Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The role of trust and risk preferences in the investment decision of farmers : evidence from surveys… Gong, Yazhen 2010

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2010_spring_gong_yazhen.pdf [ 1.14MB ]
Metadata
JSON: 24-1.0069996.json
JSON-LD: 24-1.0069996-ld.json
RDF/XML (Pretty): 24-1.0069996-rdf.xml
RDF/JSON: 24-1.0069996-rdf.json
Turtle: 24-1.0069996-turtle.txt
N-Triples: 24-1.0069996-rdf-ntriples.txt
Original Record: 24-1.0069996-source.json
Full Text
24-1.0069996-fulltext.txt
Citation
24-1.0069996.ris

Full Text

 The Role of Trust and Risk Preferences in the Investment Decision of Farmers: Evidence from Surveys and Field Experiments in Rural China   by Yazhen Gong  B.A., Beijing Forestry University, 1995 M.Sc., University of the Philippines, 2002   A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY  in  The Faculty of Graduate Studies (Forestry)    THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)   April 2010   ©Yazhen Gong 2010 \  ii  Abstract  In rural China, forest and agricultural lands are undergoing major reforms, leading to many new challenges. I chose to understand the factors affecting a farmer’s investment decision on their lands. The first manuscript focuses on the role of property rights, social capital and contrac- tual rules in the design and implementation of the world’s first Clean Development Me- chanism (CDM) afforestation project in China.  Using village-level surveys, I found: [1] although the project facilitates participation through carbon pooling and a share-holding system, much of the project land remained unforested; [2] the primary reasons for the un- forested lands included constrained contractual rules, property rights allocation disputes, and low levels of social capital in some villages. The second manuscript identifies determining factors of villagers’ trust in their peer members.   I used a unique data set collected by household surveys and trust games con- ducted in 30 administrative villages in Yunnan Province of southwestern China.  I found: [1] the survey measure and experimental measure of trust were strongly correlated; [2] evidence that the positive individual social interactions and positive past experiences had a  significant and positive effect on the villagers’ mutual trust; [3] that while formal vil- lage institutions might have gradually substituted for the traditional informal institutions, the formal institutions has not effectively replaces the roles of informal village institu- tions in maintaining villagers’ mutual trust; [4] openness to the outside world and the market could have eroded mutual trust among rural villagers in Yunnan Province.  iii    In the third manuscript, I used a data set collected through artefactual field expe- riments and household surveys from 30 villages in Yunnan Province.  I linked the expe- rimental measure of risk preference to household investment decisions on input use on farmlands. Major findings in this chapter include: [1] farmland size was negatively and significantly correlated with the intensity of both chemical fertilizer and pesticide use; [2] a family head’s risk aversion level was significantly and positively correlated with inten- sity of both chemical fertilizer use and pesticide use; [3] a family’s connection to social networks negatively and significantly predicts its decision regarding the intensity of pes- ticide use.  iii  Table of Contents  Abstract .............................................................................................................................. ii Table of Contents ............................................................................................................. iii List of Tables .................................................................................................................... vi List of Figures .................................................................................................................. vii Acknowledgements ........................................................................................................ viii Dedication ......................................................................................................................... ix Co-authorship Statement ................................................................................................. x Chapter 1  Introduction.................................................................................................... 1 1.1. Literature Review ................................................................................................. 3 1.1.1. Barriers to Farmers’ Investment Decisions: Risk and Transaction Costs ..... 3 1.1.2. Social Capital ................................................................................................ 5 1.1.3. Trust .............................................................................................................. 6 1.1.4. Field Experiment Method and External Validity .......................................... 7 1.2. Research Objectives ............................................................................................. 9 1.3. Dissertation Structure ........................................................................................... 9 Bibliography .................................................................................................................... 12 Chapter 2  Participation in the World’s First Clean Development Mechanism Forest Project: The Role of Property Rights, Social Capital and Contractual Rules .......... 21 2.1. Introduction ........................................................................................................ 22 2.2. Requirements for Participation in a CDM Forest Project .................................. 25 2.3. Background Information of Guangxi CDM Project ........................................... 33 2.3.1 Project Environmental and Development Objectives ................................. 34 2.3.2 Land Tenure Arrangements ........................................................................ 37 2.3.3 Project Baseline and “Additionality” .......................................................... 37 2.3.4 Pooling Arrangement and the Share-holding System ................................. 39 2.3.5 Contractual Arrangements between the Buyer and Sellers ......................... 41 2.3.6 Financial Arrangements for the Project ...................................................... 42 2.4. Data and Sampling Methods .............................................................................. 43  iv  2.5. Results and Discussion ....................................................................................... 44 2.5.1 Project Successes ........................................................................................ 44 2.5.2 Project Shortcomings .................................................................................. 50 2.6. Conclusions ........................................................................................................ 59 Bibliography .................................................................................................................... 62 Chapter 3  Individual Social Interactions, Village Institutions and Rural Villagers’ Mutual Trust: Survey and Field Experimental Evidence from Western Rural China ........................................................................................................................................... 67 3.1. Introduction ........................................................................................................ 68 3.2. The Measurement of Trust ................................................................................. 70 3.2.1 The Experimental Measure of Trust ........................................................... 71 3.2.2 The Experimental Measure and Psychological Games ............................... 73 3.3. Survey and Experimental Design ....................................................................... 74 3.3.1 The Survey Design: Measuring Stated Trust and Beliefs ........................... 74 3.3.2 The Experimental Design: Revealing Trusting Behaviors ......................... 76 3.4. Field Evidence for Player 1’s Motivations and a Theoretical Model ................ 82 3.4.1. Outcomes of the Game ................................................................................ 82 3.4.2. Field Evidence of Player 1’s Motivations ................................................... 82 3.4.3. The Theoretical Model ................................................................................ 86 3.5. Empirical Model Specifications ......................................................................... 89 3.5.1 Individual and Household Characteristics .................................................. 90 3.5.2 Identifying Potential Determinants of Beliefs ............................................ 92 3.5.3 Village Socio-economic Characteristics ................................................... 101 3.6. Results and Discussions ................................................................................... 103 3.6.1 Descriptive Statistics ................................................................................. 103 3.6.2 Main Results ............................................................................................. 105 3.6.3 Robustness Checks for the Effect of Beliefs on Trusting Behaviors ........ 115 3.7. Conclusions ...................................................................................................... 120 Bibliography .................................................................................................................. 122 Chapter 4  Risk Aversion and Farm Input Choice: Evidence from Field Experiments in China ................................................................................................... 