., Aveling, R., Brockington, D., Dickson, B., Elliott, J., Hutton, J., Roe, D., Vira, B., Wolmer, W., 2004. Biodiversity conservation and the eradication of poverty. Science 306, 1146-1149. Adhikari, K.P., Goldey, P., 2010. Social Capital and its "Downside": The Impact on Sustainability of Induced Community-Based Organizations in Nepal. World Devel. 38, 184-194. Agrawal, A., 2000. Small is Beautiful, but is Larg er Better? Forest Management Institutions in the Kumaon Himalaya, India, in: Gibson, C., McKean, M.A., Ostrom, E. (Eds.), People and Forests: Communities, Institutions, and Governance. MIT Press, Cambridge, pp. 57-85. Agrawal, A., Chhatre, A., 2006. Explaining success on the commons: Community forest governance in the Indian Himala ya. World Devel. 34, 149-166. Agrawal, A., Goyal, S., 2001. Group size and collective action - Third-party monitoring in common-pool resources. Comp Polit Stud 34, 63-93. Aiken, L.S., West, S.G., 1991. Multiple regression: Testing and interpreting interactions. Sage Publications, Incorporated. Alix-Garcia, J., De Janvry, A., Sadoulet, E., 2008. The role of deforestation risk and calibrated compensation in designing payments for environmental services. Environ. Devel. Econ. 13, 375- 394. An, L., He, G., Liang, Z., Liu, J., 2006. Impact s of demographic and socioeconomic factors on spatio-temporal dynamics of panda habita t. Biodivers. Conserv. 15, 2343-2363. An, L., Liu, J., Ouyang, Z., Linderman, M., Zhou, S., Zhang, H., 2001. Simulating demographic and socioeconomic processes on hou sehold level and implications fo r giant panda habitats. Ecol. Model. 140, 31-49. An, L., Lupi, F., Liu, J., Linderman, M.A., Hua ng, J., 2002. Modeling the choice to switch from fuelwood to electricity: Implications for giant panda habitat conservation. Ecol. Econ. 42, 445-457. An, L., Mertig, A.G., Liu, J., 2003. Adolescents leaving parental home: Psychosocial correlates and implications for conservation. Population and Environment 24, 415-444. Anselin, L., 1988. Spatial Econometrics: Methods and Models. Kluwer Academic Publishers Norwell, MA. 249 Anselin, L., Syabri, I., Kho, Y., 2006. GeoDa: An introduction to spatial data analysis. Geographical Analysis 38, 5-22. Anthony, D.L., Campbell, J.L., 2011. States, so cial capital and cooperation: looking back on Governing the Commons. Internationa l Journal of the Commons 5, 284-302. Arriagada, R.A., Sills, E.O., Pattanayak, S.K ., Ferraro, P.J., 2009. Combining qualitative and quantitative methods to evaluate participati on in Costa Rica's Program of payments for environmental services. Journal of Sustainable Forestry 28, 343-367. Axinn, W.G., Pearce, L.D., Ghimire, D., 1999. I nnovations in life history calendar applications. Soc. Sci. Res. 28, 243-264. Baland, J.-M., Platteau, J.-P., 1996. Halting degrada tion of natural resources : is there a role for rural communities? Food and Agriculture Or ganization of the United Nations, Oxford. Baland, J.M., Platteau, J.P., 1999. The ambiguous impact of inequality on local resource management. World Devel. 27, 773-788. Barbier, E.B., 2011. Pricing Nature. Annual Review of Resource Economics 3, 337-353. Barrett, C.B., Travis, A.J., Dasgupta, P., 2011. On biodiversity conservation and poverty traps. Proc. Natl. Acad. Sci. U. S. A. 108, 13907-13912. Bateman, I.J., Mace, G.M., Fezzi, C., Atkinson, G., Turner, K., 2011. Economic Analysis for Ecosystem Service Assessments. Environ. Resour. Econ. 48, 177-218. Bearer, S., Linderman, M., Huang, J.Y., An, L., He, G.M., Liu, J.Q., 2008. Effects of fuelwood collection and timber harvesting on giant pand a habitat use. Biol. Conserv. 141, 385-393. Berke, P.R., Campanella, T.J., 2006. Planning for postdisaster resiliency. Ann. Am. Acad. Pol. Soc. Sci. 604, 192-207. Berkes, F., 2007. Community-based conservation in a globalized world. Proceedings of the National Academy of Sciences 104, 15188-15193. Berry, P.A.M., Garlick, J.D., Smith, R.G., 2007. Near-global validation of the SRTM DEM using satellite radar altimetry. Remote Sens. Environ. 106, 17-27. Birch, J.C., Newton, A.C., Aquino, C.A., Cant arello, E., Echeverria, C., Kitzberger, T., Schiappacasse, I., Garavito, N.T., 2010. Cost-effectiveness of dryland forest restoration evaluated by spatial analysis of ecosystem services. Proc. Natl. Acad. Sci. USA 107, 21925-21930. Blalock, H.M., 1965. Theory Building and the Stat istical Concept of Interaction. Am. Sociol. Rev. 30, 374-380. 250 Boardman, A.E., Greenberg, D.H., Vining, A.R ., Weimer, D.L., 2006. Cost-Benefit Analysis: Concepts and Practice, 3rd ed. Prentice Hall, Upper Saddle River, New Jersey. Bollen, K.A., Noble, M.D., 2011. Structural equation models and the quan tification of behavior. Proc. Natl. Acad. Sci. U. S. A. 108, 15639-15646. Bouma, J., Bulte, E., van Soest, D., 2008. Trust and cooperation: Social capital and community resource management. J. Environ. Econ. Manage. 56, 155-166. Boyd, J., Banzhaf, S., 2007. What are ecosyst em services? The need for standardized environmental accounting units. Ecol. Econ. 63, 616-626. Boyd, R., Gintis, H., Bowles, S., 2010. Coordi nated Punishment of Defectors Sustains Cooperation and Can Proliferate When Rare. Science 328, 617-620. Brown, T.A., 2006. Confirmatory factor analysis for applied research. The Guilford Press, New York, NY. Butler, C.D., Oluoch-Kosura, W., 2006. Linking fu ture ecosystem services and future human well-being. Ecol. Soc. 11, 30. Cai, J., Jiang, Z.G., Zeng, Y., Li, C.W., Brav ery, B.D., 2008. Factors affecting crop damage by wild boar and methods of mitigation in a gian t panda reserve. Eur J Wildl Res 54, 723-728. Camfield, L., Skevington, S.M., 2008. On subjectiv e well-being and quality of life. Journal of Health Psychology 13, 764-775. Campbell, A., 1976. Subjective Measures of Well-Being. Am. Psychol. 31, 117-124. Carpenter, J.P., 2007. Punishing free-riders: How group size affects mutual monitoring and the provision of public goods. Games Econ. Behav. 60, 31-51. Carpenter, S.R., DeFries, R., Dietz, T., Mooney, H.A., Polasky, S., Reid, W.V., Scholes, R.J., 2006. Millennium Ecosystem Assessment: Research Needs. Science 314, 257-258. Carpenter, S.R., Mooney, H.A., Agard, J., Capist rano, D., DeFries, R.S., Diaz, S., Dietz, T., Duraiappah, A.K., Oteng-Yeboah, A., Pereira, H.M., Perrings, C., Reid, W.V., Sarukhan, J., Scholes, R.J., Whyte, A., 2009. Science fo r managing ecosystem services: Beyond the Millennium Ecosystem Assessment. Proc. Natl. Acad. Sci. U. S. A. 106, 1305-1312. Chamberlin, J.R., 1974. Provision of Collective G oods as a Function of Group Size. Amer. Polit. Sci. Rev. 68, 707-716. Chang, J., Wu, X., Liu, A.Q., Wang, Y., Xu, B., Yang, W., Meyerson, L.A., Gu, B.J., Peng, C.H., Ge, Y., 2011. Assessment of net ecosystem servi ces of plastic greenhouse vegetable cultivation in China. Ecol. Econ. 70, 740-748. 251 Chee, Y.E., 2004. An ecological perspective on the valuation of ecosystem services. Biol. Conserv. 120, 549-565. Chen, X., Frank, K.A., Dietz, T., Liu, J., 2012a. Weak ties, labor Migration, and environmental impacts: Toward a sociology of sustainability. Organ. Environ. 25, 3-24. Chen, X., Lupi, F., An, L., Sheely, R., Viña, A., Liu, J., 2012b. Agent-based modeling of the effects of social norms on enrollment in paym ents for ecosystem services. Ecological Modeling 229, 16-24. Chen, X., Lupi, F., He, G., Liu, J., 2009. Linking social norms to efficient conservation investment in payments for ecosystem servi ces. Proc. Natl. Acad. Sci. USA 106, 11812-11817. Chen, X., Lupi, F., Vina, A., He, G., Liu, J., 2010. Using Cost-Effective Targeting to Enhance the Efficiency of Conservation I nvestments in Payments for Ecosystem Services. Conserv. Biol. 24, 1469-1478. Chhatre, A., Agrawal, A., 2008. Forest commons and local enforcement. Proc. Natl. Acad. Sci. USA 105, 13286-13291. Colby, M.E., 1991. Environmental management in development: the evolution of paradigms. Ecol. Econ. 3, 193-213. Collins, S.L., Carpenter, S.R., Swinton, S.M., Orenstein, D.E., Childers, D.L., Gragson, T.L., Grimm, N.B., Grove, M., Harlan, S.L., Kaye, J.P ., Knapp, A.K., Kofinas, G.P., Magnuson, J.J., McDowell, W.H., Melack, J.M., Ogden, L.A., Robertson, G.P., Smith, M.D., Whitmer, A.C., 2011. An integrated conceptual framework for l ong-term social-ecologica l research. Front. Ecol. Environ. 9, 351-357. Costanza, R., d'Arge, R., de Groot, R., Farber, S., Grasso, M., Hannon, B., Limburg, K., Naeem, S., O'Neill, R.V., Paruelo, J., Raskin, R.G., Sutton, P., van den Belt, M., 1997. The value of the world's ecosystem services and na tural capital. Nature 387, 253-260. Daily, G.C., 1997. Nature's Services: Societal Dependence on natural Ecosystems. Island Press, Washington, D.C. Daily, G.C., Matson, P.A., 2008. Ecosystem services : From theory to implementation. Proc. Natl. Acad. Sci. U. S. A. 105, 9455-9456. Dalal, D.K., Zickar, M.J., 2012. Some Common Myths About Cent ering Predictor Variables in Moderated Multiple Regression and Polynomial Regression. Organ Res Methods 15, 339-362. Daniel, C., Wood, F.S., Gorman, J.W., 1999. Fitti ng Equations to Data: Computer Analysis of Multifactor Data. Wiley. Daniels, A.E., Bagstad, K., Esposito, V., Moulaert, A., Rodriguez, C.M., 2010. Understanding the impacts of Costa Rica's PES: Are we aski ng the right questions? Ecol. Econ. 69, 2116-2126. 252 De Rus, G., 2010. Introduction to Cost-benefit Analysis: Looking for Reasonable Shortcuts. Edward Elgar Pub. Diener, E., 1994. Assessing Subjective Well-Being - Progres s and Opportunities. Soc. Indic. Res. 31, 103-157. Diener, E., 2000. Subjective well-being Š The science of happiness and a proposal for a national index. Am. Psychol. 55, 34-43. Diener, E., 2009. The science of well-being: The collected works of Ed Diener. Springer, The Netherlands. Diener, E., Ryan, K., 2009. Subjective well-being: a general overview. S outh African Journal of Psychology 39, 391-406. Diener, E., Suh, E.M., Lucas, R.E., Smith, H.L., 1999. Subjective well-being: Three decades of progress. Psychol. Bull. 125, 276-302. Dietz, T., Henry, A.D., 2008. Context and the commons. Proc. Natl. Acad. Sci. USA 105, 13189- 13190. Dietz, T., Rosa, E.A., York, R., 2009. Environmentally Efficient Well-Being: Rethinking Sustainability as the Relations hip between Human Well-being and Environmental Impacts. Hum. Ecol. Rev. 16, 114-123. Dietz, T., Rosa, E.A., York, R., 2012. Environmentally efficient well-being: Is there a Kuznets curve? Appl Geogr 32, 21-28. Eagly, A.H., Kulesa, P., 1997. Attitudes, attitude stru cture, and resistance to change: implications for persuasion