127 4.1. Introduction ...................................................................................................... 128  v  4.2. Risk Aversion, Its Measurement and Relations to Farmers’ Investment Decisions .......................................................................................................... 131 4.3. Methods ............................................................................................................ 134 4.3.1. Survey Design and Data Description ........................................................ 134 4.3.2. Field Experimental Design and Data Description .................................... 138 4.4. Some Theory and Empirical Model Specifications .......................................... 143 4.5. Results and Discussions ................................................................................... 148 4.5.1. Descriptive Evidence ................................................................................ 148 4.5.2. Determinants of Household Investment Decisions on Input Use ............. 149 4.5.3. Robustness Checks.................................................................................... 156 4.6. Conclusions ...................................................................................................... 157 Bibliography .................................................................................................................. 159 Chapter 5   Conclusions ................................................................................................ 164 Bibliography .................................................................................................................. 169 Appendix 1 - Survey ..................................................................................................... 171   vi  List of Tables  2- 1  Multiple Tree Species to be Planted by the Project .................................................. 35 2- 2  Local land Users’ Potential Profit ............................................................................ 46 2- 3  Reasons for Unaccomplished Plans .......................................................................... 50 3- 1  Summary Statistics of Outcomes .............................................................................. 82 3- 2  Player 1's Reasons for Sending a Positive Amount .................................................. 84 3- 3  Player 1's Reasons for Sending Nothing .................................................................. 85 3- 4  Correlations between Village Institutions and Village Characteristics .................... 99 3- 5  Characteristics of Surveyed Administrative Villages ............................................. 103 3- 6  Descriptive Statistics for Trust-related Variables ................................................... 105 3- 7  Effect of Individual and Household Characteristics on the Amount Sent .............. 106 3- 8  Effect of Individual-level Social Interactions and Past Experiences on Trusting Behaviors ............................................................................................................... 108 3- 9  The Effect of Village Institutions on Trusting Behaviors ...................................... 112 3- 10 Surveyed Trust, Expected Return and Trusting Behaviors ................................... 118 4- 1    Summary Statistics ............................................................................................... 135 4- 2    The Ten Pair- wise Lotteries ................................................................................ 139 4- 3    Outcomes of the Risk Experiments ...................................................................... 143 4- 4    Correlations among Key Variables of Interest ..................................................... 148 4- 5    Household Decisions on Chemical Fertilizer Use (Consistent Subjects) ............. 150 4- 6    Household Decisions on Pesticide Use (Consistent Subjects) ............................. 151   vii  List of Figures  2- 1    A Location Map of Guangxi CDM Project  ........................................................... 34 2- 2   Annual and Accumulative Carbon Sink by the Project ........................................... 36 2- 3   A Share-holding System Created for Guangxi CDM Project .................................. 40      viii  Acknowledgements  I first present my thanks to all of my advisors who have offered me invaluable help and advice during my doctoral study at the University of British Columbia. I feel very grateful to my thesis major supervisor, Dr. Gary Bull, for his inspirations, supervision, excellent mentoring and tremendous support in many ways.  I am greatly indebted to Dr. Kathy Baylis, who gave me very rigorous technical training and guided me to think critically in doing solid empirical research through the whole period of my doctoral study. I owe my special thanks to Dr. Jintao Xu and his research team at Environmental Economics Program in China (EEPC), Peking University.  I must thank Dr. Xu for shar- ing his insights on resource management and environmental policies in China and his ge- nerous offering of his unique data set for me to accomplish my dissertation.  I must give my special acknowledgement to the survey team led by Dr. Xuemei Jiang and Dr. Yan Sun to collect data in rural villages in Yunnan Province.   I should also give my acknowl- edgement to Ms. Hang Yin and Mrs. Ling Li for their great assistance in data cleaning. I must thank Economy and Environment Program in Southeast Asia (EEPSEA) for its financial support for my doctoral field work and Dr. Herminia Francisco for her support. I must give acknowledgement to BIOCAP, Canada for providing me financial support to pursue my doctoral study with Dr. Gary Bull at the University of British Columbia.  viii  I especially thank Dr. Wictor Admonzwicz, who initially inspired me of conducting field experiments related to resource management in China, and gave me his valuable comments as a resource person of EEPSEA. I reserve my special thanks to my parents, my husband, my daughter, my brother and sister-in-law for their love and support.   ix     Dedication To My Beloved Lingyi (Nini) and Jinhong   x  Co-authorship Statement  Chapter 2: Title:                   Participation in the World’s First Clean Development Mechanism For- est Project: The Role of Property Rights, Social Capital and Contrac- tual Rules Authors:                 Yazhen Gong, Gary Bull and Kathy Baylis Role of co-authors: Yazhen Gong contributed to the development of research idea, data analysis and preparation for manuscript.  Gary Bull contributed to the development of research idea and survey design; Kathy Baylis contri- buted to theoretical model development. Chapter 3: Title:                         Individual Social Interactions, Village Institutions and Rural Villag- ers’ Mutual Trust--Survey and Field Experimental Evidence from Western Rural China Authors:                     Yazhen Gong, Kathy Baylis, Jintao Xu and Gary Bull Roles of co-authors:    Yazhen Gong contributed to the development of research idea, da- ta analysis and preparation for manuscript.  Kathy Baylis contri- buted to model development and data analysis; Jintao Xu contri- buted to research design and data collection; Gary Bull contributed to research design and manuscript preparation.   xi  Chapter 4: Title:                   Risk Aversion and Farm Input Choice: Evidence from Field Experi- ments in China Authors:                   Yazhen Gong, Kathy Baylis, Jintao Xu, Robert Kozak and Gary Bull Roles of co-authors:    Yazhen Gong contributed to the development of research idea, da- ta analysis, and manuscript preparation.  Kathy Baylis contributed to model development and data analysis; Jintao Xu contributed to research design and data collection; Robert Kozak contributed to statistical analysis and manuscript preparation; Gary Bull contri- buted to research design and manuscript preparation.   1    Chapter 1  Introduction  2  In China, new opportunities and challenges have emerged for forestry and agricultural sectors.  For the forestry sector, afforestation and reforestation are national strategies for climate change mitigation.   For the agricultural sector, the overuse of chemical fertilizer and chemical pesticides, which has caused environmental degradation and reduced hu- man health in China (Hu et al. 2007; Huang et al. 2003; Huang et al. 2008), is under re- view.  Although many efforts are being made to restore agricultural and forest lands through new policy instruments or help improve management practices on those lands, the efforts are fraught with difficulties.  To better understand these difficulties, it is useful for policy makers to regularly evaluate the efforts in order to gain insight on how to im- prove the policy environment and management practices.   Therefore, it is important for them to understand factors affecting a farmer’s investment decisions on their lands, espe- cially given the new land tenure system in China, in order for the forest and agriculture policy makers to evaluate the impact on humans. My dissertation focuses on investigating farmers’ investment decisions on their lands, both on community level and household level, to provide useful information for policy makers in China.  I investigate the farmers’ investment decisions at the community level based on the fact that in forestry sector, many of farmers’ land use decisions are still made at community level, even after the most recent land tenure reform1.  I also investi- gate the farmers’ investment decisions at the household level, because the investment de- cisions on agriculture lands are largely made at the household level.  Since the “House-  1 In early 1980s China adopted “Three Fix Policy” in forest sector, which was similar to “Household Re- sponsibility System” undertaken in agricultural sector.  