on environmental issues, in: Baze rman, M.H., Messick, D.M., Tenbrunsel, A.E., Wade-Benzoni, K.A. (Eds.), Environmental ethics and behavior. The New Lexington Press, San Francisco, CA, pp. 122-153. Ekin, P., 2008. The four-capital method of su stainable development evaluation. European Environment 18, 63-80. Elliott, J.R., Pais, J., 2006. Race, class, and Hurricane Katrina: Social differences in human responses to disaster. Soc. Sci. Res. 35, 295-321. Engel, S., Pagiola, S., Wunder, S., 2008. Designing payments for environmental services in theory and practice: An overview of the issues. Ecol. Econ. 65, 663-674. Escobar, A., 1999. After Nature: Steps to an Anti essentialist Political Ecology. Curr Anthrop 40, 1-30. Esteban, J., Ray, D., 2001. Collective action and the group size paradox. Amer. Polit. Sci. Rev. 95, 663-672. 253 Evenson, R.E., Mwabu, G., 1996. Household composition and expenditures on human capital formation in Kenya. Res Popul Econ 8, 205-232. Farber, S.C., Costanza, R., Wilson, M.A., 2002. Economic and ecological concepts for valuing ecosystem services. Ecol. Econ. 41, 375-392. Fehr, E., Gachter, S., 2002. Altruistic punishment in humans. Nature 415, 137-140. Fischbacher, U., Gachter, S., Fehr, E., 2001. Are people conditionally cooperative? Evidence from a public goods experiment. Econ. Letters 71, 397-404. Fisher, B., Turner, R.K., Morling, P., 2009. Defining and classifying ecosystem services for decision making. Ecol. Econ. 68, 643-653. Foley, J.A., DeFries, R., Asner, G.P., Barford, C., Bonan, G., Carpenter, S.R., Chapin, F.S., Coe, M.T., Daily, G.C., Gibbs, H.K., Helkowski, J.H., Holloway, T., Howard, E.A., Kucharik, C.J., Monfreda, C., Patz, J.A., Prentice, I.C., Ramankutty, N., Snyder, P.K., 2005. Global Consequences of Land Use. Science 309, 570-574. FONAFIFO, CONAFOR, Ministry of Environment, 2012. Less ons learned for REDD+ from PES and conservation incentive programs. Exampl es from Costa Rica, Mexico, and Ecuador. Costa Rica, Ecuador, Mexico, World Bank, Forest Trends and PROFOR, p. 164. Forestry Department of Food and Agriculture Organization of the United Nations, 2010. Global Forest Resources Assessment 2010: Country Report - China. Forestry Department of Food and Agriculture Organization of the United Nations, Rome. Fox, J., 1997. Applied Regression Analysis, Linear Models and Related Methods. SAGE Publications, Inc., Thousand Oaks. Freeman, D., Thornton, A., Camburn, D., Al win, D., Yooung-DeMacro, L., 1988. The life history calendar: A technique for collecting retrospective data, in: Clogg, C. (Ed.), Soc. Method. American Sociological Associati on, Washington, DC., pp. 37-68. Friedrich, R.J., 1982. In Defense of Multiplica tive Terms in Multiple-Regression Equations. Am. J. Polit. Sci. 26, 797-833. Gallopin, G.C., 2006. Linkages between vulnerabili ty, resilience, and adaptive capacity. Global Environ Chang 16, 293-303. Gautam, A.P., 2007. Group size, heterogeneity a nd collective action outcomes: Evidence from community forestry in Nepal. Int J Sustain Dev World Ecol 14, 574-583. Gefen, D., Straub, D.W., Boudreau, M.-C., 2000. Structural Equation Modeling and Regression: Guidelines for Research Practice Communications of the Association for Information Systems 4, 1-79. 254 Gessler, P.E., Moore, I.D., McKenzie, N.J., Ryan, P.J., 1995. Soil-landscape modeling and spatial prediction of soil attributes. Int J Geogr Inf Sci 9, 421-432. Gibson, C.C., Williams, J.T., Ostrom, E., 2005. Local enforcement and better forests. World Devel. 33, 273-284. Grant, S., Langan-Fox, J., Anglim, J., 2009. The Bi g Five Traits as Predictors of Subjective and Psychological Well-Being. Psychol. Rep. 105, 205-231. Grigg, A., 2008. Dependency and impact on ecosyst em services Œ unmanaged risk, unrealised opportunity: A briefing document for the food, be verage and tobacco sectors. Fauna & Flora International. Gross-Camp, N.D., Martin, A., McGuire, S., Ke bede, B., Munyarukaza, J., 2012. Payments for ecosystem services in an African protected area: exploring issues of legitimacy, fairness, equity and effectiveness. Oryx 46, 24-33. Gunderson, L., 2010. Ecological and Human Comm unity Resilience in Response to Natural Disasters. Ecol. Soc. 15, 18. Haan, M., Kooreman, P., 2002. Free riding and the provision of candy bars. J. Public Econ. 83, 277-291. Hanley, N., MacMillan, D., Wright, R.E., Bullock, C., Simpson, I., Parsisson, D., Crabtree, B., 1998. Contingent valuation versus choice experiments: Es timating the benefits of environmentally sensitive areas in Scotland. J. Agr. Econ. 49, 1-15. Hanley, N., Shogren, J.F., White, B., 2001. Intr oduction to Environmental Economics. Oxford University Press, New York. Hardin, G., 1968. The Tragedy of the Commons. Science 162, 1243-1248. He, G., Chen, X., Bearer, S., Colunga, M., Mertig, A., An, L., Zhou, S., Linderman, M., Ouyang, Z., Gage, S., Li, S., Liu, J., 2009. Spatial and temporal patterns of fuel collection in Wolong Nature Reserve: Implications for panda cons ervation. Landscape Urban Plann. 92, 1-9. He, G., Chen, X., Liu, W., Bearer, S., Zhou, S., Cheng, L., Zhang, H., Ouyang, Z., Liu, J., 2008. Distribution of Economic Benef its from Ecotourism: A Case St udy of Wolong Nature Reserve for Giant Pandas in China. Environ. Manage. 