However, since 1980s, forestlands are still largely under collective management.  In the early 2000s, a new wave of individualization’ of forestlands was in- itiated in Fujian and Jiangxi Provinces and then expanded over the collectively forest region in China.  3  hold Responsibility System” undertaken in agricultural sector toward individualization of agricultural lands in late 1970s, almost all of agricultural lands have been under the man- agement of individual farmers’ households. To investigate farmers’ investment decisions, I pay special attention to the role of farmers’ individual preferences, such as risk preferences, and social capital (including trust), which is a community-level attribute (Coleman 1990; Putnam 2000).  Specifically, I focus on trust and risk preferences by linking trust and risk preferences to farmers’ in- vestment decisions on two types of projects: [1] investment in afforestation on barren lands in compliance with the Clean Development Mechanism (CDM) forest projects re- quirements, and [2] investment in chemical fertilizer and pesticides on farmlands.  I use survey and experimental methods to measure trust and risk preferences. 1.1. Literature Review 1.1.1. Barriers to Farmers’ Investment Decisions: Risk and Transaction Costs Investment in agricultural production or forest projects is susceptible to risk, due to hu- man-induced or naturally-occurring factors. The list of risk factors includes: ambiguous property rights, changing government policies, uncertain market prices, extreme weather events, and pest outbreaks (Just and Zilberman 1983; Knight et al. 2003; Keenan et al. 2004; Reedy 2003).   As many small-scale poor land users in developing countries have both small plots of land and credit constraints (Eswaran and Kotwal 1990; Just and Zil- berman 1985), they cannot easily absorb many of the negative shocks associated with risk.  4  In forest carbon projects, a new type of forest management activity being considered by farmers, there is high transaction costs (Milne 1999; van Kooten et al. 2002) and this impedes the participation of small-scale and dispersed land users (Pagiola 2008).  Trans- action costs typically include the costs of project design, regulatory approval, validation, registration, information search, negotiation, signing and implementation of contracts, monitoring, insurance, verification and certification (Australian Greenhouse Office 2005; Milne 1999), costs of communicating with project partners, and costs of ensuring parties fulfill contracted obligations (Smith and Scherr 2003).   Many of these transaction costs are considered fixed costs, i.e. they are not dependant on project size (Australian Green- house Office 2005; Milne 1999).  This implies that for many small scale land users, the transaction costs are too high to participate. For small-scale farmers in developing countries, high risk combined with high trans- action cost is a formidable barrier to investment (Milne 1999; van Kooten et al. 2002). For example, a CDM forest project represents a type of investment that involves both high risk and high transaction cost (Haupt and von Lüpke 2007) and it often impedes small-scale and disperse land users from participating (Pagiola 2008). To reduce risk and transaction costs, pooling or bundling individual carbon sequestra- tion activities has been recommended.  Signing collective contracts with groups of small- holders is seen as a practical means to facilitate participation (Australian Greenhouse Of- fice 2005; Grieg-Gran et al. 2005; Wunder 2005; Pagiola 2008), because it can help spread transaction costs over a group of farmers, disperse risk over a larger geographical area in the event of disasters (Richardson 2005), and smooth against yearly fluctuations  5  across individual plots and fields (Carter 1987; Grieg-Gran et al. 2005).  In spite of its apparent advantages, pooling does require collective action, the success of which largely depends on a complex mix of property rights, contracts and social capital (Ostrom 1990; Pagiola et al. 2005; Wunder 2005).   In chapter 2, I demonstrate how the world’s first CDM forest project, implemented in China, illustrates the important mix of these three attributes and the impact on a farmer’s decisions to invest in tree planting on their barren lands. 1.1.2. Social Capital Although the exact definition of social capital is still under considerable debate, I take social capital to be broadly defined as the connections among individuals, i.e. networks, norms, trust, concerns for one’s associates and willingness to sanction violators of rules or norms (Bowles and Gintis 2002; Putnam 2000).  Social capital is considered multifa- ceted: while some conceive social capital as a person’s social characteristics, such as so- cial skills and charisma, others regard it as an individual attribute (Glaeser et al. 2000), a community-level attribute (Coleman 1990; Putnam 2000), or even an attribute of an insti- tutional environment (North 1990).  Social capital is thought to have the following com- mon characteristics: [1] it has economic consequences (Coleman 1990; Putnam 2000; Putnam et al. 1993); [2] it has both private and public good properties (Putnam 2000); [3] it can provide such functions as organizing information sharing, coordinating activities and facilitating collective decision making (Sobel 2002). Social capital plays an important role in resource management (Ostrom 1990; Pretty and Ward 2001; Baland and Platteau 2000), public good provisions (Edward and Gugerty  6  2005) and even adaptation to climate change (Adger 2003).  Social capital within rural communities provides one solution to pursue sustainable livelihood and resource man- agement that often involve a rural communities’ collective action (Pretty and Ward 2001). 1.1.3. Trust Trust is thought to be critical for social capital (Bellemare and Kröger 2007; Fukuyama 1995).  It can be viewed as an individual’s priori estimate of the probability that other people will act cooperatively for mutual benefit (Gambetta 1988).  It can influence the economic success on macro- and micro- levels by reducing transaction costs (Alesina and La Ferrara 2002), lubricating co-operation (Pretty and Ward 2001; Sobel 2002) and faci- litating the enforcement of formal or informal contracts (Asquith et al. 2008; Lyon 2003; Ostrom and Nagendra 2007).  It has been found to be associated with economic growth (Knack and Keefer 1997; La Porta et al. 1997; Zak and Knack 2001), good governance (Knack 2001), individual economic performances, including financial loan repayment rates (Karlan 2005), individual stock purchase (Guiso et al. 2008) and participation in community resource management (Boudma et al. 2008). In the past, both survey and experimental methods have been used to measure trust. On its own, the survey method has been criticized for its lack of behavioral relevance (Naef and Schupp 2009) and its introduction of the divergence between stated versus ac- tual preferences and beliefs (Bouma et al. 2008). As an alternative, the experimental me- thod has been increasingly used as a method to measure trust (Berg et al. 1995; Barr 2003; Bouma et al. 2008; Buchan et al. 2008; Croson and Buchan 1999; Eckel and Wilson 2004; Glaeser et al. 2000; Karlan 2005; Naef and Schupp 2009; Schechter 2007).  In a few in-  7  stances, the survey method and experimental method have been used together to measure trust (Glaeser et al. 2000; Naef and Schupp 2008).  However, the relationship between the two methods has been inconclusive.  Glaeser et al. (2000) found that the survey measure of trust was not correlated with trusting behaviors observed in trust games, while others found a significant correlation between the survey and experimental measures of trust (Bellemare and Kroeger 2007; Holm and Nystedt 2008). Glaeser et al. (2000) found that the survey measure of trust was not correlated with trusting behaviors observed in trust games, while Bellemare and Kröger (2007) found a significant correlation between the survey and experimental measures.   Naef and Schupp (2009) showed that the inconclusive relationship between the survey and experimental measures could attribute to the fact that they were capturing different dimensions of trust, as trust in Glaeser et al. (2000) could refer to generalized trust in the overall population, which involved strangers, or personalized trust arising from repeated interpersonal inte- ractions (Fafchamps 2006). Their experimental and survey methods both focused on measuring the single dimension of “trust in strangers” where they found the experimental and survey measures of trust was strongly correlated.  In chapter 3 of my dissertation, the survey and the experimental methods are used to measure trust and both methods focus on measuring the same dimension of trust: the “rural villagers’ mutual trust” in Yunnan Province in southwestern China. 1.1.4. Field Experiment Method and External Validity The experimental method is thought to be a more effective and convincing approach to collect data on behavioral questions, such as preferences, behavioral propensities and  8  other individual attributes than the survey method since experiments provide direct ob- servations of behavior (Carpenter 2000, 2002).   However, the conventional laboratory experiments only employ standard subject pool of students, an abstract framing and an imposed set of rules (Harrison and List 2004).  Because the conventional experimental method uses students as subjects, it is commonly criticized for its external validity; since it is difficult to draw inference from the student subject pool to the general population. In contrast, the field experiment method, which uses the field context instead of an abstract context (i.e. the classroom), is an alternative method (Harrison and List 2004). One major type of field experiment is the artefactual field experiment, in which the pro- cedure is the same as the conventional experiments, but with non-student subject pool.  