42, 1017-1025. Henrich, J., Gil-White, F.J., 2001. The evolution of prestige - Freely conferred deference as a mechanism for enhancing the benefits of cultu ral transmission. Evol Hum Behav 22, 165-196. Houck, O.A., 1985. Rising water: the national flood insurance program and Louisiana. Tul. L. Rev. 60, 61-164. 255 Hughes, R., Flintan, F., 2001. Integrating Conservation and Development Experience: A Review and Bibliography of the ICDP Literature. International Institute for Environment and Development, London. Hull, V., Xu, W.H., Liu, W., Zhou, S.Q., Vina, A., Zhang, J.D., Tuanmu, M.N., Huang, J.Y., Linderman, M., Chen, X.D., Huang, Y., Ouyang, Z.Y., Zhang, H.M., Liu, J.G., 2011. Evaluating the efficacy of zoning designati ons for protected area manage ment. Biol. Conserv. 144, 3028- 3037. Hwang, S.-H., 2011. Larger groups may allevi ate collective action problems [Version: 2011.10.11]. Research Institute for Market Economy, Sogang University, Seoul, Korea. IPCC, 2012. Managing the risks of extreme events and disasters to advance climate change adaption, in: Field, C.B., V. Barros, T.F. Stocker,, D. Qin, D.J.D., K.L. Ebi, M.D. Mastrandrea, K.J. Mach, G.-K. Plattner, S.K. Allen,, M. Tignor, a.P.M.M. (Eds.), A Special Report of Working Groups I and II of the Intergovernme ntal Panel on Climate Change. Cambridge University Press, New York, NY, USA. Isaac, R.M., Walker, J.M., Williams, A.W., 1994. Group Size and the Voluntary Provision of Public Goods: Experimental Evidence Utiliz ing Large Groups. J. Public Econ. 54, 1-36. IUCN, 1980. World Conservation Strategy: Living Resource C onservation for Sustainable Development. International Union for Conserva tion of Nature and Natural Resources, Gland, Switzerland. IUCN, UNEP-WCMC, 2012. The World Database on Protected Areas (WDPA), Februrary 2012 ed. United Nations Environment Programme (UNE P) - World Conservation Monitoring Centre (WCMC), Cambridge, UK. Jack, B.K., Kousky, C., Sims, K.R.E., 2008. Designing payments for ecosystem services: Lessons from previous experience with incentive-based mechanisms. Proc. Natl. Acad. Sci. U. S. A. 105, 9465-9470. Jeanrenaud, S., 2002. People-oriented approaches in global conservation: is the leopard changing its spots? International Institute for Environment and Development, London. Jensen, J.R., 1996. Introductory digital image pr ocessing, a remote sensing perspective, Second ed. Prentice Hall, New Jersey. Jordan, S.J., Hayes, S.E., Yoskowitz, D., Smith, L.M., Summers, J.K., Russell, M., Benson, W.H., 2010. Accounting for natural resources and environmental sustainability: linking ecosystem services to human well-being. Environ. Sci. Technol. 44, 1530-1536. Kahneman, D., Krueger, A.B., 2006. Developments in the measurement of subjective well-being. J. Econ. Perspect. 20, 3-24. 256 Kalacska, M., Sanchez-Azofeifa, G.A., Rivard, B., Calvo-Alvarado, J.C., Quesada, M., 2008. Baseline assessment for environmental services payments from satellite imagery: A case study from Costa Rica and Mexico. J. Environ. Manage. 88, 348-359. Kareiva, P., Tallis, H., Ricketts, T.H., Daily, G.C., Polasky, S., 2011. Natural Capital: Theory and Practice of Mapping Ecosystem Serv ices. Oxford University Press. Kates, R.W., Travis, W.R., Wilbanks, T.J., 2012. Transformational adaptation when incremental adaptations to climate change are insufficient. Proc. Natl. Acad. Sci. U. S. A. 109, 7156-7161. Kinzig, A.P., Perrings, C., Chapin, F.S., Polasky, S., Smith, V.K., Tilman, D., Turner, B.L., 2011. Paying for Ecosystem Services-Promise and Peril. Science 334, 603-604. Kiss, A., 2004. Is community-based ecotourism a good use of biodiversity conservation funds? Trends Ecol. Evol. 19, 232-237. Klein, C.A., Zellmer, S.B., 2007. Mississippi River stories: lessons from a century of unnatural disasters. Southern Methodist University Law Review 60, 1471-1538. LeSage, J., Pace, R.K., 2009. Introduction to Spatial Econometrics. CRC Press, Boca Raton. Li, J., Feldman, M.W., Li, S.Z., Daily, G.C., 2011. Rural household income and inequality under the Sloping Land Conversion Program in western China. Proc. Natl. Acad. Sci. U. S. A. 108, 7721-7726. Li, W., Zhao, X., 1989. China's Nature Reserves. Foreign Language Press, Beijing. Li, Y., Viña, A., Yang, W., Chen, X., Zhang, J., Ouyang, Z., Liu, J., 2013. Effects of conservation policies on forest c over change in panda habitat re gions, China. Land Use Policy 33, 42-53. Lichstein, J.W., Simons, T.R., Shriner, S.A., Franzreb, K.E., 2002. Spatial autocorrelation and autoregressive models in ecology. Ecol. Monogr. 72, 445-463. Lin, N., 2001. Social capital: A th eory of social structure and acti on. Cambridge University Press, New York, NY. Linderman, M., Bearer, S., An, L., Tan, Y.C., Ou yang, Z.Y., Liu, H.G., 2005a. The effects of understory bamboo on broad-scale estimates of gi ant panda habitat. Biol. Conserv. 121, 383-390. Linderman, M.A., An, L., Bearer, S., He, G.M ., Ouyang, Z.Y., Liu, J.G., 2005b. Modeling the spatio-temporal dynamics and interactions of ho useholds, landscapes, and giant panda habitat. Ecol. Model. 183, 47-65. Liu, J., 2010. China's road to sustainability. Science 328, 974-974. Liu, J., Daily, G.C., Ehrlich, P.R., Luck, G.W., 2003a. Effects of household dynamics on resource consumption and biodiversity. Nature 421, 530-533. 