In measuring social capital, some studies have attempted to combine survey method with artefactual field experiment method (Cardenas and Carpenter 2008; Bellemare and Kröger 2007).  In spite of this initial attempt, there remains a dearth of studies which use both the field experiment and survey methods to collect social capital data especially in the context of developing countries. The artefactual field experiment method has been used to measure one important in- dividual psychological trait: an individual’s attitude towards risk. The risk experiments have been conducted in field settings using farmers as subjects (Bingswanger 1980; Gris- ley and Kellog 1987; Hill 2009; Knight et al. 2003; Liu 2008; Schechter 2003).  In spite of the above studies, more empirical evidence is still needed for the external validity of field experiments used to measure risk preference.  9  In chapter 4 of my dissertation, I explore the following question: Can risk preference, measured by field experiments, strongly predict risk behavior revealed by a farmer’s in- vestment decisions?  I provided the external validity of a field experiment by linking the experimental measure of farmers’ risk preference to their risk behavior revealed in their on-farm investment decisions. 1.2. Research Objectives The dissertation has three main objectives: [1]  To analyze the roles of property rights, contractual rules and social capital in farmers’ decision to participate in a CDM forest project. [2]  To measure and analyze the determinants of rural villagers’ mutual trust. [3]  To analyze the effect of risk preference on a farmer’s on-farm investment de- cisions. 1.3. Dissertation Structure This dissertation follows the guidelines of manuscript-based thesis (University of British Columbia-Faculty of Graduate Studies 2009).  The main body of the dissertation is com- posed of 3 manuscripts In the first manuscript (Chapter 2: Participation in the World’s First Clean Develop- ment Mechanism Forest Project-The Role of Property Rights, Social Capital and Con- tractual Rules), I use a farmers’ decision on whether to invest their barren lands in a Clean Development Mechanism (CDM) forest project by analyzing determining factors  10  for farmers’ participation.  I start with a simple Coasean framework and then extend it to include institutional components.  Using village-level survey data, I find that: [1] al- though carbon pooling and a share-holding system created for the CDM project helped to address the high transaction costs and risk that otherwise would have been formidable, obstacles remained for the small-scale farmers to participate in and benefit from the project; [2] the primary reasons for the lands remaining unforested were constrained con- tractual rules, property rights allocation disputes and low levels of social capital in some villages. In the second manuscript (Chapter 3: Individual Social Interactions, Village Institu- tions and Rural Villagers’ Mutual Trust--Survey and Field Experimental Evidence from Western Rural China), I pay specific attention to the measurement and determinants of trust, a cognitive social capital.  I conducted both surveys and trust games in 30 adminis- trative villages in Yunnan Province in Southwest China.  I used the amount sent by the senders in trust games as an experimental measure of trust and created an indicator from surveyed questions on trust as a survey measure of trust.  The survey and experimental measures focused on the same dimension of trust, the rural villagers’ mutual trust.  Con- sidering there is a scant of researches on the determinants of trust, in the second chapter, I mainly intend to analyze the determinants of trust.   Using the unique data collected in Yunnan,  I find strong evidence that:  [1] the positive individual social interactions and past experiences with each other could have significant and positive effect on the amount sent by the senders; [2] the formal village institutions had negative and significant effect on the amount sent by the senders while informal institutions had positive but insignifi- cant effect on it; [3] the degree of openness to the outside world and the market had a  11  negative effect on the amount sent.  I also found strong evidence that the survey measure and experimental measure of trust were strongly correlated with each other; In the third manuscript (Chapter 4: Risk Aversion and Farm Input Choice: Evidence from Field Experiments in China), I focus on the experimental measure of risk and its external validity.  I collected a unique data set through artefactual field experiments and household surveys from 30 villages in the Yunnan Province of southwestern China and linked the experimental measure of risk preference to household investment decisions regarding the intensity of chemical fertilizer and pesticide use on farmlands.  I find that: [1] farmland size was negatively and significantly correlated with the intensity of both chemical fertilizer and pesticide use; [2] a family head’s risk aversion level was signifi- cantly and positively correlated with intensity of both chemical fertilizer use and pesti- cide use; [3] a family’s connection to social networks negatively and significantly pre- dicts its decision regarding the intensity of pesticide use.  The findings indicate that field experiments can be informative for real-world decisions.  12  Bibliography Adger, W.N. (2003).  Social capital, collective action, and adaptation to climate change. Economic Geography, 79, 387-404. Alesina, A. & La Ferrara, E. (2002).  Who trusts others?  Journal of Public Economics, 85(2), 207-234. Asquith, N.M., Vargas, M.T., & Wunder, S. (2008).  Selling two environmental services: in-kind payments for bird habitat and watershed protection in Los Negros, Bolivia. Ecological Economics, 65(4), 675-684. Australian Greenhouse Office (2005). Planning Forest Sink Projects: A Guide to Carbon Pooling and Investment Structures.  Retrieved on April 15, 2010, from http://www.greenbiz.com/business/research/report/2005/06/07/planning-forest-sink- projects-guide-carbon-pooling-and-investment-structu. Baland, J.M., & Platteau, J.P. (2000).  Halting degradation of natural resources: Is there a role for rural communities?  New York: Oxford University Press. Barr, A. (2003). Trust and expected trustworthiness: experimental evidence from Zim- babwean villages. Economic Journal, 113(489), 614-630. Bellemare, C., & Kröger, S. (2007).  On representative social capital.  European Econom- ic Review, 51(1), 183-202. Berg, J., Dickhaut, J., & McCabe, K. (1995). Trust, reciprocity and social history. Jour- nal of Economic Behavior & Organization, 10, 122-142.  13  Binswanger, H.P. (1980).  Attitudes toward risk: experimental measurement in rural India. American Journal of Agricultural Economics, 62(3), 395-407. Bouma, J., Bulte, E., & van Soest, D. (2008).  Trust and cooperation: social capital and community resource management.  Journal of Environmental Economics and Man- agement, 56(2), 155-166. Bowles, S., & Gintis, H. (2002). Social capital and community governance.  The Eco- nomic Journal, 112, F419-F436. Buchan, N.R., Croson, R. & Solnick, S. (2008).  Trust and gender: An examination of behavior and beliefs in the investment game.  Journal of Economic Behavior & Or- ganization, 68(3-4), 466-476. Cardenas, J.C., & Carpenter, J. (2008).  Behavioral development economics: lessons from field labs in the developing world.  The Journal of Development Studies, 44(3), 311- 338. Carpenter, J.P. (2000).  Negotiation in the commons: incorporating field and experimen- tal evidence into a theory of local collective action.  Journal of Institutional and Theoretical Economics, 156 (4), 661-683. Carpenter, J.P. (2002).   Measuring social capital: adding field experimental methods to the analytical toolbox.  In E. Elgar (Eds.) Social capital, economic development and the environment.  Sunder Ramaswamy: Jonathan Isham and Thomas Kelly. Carter, M.R., 1987. Risk sharing and incentives in the decollectivization of agriculture. Oxford Economic Papers 39, 577–595.  14  Coleman, J. S. (1990). Foundations of social theory. Cambridge, MA: Harvard Universi- ty Press. Croson, R. & Buchan, N. (1999).  Gender and culture: International experimental evi- dence from trust games.  American Economic Review, 89(2), 386-391. Eckel, C. C., & Wilson, R. K. (2004). Is trust a risky decision? Journal of Economic Be- havior and Organization, 55(4), 447-465. Edward, M., & Gugerty, M. K. (2005). Ethnic diversity, social sanctions, and public goods in Kenya.  Journal of Public Economics, 89, 2325-2368. Eswaran, M. & Kotwal, A. (1990).  Implications of credit constraints for risk behavior in less Developed economies.  Oxford Economic Papers, 42(2), 473-482. Fafchamps, M. (2006).  Development and social capital.  Journal of Development Studies, 42(7), 1180-1198. Fukuyama, F. (1995). Trust: the social virtues and the creation of prosperity. London: Hamish Hamilton. Gambetta, D (edited) (1988). Trust: making and breaking cooperative relations.  Oxford: Basil Blackwell. Glaeser, E.L., Laibson, D.I., Scheinkman, J.A. & Soutter, C.L. (2000).  Measuring trust. The Quarterly Journal of Economics, 115 (3), 811-846.  15  Grieg-Gran, M., Porras, I. & Wunder, S (2005). How can market mechanisms for forest environmental services help the poor? Preliminary lessons from Latin America. World Development 33(9), 1511-1527. Grisley, W., & Kellog, E. (1987).  Risk-taking preferences of farmers in Northern Thail- and: measurements and implications.  Agricultural Economics, 1, 127-142. Guiso, L., Sapienza, P., & Zingales, L. (2008).  Trusting the stock market.  Journal of Finance, 63(6), 2557-2600. Harrison, G., & List, J. A. (2004). Field experiment. Journal of Economic Literature, 42, 1009-1055. Haupt, F., & von Lüpke, H (2007).  Obstacles and opportunities for afforestation and re- forestation projects under the Clean Development Mechanism of the Kyoto Protocol. FAO Advisory Committee on Paper and Wood Products. Hill, R.V. (2009).  Using stated preferences and beliefs to identify the impact of risk on poor households.  The Journal of Development Studies, 45(2), 151-171. Holm, H., & Nystedt, P. (2008). Trust in surveys and games-A methodological contribu- tion to the influence of money and location. Journal of Economic Psychology, 29, 522-542. Hu, R., Cao, J., Huang, J., Peng, S., Huang, J., Zhong, X., Zou, Y., Yang, J., & Beresh, R.J. (2007).  Farmers participatory testing of standard and modified site-specific ni- trogen management for irrigated rice in China.  Agricultural Systems, 94, 331-340.  16  Huang, J., Rozelle, & Lu, F. (2003).  Farm pesticide use, rice production and human health.  In T.W. Mew, D.S. Brar, S. Peng, D. Dawe & B. Hardy (Eds.), Rice Science, innovations and impact for livelihood (pp. 901-918).  International Rice Research In- stitute. Huang, J., Hu, R., Cao, J., & Rozelle, S. (2008).  Training programs and in-the-field guidance to reduce China’s overuse of fertilizer without hurting profitability.  Jour- nal of Soil and Water Conservation, 63 (5), 165A-167A. Just, R.E., & Zilberman, D. (1983).  Stochastic structure, farm size, and technology adop- tion in developing agriculture.  Oxford Economic Papers, 35(2), 307-328. Just, R.E., & Zilberman, D. (1985).  Risk aversion, technology choice, and equity effects of agricultural policy.  American Journal of Agricultural Economics, 67 (7), 435-440. Karlan, D. S. (2005). Using experimental economics to measure social capital and predict financial decisions.  American Economic Review, 95 (5), 1688-1699. Keenan, R.J., Davey, S., Grieve, A., Moran, B., Donaldson, J. (2004). Market mechan- isms and assessment methods for environmental services from private forests in Aus- tralia. In Proceedings of the workshop forests for poverty reduction: opportunities with clean development mechanism, environmental services and biodiversity, Seoul, Korea. Food and Agriculture Organization of the United Nations (FAO) (edited), Bangkok, pp. 49-60. Knack, S. (2001).  Aid dependence and the quality of governance: Cross-country empiri- cal tests.  Southern Economic Journal, 68(2), 310-329.  17  Knack, S., & Keefer, P. (1997). Does social capital have an economic payoff? A cross- country investigation. The Quarterly Journal of Economics, 112(4), 1251-1288. Knight, J., Weir, S. & Woldehanna, T. (2003).  The role of education in facilitating risk- taking and innovation in agriculture.  The Journal of Development Studies, 39 (6), 1- 22. La Porta, R.  Lopez-de-Silane, F., Shleifer, A., Vishny, R.W. (1997).  Trust in large or- ganizations.  American Economic Review, 87(2), 333-338. Liu, E.M. (2008). Time to change what to sow: risk preferences and technology adoption decisions of cotton farmers in China.  Working Papers 1064, Department of Eco- nomics, Princeton University. Lyon, F. (2003).  Community groups and livelihoods in remote rural Areas of Ghana: How small-scale farmers sustain collective action.  Community Development Journal, 38, 323-331. Milne, M. (1999). Transaction costs of forest carbon projects.  Working Paper CC05. ACIAR project ASEM 1999/093.  University of New England.  Retrieved on April 15, 2010, from  http://www.une.edu.au/carbon/CC05.PDF. Naef, M., & Schupp, J. (2009).  Measuring trust: experiments and surveys in contrast and combination.  Discussion Paper (4087).  Bonn, Germany: The Institute for the Study of Labor (IZA).  18  North, D.C. (1990).  Institutions, institutional change and economic performance.  Cam- bridge: Cambridge University Press. Ostrom, E. (1990).  Governing the commons: the evolution of institutions for collective action.  Cambridge: Cambridge University Press. Ostrom, E., & Nagendra, H. (2007).  Governing the commons in the new millennium: a diversity of institutions for natural resource management. In C. J. Cleveland (Eds.), Encyclopedia of Earth.  Washington, DC: Environmental Information Coalition, Na- tional Council for Science and the Environment. Pagiola, S. (2008).  Payments for environmental services in Costa Rica.  Ecological Eco- nomics, 65 (4), 712-724. Pagiola, S., Arcenas, A., & Platais, G. (2005). Can payments for environmental services help reduce poverty? an exploration of the issues and the evidence to date from Latin America.  World Development, 33 (2), 237-253. Pretty, J., & Ward, H. (2001).  Social capital and the environment.  World Development, 29, 209-227. Putnam, R. (2000). Bowling alone: the collapse and revival of American community. New York: Simon and Schuster. Putnam, R.D., Robert, L., & Raffaella, N. (1993). Making democracy work: Civic tradi- tions in modern Italy. Princeton, N.J.: Princeton University Press.  19  Reedy, H. (2003).  Feasibility assessment of afforestation for carbon sequestration: ex- ploring options for aggregating and selling afforestation carbon credits from small landowners.  Canadian Forest Service: Pacific Forest Center. Richardson, G. (2005).  The prudent village: risk pooling institutions in medieval English agriculture.  Journal of Economic History, 65, 386-413. Schechter, L. (2007). Traditional trust measurement and the risk confound: An experi- ment in rural Paraguay. Journal of Economic Behavior and Organization, 62(2), 272-292. Smith, J., & Scherr, S. J. (2003).  Capturing the value of forest carbon for local livelih- oods. World Development, 31(12), 2143-2160. Sobel, J. (2002). Can we trust social capital?  Journal of Economic Literature, 40(1), 139-154. Uphoff, N. (2000). Understanding social capital: Learning from the analysis and expe- rience of participation. In P. Dasgupta, & I. Serageldin (Eds.), Social capital: A mul- tifaceted perspective (pp. 215-249). Washington, D.C.: World Bank. University of British Columbia-Faculty of Graduate Studies (2009).  Manuscript-based thesis.  Retrieved on October 09, 2009, from http://www.grad.ubc.ca/students/thesis/index.asp?menu=002,002,000,000. van Kooten, G.C., Shaikh, S.L., Suchanek, P. (2002). Mitigating climate change by plant- ing trees: the transaction costs trap.  Land Economics 78 (4), 559–572.  20  Wunder, S. (2005). Payments for environmental services: some nuts and bolts.  CIFOR Occasional Paper No. 42.  Retrieved on April 15, 2010, from http://www.cifor.cgiar.org/publications/pdf_files/OccPapers/OP-42.pdf. Zak, P.J., & Knack, S. (2001).  Trust and growth.  Economic Journal, 111(470), 295-321.    21    Chapter 2  Participation in the World’s First Clean Development Mechanism Forest Project: The Role of Property Rights, Social Capital and Contractual Rules2  2 A version of this chapter has been published.  Gong, Y., Bull, G. and Baylis, K. (2010).  Participation in the world’s first Clean Development Mechanism forest project: the role of property rights, social capital and contractual rules.  Ecological Economics, 69 (6), 1292-1302.  22  2.1. Introduction Clean Development Mechanism (CDM) forest projects are intended to be both a Payment for an Environmental Service (PES) and an instrument to facilitate sustainable develop- ment in developing countries3 (Haupt and von Lupke 2007; Smith and Scherr 2003).  To facilitate sustainable development, these projects should benefit both the environment and rural land users, provided the land users are willing and able to participate.  However, high transaction costs4 and large uncertainties often bar small-scale and poor land users in developing countries from making what are inherently long-term and often expensive in- vestments in forestry (Pagiola et al. 2005; Wunder 2005). While project design can par- tially address these obstacles to participation (Pagiola et al. 2005; Tschakert et al. 2007; Wunder 2005), the relationship between the buyer and seller of the environmental service is affected by the institutional framework (Muradian et al 2010).  Using the world’s first CDM forest project in Guangxi province in China as our case study, we illustrate how social capital, property rights and contractual rules impact the local land users’ willing- ness to participate. Forest carbon sequestration projects face many formidable obstacles due, in part, to substantial transaction costs, large risk and uncertainties, long time horizons and high es- tablishment costs (Cacho et al. 2005; Haupt and von Lupke 2007; Keenan et al. 2004; Milne 1999).  Uncertainties arise from ambiguous property rights, vague or rapidly  3 In theory, CDM can be considered a particular type of PES that uses economic incentives to both enhance the environmental service and provide a global public good (see Engel et al. 2008). 4 Usually a statement of the Coase theorem uses the term transaction costs.  Since we refer to transaction costs as a specific set of costs associated with project implementation, and not those costs associated with bargaining, we use the term bargaining costs to capture search, negotiation and, excludability, and costs of establishing property rights needed for a contract to be feasible.  23  changing government policies and unknown carbon market prices (Keenan et al. 2004; Reedy 2003). Additionally, there is underlying risk from human-induced and natural dis- asters (Keenan et al. 2004; Reedy 2003).  Since many small-scale poor land users in de- veloping countries have only small plots of land and credit constraints (Just and Zilber- man 1985), they cannot easily absorb negative shocks.  