257 Liu, J., Diamond, J., 2005. China's environmen t in a globalizing world. Nature 435, 1179-1186. Liu, J., Dietz, T., Carpenter, S.R., Alberti, M., Folke, C., Moran, E., Pell, A.N., Deadman, P., Kratz, T., Lubchenco, J., Ostrom, E., Ouyang, Z., Provencher, W., Redman, C.L., Schneider, S.H., Taylor, W.W., 2007a. Comp lexity of coupled human and natural systems. Science 317, 1513-1516. Liu, J., Dietz, T., Carpenter, S.R., Folke, C., Alberti, M., Redman, C.L., Schneider, S.H., Ostrom, E., Pell, A.N., Lubchenco, J., Taylor, W.W., Ouya ng, Z.Y., Deadman, P., Kratz, T., Provencher, W., 2007b. Coupled human and natural systems. Ambio 36, 639-649. Liu, J., Li, S., Ouyang, Z., Tam, C., Chen, X ., 2008. Ecological and socioeconomic effects of China's policies for ecosystem services. Proc. Natl. Acad. Sci. USA 105, 9477-9482. Liu, J., Linderman, M., Ouyang, Z., An, L., Yang , J., Zhang, H., 2001. Ecological degradation in protected areas: The case of Wolong Nature Reserve for giant pandas. Science 292, 98-101. Liu, J., Ouyang, Z., Pimm, S., Raven, P., Wa ng, X., Miao, H., Han, N., 2003b. Protecting China's Biodiversity. Science 300, 1240-1241. Liu, J., Ouyang, Z., Tan, Y., Yang, J., Zhou, S., 1999a. Changes in human population structure and implications for biodiversity conserva tion. Population and Environment 21, 45-58. Liu, J., Ouyang, Z., Taylor, W.W., Groop, R ., Tan, Y., Zhang, H., 1999b. A framework for evaluating the effects of human factors on wildlif e habitat: the case of giant pandas. Conserv. Biol. 13, 1360-1370. Liu, J., Ouyang, Z., Yang, W., Xu, W., Li, S., 2013. Evaluation of Ecosystem Service Policies from Biophysical and Social Perspectives:The Case of China, in: Levin, S.A. (Ed.), Encyclopedia of Biodiversity (second edition), ed. Academic Press, Waltham, MA, pp. 372-384. Liu, J., Yang, W., 2012. Water Sustainability for China and Beyond. Science 337, 649-650. Liu, W., Vogt, C.A., Luo, J., He, G., Frank, K.A., Liu, J., 2012. Drivers and Socioeconomic Impacts of Tourism Participation in Protected Areas. PLoS ONE 7, e35420. Locke, H., Dearden, P., 2005. Rethinking protected area categories and the new paradigm. Environ. Conserv. 32, 1-10. Lowe, S.R., Chan, C.S., Rhodes, J.E., 2010. Pre-Hurricane Perceived Social Support Protects Against Psychological Distress: A Longitudinal Analysis of Low-Income Mothers. J. Consult. Clin. Psychol. 78, 551-560. MA, 2005. Ecosystems & Human Well-being: S ynthesis (Millennium Ecosystem Assessment). Island Press, Washington, DC. Maslow, A.H., 1943. A theory of human motivation. Psychol. Rev. 50, 370-396. 258 Masten, A.S., Obradovic, J., 2008. Disaster Prep aration and Recovery: Lessons from Research on Resilience in Human Development. Ecol. Soc. 13. Mathew, S., Boyd, R., 2011. Punishment sustains large-scale cooperation in prestate warfare. Proc. Natl. Acad. Sci. USA 108, 11375-11380. McConnell, W.J., Millington, J.D.A., Reo, N.J., Alberti, M., Asbjornsen, H., Baker, L.A., Brozovi, N., Drinkwater, L.E., Drzyzga, S.A., Fragoso, J., Holland, D.S., Jantz, C.A., Kohler, T.A., Maschner, H.D.G., Monticino, M., Podestá, G., Pontius, R.G., Jr., C.L.R., Sailor, D., Urquhart, G., Liu, J., 2011. Research on Coupled Human and Natural Systems (CHANS): Approach, Challenges, and Strategies. Bull Ecol Soc Am 92, 218-228. McLaughlin, P., 2011. Climate Change, Adaptati on, and Vulnerability: Reconceptualizing Societal-Environment Interaction Within a So cially Constructed Adaptive Landscape. Organ Environ 24, 269-291. McLaughlin, P., Dietz, T., 2008. Structure, agen cy and environment: Toward an integrated perspective on vulnerability. Glob al Environ Chang 18, 99-111. McShane, T.O., Wells, M.P., 2004. Getting Biod iversity Projects to Work: Towards More Effective Conservation and Development. Colu mbia University Press, New York, p. 447. Meguro, T., 2009. Re-conceptualization of wildlife conservation: toward resonation between subsistence and wildlife. African Centre for Technology Studies, Nairobi, Kenya. Milinski, M., Semmann, D., Krambeck, H.J., 2002. Reputation helps solve the 'tragedy of the commons'. Nature 415, 424-426. Millington, J.D.A., Walters, M.B., Matonis, M.S., Liu, J.G., 2013. Modelling for forest management synergies and trade-offs: Northern hardwood tree regeneration, timber and deer. Ecol. Model. 248, 103-112. Muthén, L.K., Muthén, B.O., 1998-2010. Mplus User's Guide, Sixth ed. Muthén & Muthén, Los Angeles, CA. Myers, N., Mittermeier, R. A., Mittermeier, C.G., da Fonseca, G.A.B., Kent, J., 2000. Biodiversity hotspots for conservation priorities. Nature 403, 853-858. Naidoo, R., Ricketts, T.H., 2006. Mapping the economic costs and benefits of conservation. PLoS Biol. 4, 2153-2164. National Bureau of Statistics of China, 2011. Rural Household Survey Instrument [in Chinese]. National Bureau of Statistics of China, Beijing. Nelson, E., Mendoza, G., Regetz, J., Polasky, S ., Tallis, H., Cameron, D.R., Chan, K.M.A., Daily, G.C., Goldstein, J., Kareiva, P.M., Lonsdorf, E., Naidoo, R., Ricketts, T.H., Shaw, M.R., 2009. Modeling multiple ecosystem services, biodiversity conservation, commodity production, and tradeoffs at landscape scales. Fr ont. Ecol. Environ. 7, 4-11. 