Thus, risk could act as a formida- ble barrier to project participation. Pooling or bundling individual activities and signing collective contracts with groups of smallholders spreads transaction costs over a large group and can be a practical means for small-scale land users to participate (Grieg-Gran et al. 2005). Nonetheless, pooling requires collective action, the success of which largely depends on a mix of property rights, contracts and social capital (Ostrom 1991; Pagiola et al. 2005; Wunder 2005). These three components are not independent. Contracts operate within a regime of prop- erty rights (North 1990) and social capital can determine individual’s ability to enforce contracts through social structures (Kumar and Matsusaka 2008). Thus, to design a suc- cessful CDM forest project, we need to understand the important roles played by social capital, property rights and contractual rules in facilitating participation5. Our case study is the world’s first CDM forest project, Facilitating Reforestation for Guangxi Watershed Management in Pearl River Basin in China6 (hereafter called the  5 Although participation can refer to a broad range of very important issues such as empowerment or legi- timacy, we focus on the narrow definition of participation as agreement to include one’s land in the CDM program. 6 China has been taking the leading role in the global CDM market, accounting for 48% of global volume of certified emission reductions by 2012 and is the current largest host country for CDM projects with respect to projected volume from registered projects (UNEP Risoe Center, 2009).  In addition, over the last 20 years China has successfully expanded its forest resources through afforestation and reforestation.  It is  24  Guangxi CDM project). It is designed to reforest 4,000 ha of remote and seriously de- graded lands. The project also has an integral component of enhancing local livelihoods. In spite of its success as of September 2007, when that the reforestation was due to be complete, only about 55% of the designed land was reforested. Major obstacles to refo- resting the remaining 45% of the land include difficulties in rehabilitating very remote and seriously degraded lands, withdrawal of local land users unsatisfied with the original contractual rules, low trust among involved stakeholders, and unresolved land tenure is- sues. The successes and the obstacles faced by the project are essentially attributable to the roles played by property rights, social capital and contractual rules6.  Therefore, this paper focuses on analyzing these three aspects in the design and implementation of the project. In the next section, we first discuss a CDM project as a form of a simple Coasean contract7, and present general requirements for land-user participation. Next, we broaden the model to include social and legal institutions such as property rights and contractual rules, and develop two propositions about the roles played by these institutional factors. We then give some background about the project itself, and discuss our data.  Last, we use our data to discuss our propositions. Given the scarcity of quantitative information,  therefore not surprising that China has become a hot spot for testing the theory and technology of CDM forest projects. 7 Simplified contractual structure is one important factor that mitigates the effectiveness of the implementa- tion of China’s sloping land conversion program, the largest reforestation program in the developing world that uses a public payment scheme to directly engage millions of rural households as core agents of project implementation (Bennett, 2008).  It is suggested that importance of local communities’ participation in program design and implementation and household granted full autonomy in participation choices should be recognized.   25  the propositions are evaluated using anecdotal evidence and information collected by per- sonal interviews and focus group discussions in the project area. 2.2. Requirements for Participation in a CDM Forest Project PES schemes are justifiably diverse.  To enable participation, they must not only provide sufficient financial incentives to land-users, they also need to incorporate the complex institutions in which they operate (Kosoy and Corbera 2010; Murandian et al.2010; Vatn 2010).  As with all PES projects, CDM forest projects are influenced by both transaction costs and risk-sharing arrangements and therefore involve the interaction of agents, which occur in the context of formal and informal institutions(Kosoy and Corbera 2010), be they contracts or a set of norms.  Therefore, a narrow Coasean approach that purely fo- cuses on financial incentives to solve the problem of externalities without considering institutional factors, such as property rights or social capital, will miss important compo- nents needed for a CDM scheme to be successful.  In particular, we argue that misunders- tanding these institutional factors has led to excessive bargaining costs and under- participation in the forest CDM regime in China.  In this section, we start with a simple Coasean framework, and then extend it to include institutional components, with a focus on facilitating land-user participation. For PES projects to induce changes in land use, at a minimum, land-users must partic- ipate (Pagiola et al. 2007). From a Coasean standpoint, there are two key requirements for efficient participation.  First, the combination of social surplus to be garnered by the change has to be sufficiently large and second, the bargaining costs have to be sufficient-  26  ly low, so that the residual surplus after bargaining is still positive (or at least not nega- tive). Although the Coase theorem officially addresses an efficient outcome, we focus here on the weaker requirements for simple land-user participation. We make the reasonable assumption that land-user participation is needed for an efficient outcome, assuming there are true benefits, measurable or immeasurable, to be gained by CDM-type projects8.  We secondly make the simplifying assumption that the price set by the carbon purchaser re- flects the true cost of the externality, and that land-users receive any surplus. We walk through the components of that surplus and list a simple requirement for it to be non- negative. We next discuss the components of bargaining cost, and what is needed for them to be low enough to allow a transaction to take place. For land-user i to participate, her expected utility from participating, E(Upi ) is at least as large as the expected utility from the next best alternative land use, or reservation utili- ty, denoted as E(Uri)  ( ) ( )pi riE U E U≥        (2.1) Utility is assumed to be a function of the revenue from participating in the project, less the physical cost, thus )( cpqUU ii T pi θδ −= , where δ is a discount factor, T is the  8 In practice, the full social surplus generated by PES-style payments is often not measurable, in part be- cause of ancillary costs or benefits associated with landowner activities (Norgaard & May, 2009-this issue). One can think of the measurability as another, perhaps infinitely large, transaction cost.  However, given the carbon buyer in this case has set a fixed price for carbon, we assume that this price at least reflects the buyer’s perceived surplus from the transaction.   27  time before the carbon payment is made, θi is a parameter running from 0 to 1, reflecting the distribution of costs among land users, and revenue is stochastic. Further, assume res- ervation utility, Uri comes from income associated with the second-best land use, yri, and that while this income is also uncertain, at least part of it is expected in the same time- period as the decision is being made (such as from this year’s crop sales).  For simplicity, assume the physical costs of the project, such as site preparation, road building and tree- planting are known. Thus, a necessary requirement for participation can be written as: ] 1 )1( [()][( δ δθδ − −≥− T ri ii T yEUcpqEU      (2.2) Components of equation (2.2) are as follows: [1]  To induce participation, the expected payment needs to be large enough to cover the project physical costs and the opportunity cost. In other words, land users will only be willing to participate in the project if their par- ticipation makes them better off (Pagiola et al. 2005). As such, the expected utility from participation needs to be at least equal to the expected utility of their second-best land use (Engel et al. 2008; Pagiola et al. 2005; Pagiola et al. 2007). Heterogeneity of physical costs, such as site preparation, and of opportunity costs complicate matters. To induce high-cost land users to join, [1] prices either need to be sufficient to include all participants; or [2] payments need to be differentiated in space and/or across agents on the basis of costs (Engel et al. 2008; Pagiola 2008).  If an average price for carbon is used, high-cost land users will not participate.  Similarly, those with  28  high opportunity cost (or reservation utilities), will not participate.   However, making higher payment to high cost land users to encourage their participation only makes sense if the benefit of their participation still exceeds the payment, but that payment can be re- distributed among participants to increase the amount of land.  If the opportunity costs of participation are known, these differential payments can explicitly be integrated into the contractual design, or, if unknown, they can be implemented through a bidding system. [2]  Along with sufficient or differentiated payments, to meet the participation con- straint in equation 2.2, the transaction cost of participation needs to be relatively low. Differentiated payments are not costless, however.  Along with requiring more infor- mation about land-users, such as opportunity cost, differentiated payments may raise the specter of unfairness (Pagiola et al. 2005).  If one village sees its neighbors being paid twice as much for their land just because the other village is closer to town, they might perceive the program as being unfair.  To fully cover differences in opportunity cost, of- ten better-off land users will be paid more than their poorer cousins purely because they are blessed with more lucrative options for their land.  