259 Olander, L.P., Gibbs, H.K., Steininger, M., Swenson, J.J., Murray, B.C., 2008. Reference scenarios for deforestation and forest degradation in support of REDD: a review of data and methods. Environ Res Lett 3, 025011. Oliver, P.E., Marwell, G., 1988. The Paradox of Group-Size in Collective ActionŒa Theory of the Critical Mass. II. Am . Sociol. Rev. 53, 1-8. Olson, M., 1965. The Logic of Collective Action. Harvard University Press, Cambridge, MA. Ostrom, E., 1990. Governing the Commons: The Evol ution of Institutions for Collective Action. Cambridge University Press, New York. Ostrom, E., 2005. Understanding Institutional Divers ity. Princeton University Press, Princeton, NJ. Ostrom, E., 2007. A diagnostic approach for goi ng beyond panaceas. Proc. Natl. Acad. Sci. USA 104, 15181-15187. Ostrom, E., 2009. A General Framework for Anal yzing Sustainability of Social-Ecological Systems. Science 325, 419-422. Ostrom, E., Burger, J., Field, C.B., Norgaar d, R.B., Policansky, D., 1999. Revisiting the commons: Local lessons, global challenges. Science 284, 278-282. Ostrom, E., Gardner, R., Walke, J., 1994. Ru les, Games, and Common-Pool Resources. University of Michigan Press, Ann Arbor. Ostrom, E., Janssen, M.A., Anderies, J.M., 2007. Going beyond panaceas. Proc. Natl. Acad. Sci. U. S. A. 104, 15176-15178. Ouyang, Z., Xu, W., Wang, X., Wang, W., Dong, R., Zheng, H., Li, D., Li, Z., Zhang, H., Zhuang, C., 2008. Impact assessment of Wenchua n Earthquake on ecosystems. Acta Ecol Sin 28, 5801-5809. Pareto, V., [1906] 1927. Manual of politic al economy. A.M. Kelley, New York. Pearl, J., 2000. Causality: Models, Reasoning, and Inference. Cambridge University Press, Cambridge. Pecorino, P., Temimi, A., 2008. The group size para dox revisited. J. Public Econ. Theory 10, 785-799. Pereira, E., Queiroz, C., Pereira, H.M., Vicente, L., 2005. Ecosystem services and human well- being: a participatory study in a mountain community in Portugal. Ecol. Soc. 10, 14. Phillips, A., 2003. Turning ideas on their head: Th e New Paradigm For Protected Areas. George Wright Forum 20, 8-32. 260 Portes, A., 1998. Social Capital: Its origins and applications in modern sociology. Annual Review of Sociology 24, 1-24. Portes, A., Landolt, P., 1996. The Downside of Social Capital. The American Prospect, 18-21, 94. Poteete, A.R., Janssen, M.A., Ostrom, E., 2010. Working Together: Collective Action, the Commons, and Multiple Methods in Practice. Princeton University Press, Princeton. Poteete, A.R., Ostrom, E., 2004. Heterogeneity, group size and collective action: The role of institutions in forest management. Dev. Change 35, 435-461. Power, A.G., 2010. Ecosystem services and agriculture: tradeoffs and synergies. Philos T R Soc B 365, 2959-2971. Pretty, J., 2003. Social capital and the collective management of resources. Science 302, 1912- 1914. R Development Core Team, 2011. R: A Program ming Environment for Data Analysis and Graphics. Ranis, G., Stewart, F., Ramirez, A., 2000. Ec onomic growth and human development. World Devel. 28, 197-219. Raudsepp-Hearne, C., Peterson, G.D., Bennett, E.M., 2010. Ecosystem service bundles for analyzing tradeoffs in diverse landscapes. Proc. Natl. Acad. Sci. U. S. A. 107, 5242-5247. Ready, R., Navrud, S., 2006. International benefit transfer: Methods and validity tests. Ecol. Econ. 60, 429-434. Redford, K.H., Adams, W.M., 2009. Payment for eco system services and th e challenge of saving nature. Conserv. Biol. 23, 785-787. Richard, T.C., Nicholas, E.F., Norman, F.M., 2001. Contingent Valuation: Controversies and Evidence. Environ. Resource Econ. 19, 173-210. Rosa, E.A., Dietz, T., Diekmann, A., Jaeger , C.C., 2010. Human Footprints on the Global Environment: Threats to Su stainability. MIT Press, Cambridge, Massachusetts. Rowe, K., 2006. The measurement of composite va riables from multiple indicators: Applications in Quality Assurance and Accreditation SystemsŒChildcare. Australian Council for Educational Research, Camberwell, Victoria. Rowe, K.S., Rowe, K.J., 1997. Norms for parent al ratings on Conners' Abbreviated Parent-Teacher Questionnaire: implications for the design of behavioral rating inventories and analyses of data derived from them. J. Abnorm. Child Psychol. 25, 425-451. Rustagi, D., Engel, S., Kosfeld, M., 2010. Conditional Cooperation and Costly Monitoring Explain Success in Forest Commons Management. Science 330, 961-965. 261 Sanchez-Azofeifa, G.A., Pfaff, A., Robalino, J.A., Boomhower, J.P., 2007. Costa Rica's payment for environmental services program: Intention, implementation, and impact. Conserv. Biol. 21, 1165-1173. Sandoval, J., 1981. Format effects in two teacher rating scales of hyperactivity. J. Abnorm. Child Psychol. 9, 203-218. Schaller, G.B., Hu, J., Pan, W., Zhu, J., 1985. The Giant Pandas of Wolong. University of Chicago Press, Chicago. Scherr, S.J., Bennett, M.T., Loughney, M., Canby, K., 2006. Developing Future Ecosystem Service Payments in China: Lessons Learned from International Experience. Schielzeth, H., 2010. Simple means to improve th e interpretability of regression coefficients. Methods Ecol Evol 1, 103-113. Scott, M.J., Bilyard, G.R., Link, S.O., Ulibarri, C.A., Westerdahl, H.E., Ricci, P.F., Seely, H.E., 1998. Valuation of Ecological Resources and Functions. Environ. Manage. 22, 49-68. Scullion, J., Thomas, C.W., Vogt, K.A., Perez- Maqueo, O., Logsdon, M.G., 2011. Evaluating the environmental impact of payments for ecosystem services in Coatepec (Mexico) using remote sensing and on-site interviews. Environ. Conserv. 38, 426-434. Shackleton, C., Shackleton, S., Gambiza, J., Nel, E., Rowntree, K., Urquhart, P., 2008. Links between Ecosystem Services and Poverty Alleviatio n: Situation analysis for arid and semi-arid lands in southern Africa. Ecosystem Servi ces and Poverty Reduction Research Programme: DFID, NERC, ESRC. Shaw, R., Goda, K., 2004. From disaster to sust ainable civil society: The Kobe experience. Disasters 28, 16-40. Shrestha, R., Rosenberger, R., Loomis, J., 2007. Benefit Transfer Using Meta-Analysis in Recreation Economic Valuation. Econ Non-Mark Good 9, 161-177. Simon, H., 1953. Causal orderi ng and identifiability, in: Hood, W.C., Koopmans, T.C. (Eds.), Studies in Econometric Method. Wiley, New York, pp. 49-74. Smit, B., Wandel, J., 2006. Adaptation, adap tive capacity and vulnerability. Global Environ Chang 16, 282-292. Southwood, K.E., 1978. Substantive Theory and Sta tistical Interaction - 5 Models. Amer J Sociol 83, 1154-1203. SPGPWERR, 2008. The State Overall Planning for Post-Wenchuan Earthquake Restoration and Reconstruction (In Chinese and English). State Planning Group of Post-Wenchuan Earthquake Restoration and Reconstruction (SPGPWERR), Beijing. 262 State Forestry Administration of China, 2005. Report of Forest Resources in China. China Forestry Press, Beijing. State Planning Group of Post-Wenchuan Earthqua ke Restoration and Reconstruction, 2008. The State Overall Planning for Post-Wenchuan Earth quake restoration and Reconstruction, Beijing. Summers, J.K., Smith, L.M., Case, J.L., Linthur st, R.A., 2012. A review of the elements of human well-being with an emphasis on the cont ribution of ecosystem services. Ambio 41, 327- 340. Swallow, B.M., Sang, J.K., Nyabenge, M., Bundotich, D.K., Duraiappah, A.K., Yatich, T.B., 2009. Tradeoffs, synergies and traps among ecosystem services in the Lake Victoria basin of East Africa. Environ Sci Policy 12, 504-519. Tallis, H., Kareiva, P., Marvier, M., Chang, A., 2008. An ecosystem services framework to support both practical conservation and economic deve lopment. Proc. Natl. Acad. Sci. U. S. A. 105, 9457-9464. Task Force for Eco-Compensation Mechanisms and Polices in China, 2007. Eco-compensation Mechanisms and Polices in China. Science Press, Beijing. TEEB, 2010. The economics of ecosystems and biodi versity: mainstreami ng the economics of nature: a synthesis of the approach, conclusions and recommenda tions of TEEB. Progress Press, Malta. Todd, S., 1992. Collective Action: Theory and App lications. University of Michigan Press, Ann Arbor. Tuanmu, M.-N., 2012. Spatiotemporal dynamics of giant panda habitat: implications for panda conservation under a changing environment. PhD Dissertation, Department of Fisheries and Wildlife. Michigan State University, East Lansing. Tuanmu, M.N., Vina, A., Bearer, S., Xu, W.H., Ouyang, Z.Y., Zhang, H.M., Liu, J.G., 2010. Mapping understory vegetation using phenological characteristics derived from remotely sensed data. Remote Sens. Environ. 114, 1833-1844. Tucker, C.M., 2010. Learning on Governance in Forest Ecosystems: Lessons from Recent Research. International Journal of the Commons 4, 687-706. Turner, B.L., Kasperson, R.E., Matson, P.A., McCarthy, J.J., Corell, R.W., Christensen, L., Eckley, N., Kasperson, J.X., Luers, A., Martello, M.L., Polsky, C., Pulsipher, A., Schiller, A., 2003. A framework for vulnerability analysis in su stainability science. Proc. Natl. Acad. Sci. U. S. A. 100, 8074-8079. Turner, B.L., Lambin, E.F., Reenberg, A., 2008. Land Change Science Special Feature: The emergence of land change science for global environmental change and sustainability. Proc. Natl. Acad. Sci. USA 105, 2751-2751. 263 Uchida, E., Rozelle, S., Xu, J.T., 2009. Conservation Payments, Liquidity Constraints, and Off- Farm Labor: Impact of the Grain-for-Green Pr ogram on Rural Households in China. Am. J. Agric. Econ. 91, 70-86. Uchida, E., Xu, J.T., Rozelle, S., 2005. Grain for green: Cost-effectiveness and sustainability of China's conservation set-aside program. Land Econ. 81, 247-264. van Emmerik, I.J.H., 2006. Gender differences in the creation of different types of social capital: A multilevel study. Social Netwks 28, 24-37. Vemuri, A.W., Costanza, R., 2006. The role of human, social, built, and natural capital in explaining life satisfaction at the country leve l: Toward a National Well-Being Index (NWI). Ecol. Econ. 58, 119-133. Viña, A., Bearer, S., Chen, X., He, G., Linderman , M., An, L., Zhang, H., Ouyang, Z., Liu, J., 2007. Temporal changes in giant panda habitat connectivity across boundaries of Wolong Nature Reserve, China. Ecol. Appl. 17, 1019-1030. Viña, A., Bearer, S., Zhang, H., Ouyang, Z., Liu, J., 2008. Evaluating MODIS data for mapping wildlife habitat distribution. Remote Sens. Environ. 112, 2160-2169. Viña, A., Chen, X., McConnell, W., Liu, W., Xu, W., Ouyang, Z., Zhang, H., Liu, J., 2011. Effects of natural disasters on conservation po licies: The case of the 2008 Wenchuan Earthquake, China. Ambio 40, 274-284. Viña, A., Chen, X., Yang, W., Liu, W., Li, Y., Ouyang, Z ., Liu, J., 2013. Improving the efficiency of conservation policies with the use of surrogates derived from remotely sensed and ancillary data. Ecol. Indicators 26, 103-111. Wade, R., 1988. Village republics: economic cond itions for collective action in South India. Cambridge University Press, Cambridge. Wahab, S., Pigram, J.J., 1997. Tourism, Development and Growth. Routledge, London. Wallace, K., 2008. Ecosystem services : Multiple classifications or confusion? Biol. Conserv. 