Perceived fairness is particularly important in China (Bhalla 1992). Even with a payment that satisfies equation 2.2, the transaction costs of forest seques- tration projects could be large enough to deter participation.  Large forest establishment cost and fixed transactions costs associated with carbon sequestration projects often pro- hibit poor and small-scale land users from participating in the CDM forest projects (Kee- nan et al. 2004; Reedy 2003). One way to reduce transaction costs is to bundle individual  29  (carbon sequestration) activities to create the economies of scale and provide a viable means for the small-scale land users to participate (Australian Greenhouse Office 2005; Grieg-Gran et al. 2005; Wunder 2005; Pagiola 2007).  Bundling requires an incentive- compatible contract detailing income sharing arrangements or high degree of social capi- tal.  However, due to the impossibility of writing detailed long-term contracts or low trust for their trading partners, land users may choose to hold up their participation in the bun- dled CDM forest project.  We will revisit these points shortly. A further component of equation 2.2 is the discount factor9 faced by the land-user. Note that if the carbon payment is expected later than the arrival of the first income from the second-best use, a high discount factor will shrink the utility of the expected carbon payment more than the utility net present value of the second-best income stream.  Fur- ther, note that a typical CDM forest project requires a high rate of financing upfront, while revenues and benefits occur several years hence (Haupt and von Lüpke 2007).  To participate, sellers need to have the liquidity to make this up-front investment, or buyers may need to pay this initial cost.  Thus, having the only payment for carbon conditioned on the delivery of future carbon credits will discourage participation among the smaller land users. [3]  The payment schedule adopted by buyers from the developed countries need to take the sellers’ constraints into account, perhaps through some form of upfront payment.  9 We have modeled the discount factor in its most simple form here.  However, more complex and perhaps accurate measures of discounting, such as hyperbolic discounting, will have the same effect of making the future carbon payment relatively less attractive than the income stream from annual crop harvests.  30  [4]  Credit constraints also mean that land users are sensitive to the risk of both the carbon payment and the second-best land-use option. [5]  Land users’ anticipated bargaining power in trading relation needs to be high enough or their outside options ought to be few in order for them to have incen- tives to participate. Land users may still decide to hold up their participation due to their anticipated low bargaining power ex post in their relation with trading partners, even though they could gain a potential profit from investing their barren lands to the CDM forest project through certain appropriate arrangements, such as bundling, high enough payment made to them and flexible payment methods to consider differentiated costs.  Their decision of holding up participation could largely be attributable to the impossibility of writing, monitoring, and enforcing complete contracts (Joskow 1988; Williamson 1985) together with asset specificity of the investment.  The asset specificity in the particular context of the CDM forest project lies in the fact that CDM forest project requires barren lands be planted with specific tree species for particular purposes, essentially including carbon sequestra- tion..  Suppose the land users are able to participate in the CDM forest project through bundling arrangement with the help of a bundler, such as a local forest company, they are more deeply involved in the project than their trading partner, the bundler.  Given the in- complete contracts, if they are uncertain of their bargaining power in the relation with the bundler ex post, they would be worried that the bundler would demand for a higher share of the profits in the future and thus decide to hold up their participation in the project at the current stage.  31  The land users could also hold up their participation if they anticipate some potential out- side options for their land.  When the land users foresee some potential outside options in the future, their reservation utility, which is essentially the right-hand component of the equation 2.2, will be increased.  As a result, they need a higher payment for them to par- ticipate in the CDM forest project.; otherwise, they choose not to participate in the project at the current stage. [6]  Related to both the expected revenue and property rights need to be sufficiently secure. Assume for the moment that there is some risk of the land users losing their property rights to the contracted land, and ( ) ( ( ))T TiE pq pE qδ α δ=  where α represent the proba- bility of retaining property rights and 0<α<1.  Unless a land user is confident of her prop- erty rights, she will need a larger payment or a larger potential surplus for her to be will- ing to participate. Note that the probability of retaining property rights not only affects a land user’s expected income from carbon, it also affects their income from other land uses.  Thus, an alternative approach is if the carbon contract itself can strengthen the land user’s property rights claim, increasing alpha not only on the carbon payment, but on ex- pected revenue from other land uses, the land-user may accept a much smaller carbon payment or no payment at all. If the first half of the Coase theorem deals with the available surplus, the second half of the Coase theorem deals with bargaining costs. Therefore, just having a positive sur- plus available to land users is not sufficient to ensure an efficient outcome, or, even a ne- cessary component of that outcome: participation. We take a broad view of bargaining  32  costs to explicitly include institutional factors often ignored in a Coasean framework. Bargaining costs include the standard search, negotiation and enforcement costs, where we think of part of the negotiation cost as being the cost of establishing and agreeing on property rights.  We posit that bargaining costs are decreasing in social capital and in formal institutions, such as clearly defined property rights and well-crafted contracts. Due to bounded rationality and opportunism of human behavior, relational contracts are often necessary (Williamson 1985).   However, in reality, contracts are typically in- complete (North 1990). In the presence of incomplete contracts and opportunistic beha- vior, trust, an integral part of social capital, can lubricate co-operation (Pretty and Ward 2001; Sobel 2002) and could play an important role in enforcing formal or informal con- tracts (Asquith et al. 2008; Ostrom and Nagendra 2007; van Kooten et al. 2006). From these necessary conditions, we develop the following related propositions for CDM forest projects: Proposition 1. Having positive surplus is not sufficient to result in an efficient agreement. This proposition is a simple outgrowth of the Coase theorem.  When bargaining costs are too high, even agreements that would appear to benefit all parties are not consum- mated.  We detail several examples of this in our results section. Proposition 2. Bargaining costs can be reduced either by high social capital, clear for- mal institutions, such as a clear delineation of property rights, or through a standardized contract.  33  A standardized contract, such as the formal pooling arrangement and share-holding system created for the Guangxi CDM project, can reduce bargaining costs and enable participation, even in the absence of other formal institutions or social capital. Corollary 1.  A standardized contract does not always have the flexibility to reach an efficient outcome. Because of heterogeneity of physical costs, opportunity costs and social capital, a standardized contract may not be incentive compatible for all land users, and therefore relying on the standardized contract may leave some parties out of the bargain. Corollary 2. With sufficient social capital, the lack of formally-defined property rights can be overcome. Even when property rights are not clearly established, if there is sufficient social capi- tal and sufficient potential surplus to be had from an agreement, an agreement can be reached. 2.3. Background Information of Guangxi CDM Project The Guangxi CDM project is designed to plant 4,000 ha of multiple-use forests on se- riously degraded and remote lands to reach multiple environmental and developmental objectives. It is implemented in Cangwu and Huanjiang Counties in China’s Southern Province of Guangxi (Figure 2-1).   34   Figure 2- 1:  A Location Map of Guangxi CDM Project  (Source: Guangxi CDM Project Development Document 2006)   2.3.1 Project Environmental and Development Objectives Using environmentally friendly techniques to plant a combination of six tree species, the project has the multiple objectives of sequestering carbon, enhancing biodiversity, reduc- ing soil erosion, and improving local livelihoods. Of the total designated area of 4,000 ha, 3,000 ha are to be planted with a mixture of five native tree species and 1,000 ha planted with one single tree species, eucalyptus (Table 2-1).     35   Table 2- 1: Multiple Tree Species to be Planted by the Project Tree species Cangwu Huanjiang Species ratio Rotation age (ha) (ha) (years) Eucalyptus 500 500 1:1 10:10 Sweetgum: Chinese red pine 0 1050 6:4 17:30 Sweetgum : Chinese fir 0 450 6:4 17:30 Chinese red pine : oak 900 0 6:4 30:7 Chinese red pine: Chinese Gugertree 600 0 6:4 30:17 Total 2000 2000 (Source: Guangxi CDM Project Development Document 2006) Remarks: In columns 3and4, the first figure refers to the first species to the left, and the second, to the second species listed in column 1.   