141, 353-354. Wallace, K.J., 2007. Classification of ecosystem se rvices: Problems and solutions. Biol. Conserv. 139, 235-246. Ward, F.A., Pulido-Velazquez, M., 2008. Water c onservation in irrigation can increase water use. Proc. Natl. Acad. Sci. U. S. A. 105, 18215-18220. Wellman, B., Frank, K., 2001. Network Capital in a Multi-Level World: Getting Support in Personal Communities, in: Lin, N., Cook, K., Bu rt, R. (Eds.), Social Capital: Theory and Research. Aldine DeGruyter, Chicago, pp. 233-273. 264 Wellman, B., Wortley, S., 1990. Different Stroke s from Different Folks - Community Ties and Social Support. Amer J Sociol 96, 558-588. Whisman, M.A., McClelland, G.H., 2005. Designing, testing, and interpreting interactions and moderator effects in family resear ch. J. Fam. Psychol. 19, 111-120. Willis, K.J., Gillson, L., Knapp, S., 2007. Biodive rsity hotspots through time: an introduction. Philos T R Soc B 362, 169-174. Wilson, M.A., Hoehn, J.P., 2006. Valuing envi ronmental goods and services using benefit transfer: The state-of-the art and science. Ecol. Econ. 60, 335-342. Wolong Nature Reserve, 2005. Development Hist ory of Wolong Nature Reserve [in Chinese]. Sichuan Science and Technology Press, Chengdu. Wong, D.F.K., Li, C.Y., Song, H.X., 2007. Rura l migrant workers in urban China: living a marginalised life. Int. J. Soc. Welfare 16, 32-40. Wooldridge, J.M., 2002. Econometric analysis of cross section and panel data. MIT Press, Cambridge, MA. Wright, S., 1921. Correlation and ca usation. J Agric Res 20, 557-585. Wunder, S., 2005. Payments for environmental se rvices: some nuts and bolts. CIFOR Occasional Paper No.42. Center for International Forestry Research Bogor, Indonesia. Wunder, S., 2007. The efficiency of payments for environmental services in tropical conservation. Conserv. Biol. 21, 48-58. Wunder, S., Alban, M., 2008. Decentralized paymen ts for environmental services: The cases of Pimampiro and PROFAFOR in Ecuador. Ecol. Econ. 65, 685-698. Wunscher, T., Engel, S., Wunder, S., 2008. Spat ial targeting of payments for environmental services: A tool for boosting conservation benefits. Ecol. Econ. 65, 822-833. WWF, 2007. Terrestrial Ecoregions. Xinhua News Agency, 2008a. Death toll from Ch ina's May earthquake remains unchanged at 69,227. Xinhua News Agency, 2008b. Direct economic loss of the Wenchuan Earthquake is estimated to be above one trillion. Xu, J.T., Katsigris, E., White, T.A., 2002. Impl ementing the natural forest protection program and the sloping lands conversion program. Wester n China Forests and Grassland Taskforce of China Council for International Cooperation on Environment and Development, Beijing. Xu, J.T., Yin, R.S., Li, Z., Liu, C., 2004. Sloping land conversion: Cost-effectiveness, structural adjustment, and economic sustainability. China Economics Quarterly 4, 139-162 (in Chinese). 265 Yang, W., Chang, J., Xu, B., Peng, C., Ge, Y ., 2008. Ecosystem service value assessment for constructed wetlands: A case study in Hangzhou, China. Ecol. Econ. 68, 116-125. Yang, W., Dietz, T., Kramer, D.B., Chen, X ., Liu, J., 2013a. Going beyond the Millennium Ecosystem Assessment: an index system of human well-being. PLoS ONE 8, e64582. Yang, W., Dietz, T., Kramer, D. B., Ouyang, Z., Liu, J., 2013b. An integrated approach to understand the linkages between ecosystem serv ices and human well-being (Under review). Yang, W., Dietz, T., Liu, W., Luo, J., Liu, J., 2013c. Going beyond the Millennium Ecosystem Assessment: an index system of human depe ndence on ecosystem services. PLoS ONE 8, e64581. Yang, W., Dietz, T., Luo, J., Liu, W., Ouya ng, Z., Liu, J., 2013d. Interaction effects of conservation and development policies on rural ho usehold income and income structure. (Under review). Yang, W., Liu, W., Viña, A., Luo, J., He, G ., Ouyang, Z., Liu, J., 2013e. Performance and prospects on payments for ecosystem services programs: evidence from China. J. Environ. Manage. 127, 86-95. Yang, W., Liu, W., Viña, A., Tuan mu, M., He, G., Dietz, T., Liu, J., 2013f. Nonlinear effects of group size on collective action and resource outcomes. Proc. Natl. Acad. Sci. USA 110, 10916- 10921. Yin, R.S., 2009. An Integrated Assessment of China's Ecological Restorat ion Programs. Springer, Dordrecht, The Netherlands. Zhang, H., Li, D., Wei, R., Tang, C., Tu, J ., 1997. Advances in conservation and studies on reproductivity of giant pandas in Wolong. Sichuan J Zool 16, 31-33. Zhang, L., Wang, N., 2003. An initial study on habitat conservation of Asian elephant (Elephas maximus), with a focus on human elephant conf lict in Simao, China. Biol. Conserv. 112, 453- 459. Zhang, W., Ricketts, T.H., Kremen, C., Carney , K., Swinton, S.M., 2007. Ecosystem services and dis-services to agriculture. Ecol. Econ. 64, 253-260. Zhang, X.Q., Zhu, F., 2011. Group Size and Incentiv es to Contribute: A Natural Experiment at Chinese Wikipedia. Am. Econ. Rev. 101, 1601-1615.