The estimated total amount of carbon to be sequestered is 0.77 megatons (Mt) of CO2 equivalent (CO2e) over a 30-year crediting period (2006-2035). A relatively smooth an- nual net carbon sink curve is to be generated over the crediting period (2006-2035) by planting six tree species with different growth rates, carbon sequestration rates, and rota- tion periods (Table 2-1). For example, oak, eucalyptus and sweetgum, have higher growth rates and a shorter rotation period, leading to a higher rate of carbon sequestration in the early stage of the crediting period, while Chinese fir and Chinese red pine have lower growth rate in the early stage but longer rotation period and yield a higher total amount of carbon over the crediting period. Fig. 2-2 presents the annual and accumula- tive carbon sink of the project.  36   Figure 2- 2: Annual and Accumulative Carbon Sink by the Project (Source: Adapted from Guangxi CDM Project Development Document 2006).  Other environmental benefits, biodiversity enhancement and soil erosion control, are also expected to be generated by the project.  The forest coverage in project area will be increased by 1.34% as a result of the establishment of the 4,000 ha of multiple use forests. The regeneration of the previously seriously degraded land is expected to improve soil erosion control in the project area.   Given that some of the regenerated lands are adjacent to a national nature reserve, the regenerated forests can serve as corridors for wildlife and thus help enhance the viability of wildlife populations in the protected area (PDD 2006). Local livelihoods are to be improved mainly through direct transfers and new em- ployment in tree planting, weeding, harvest timber, collecting resin and forest manage- ment.  About 20,000 local farmers of 5,000 households are expected to benefit from the project (PDD 2006).  The expected total revenue from sales of carbon credits, timber and -100000 0 100000 200000 300000 400000 500000 600000 700000 800000t CO 2- e Year Net CO2 removal (tC) Carbon stock  37  pine resin is approximately US$ 5.5 million: US$ 3.5 million from sales of timber and pine resin; and US$ 2.0 million from sales of certified carbon credits. Local land users divide the expected revenue with three local forest companies at shares set under the share-holding system (details to follow). 2.3.2 Land Tenure Arrangements The 4,000 ha of land involved in the project (2,000 ha in Cangwu and 2,000 ha in Huan- jiang) have two types of land tenure arrangements: individually managed lands (1098.4 ha) and communal lands (2,901.6 ha). Approximately 55% (1098.4 ha) of the lands in Cangwu are individually-held and the remaining 45% (901.6 ha) are communal. All of the 2000 ha of lands in Huanjiang are communal lands. The individually managed lands have been contracted to single households under a 50-year contract period since early 1980s. The communal lands are under the management of natural villages10, which are sub-units of the administrative villages.   Many communal lands in Huanjiang do not have boundaries between 2 adjacent natural villages, resulting in tenure disputes.  Some communal land boundaries are also unclear in Cangwu. 2.3.3 Project Baseline and “Additionality” For a CDM project to deliver real carbon benefits, it must demonstrate “additionality”11, implying that the carbon sequestration added by a CDM project would not have occurred  10 The Guangxi project involves both communal lands managed by natural villages and lands managed by individual famer households.  In Chinese context, a natural villages are communities is a community where people live together, while an administrative village consists of several natural villages, some of which might be far away from each other. 11 To prove “additionality”, the project developers need to show that either the project is not economically or financially attractive without the additional payment, or it would not be able to overcome legal, technol- ogical or ecologically barriers without the income of carbon credits (Haupt and von Lupke 2007).  38  otherwise (Haupt and von Lupke 2007).  The stated baseline of the Guangxi CDM project is that the barren lands would “remain abandoned and degraded” (section B.2 in PDD 2006). Before the CDM forest project, most of the 4,000 ha of the lands were predomi- nantly covered with grasses or shrubs, except for about 35 ha of lands covered with scat- tered trees.  While these lands, which were mainly deforested in 1960s, are restricted to forest use under Chinese law, they have not been reforested before the CDM project due to the lack of private or public investment. The Guangxi CDM project is potentially additional in that without carbon financing, the project would not be economically attractive and would likely face legal, institutional or other barriers.  Before the CDM project, the private business sector had little incentive to reforest the seriously degraded and remote lands due to the low rate of return on the investment (section B.3 of PDD 2006).  Along with the carbon financing, the provincial and local governments contributed additional financial support and coordination to over- come some previous financial and institutional barriers12 faced by reforestation projects. In this sense, the provincial and local governments have acted as intermediaries, facilitat- ing the project.  In terms of economic returns, without the mix of carbon financing and government financial support, investment on regenerating seriously degraded and remote lands only has an internal rate of investment (IRR) of 8.53%, which is lower than the re- quired rate of return (12%) set by the Chinese government for forest investment (section B.3 of PDD 2006). With the additional revenue from carbon credits sold at a fixed price  12 Institutional barriers include the fact that local land users were not capable of directly finding and nego- tiating with the buyers regarding sales of carbon credits (section B.3 of PDD 2006).  Local governments play an important role of coordinating between the local forest companies and the local land users to form a share-holding system.   39  of US$4.5/ton of CO2e for the crediting period of 2006-2035 and with incomes from sales of carbon credits combined with government financing, the IRR of the investment increases to 15.02% (section B.3 of PDD 2006), a rate that makes forest investment at- tractive. 2.3.4 Pooling Arrangement and the Share-holding System The pooling arrangement and the share-holding system are unique components of the Guangxi CDM project. The pooling arrangement bundles barren lands from 27 villages13, 15 from Cangwu and 12 from Huanjiang, to form the total project size of 4,000 ha. The share-holding system was created between the local land users and three local forest companies. Fig. 2 presents the framework of the share-holding system.  13 The villages that were identified in the PDD were chosen based on the availability of relatively large pieces of contiguous barren lands in the villages.  In Huanjiang, pooling only involves communal lands. In Cangwu, the situation is more complicated: in 4 villages, all involved lands are communal lands; in 5 vil- lages, all involved lands are individually-managed lands; in the remaining 6 villages, both communal and individual lands are involved.   40   Figure 2- 3: A Share-holding System Created for Guangxi CDM Project  The share-holding system involves two groups of stakeholders: the three local forest companies, two from Cangwu and one from Huanjiang, and the local land users from 27 villages. The share-holding contract has three components: [1] the local forest companies are fully responsible for paying for project development and monitoring costs, production costs (including capital and labor) needed for reforestation and forest management, and take all responsibility for harvest and product sales, and providing necessary technical support and training to the local land users regarding the carbon project. Local land users only need to contribute their individually managed or communal barren lands. [2] The forest companies are allowed to establish 1,000 ha of eucalyptus plantations by leasing lands from local land users at the local land rent prevailing in 2005 (60 yuan/ha/year) for   Local forest companies  Capital& techniques Land Shared income  Carbon credit   Timber  Pine resins  Output  Shared income  Input  Local communities  41  a period of 2006-2035 and claim all revenues from these plantations, get subsidized loans and government financial support; [3] Local land users will receive 40% of the income from the future sales of timber, pine resin and 60% of the income from sales of carbon credits produced from 3,000 ha of forests with five native tree species other than eucalyp- tus, while the three local forest companies will claim all revenues from the sales of prod- ucts from 1,000 ha of lands to be planted with eucalyptus. 2.3.5 Contractual Arrangements between the Buyer and Sellers To pool the carbon credits and reduce transaction costs, collective contractual arrange- ments are signed between buyers and sellers. The collective contracts were signed fol- lowing three steps. First, the buyer, the World Bank’s BioCarbon Fund, signed a contract with Luhuan Forestry Development Company from Huanjiang County, who acts as an intermediary, represents all sellers under the share-holding system. Next, the Luhuan Fo- restry Development Company signed individual contracts with further intermediaries, the other two forest companies, Kangyuan and Fuyuan forest farms from Cangwu County. For communal lands, all three forest companies signed contracts with natural village leaders who then determined how revenue from the communal lands would be shared among their community members; for individual lands, the forest companies signed the contracts directly with household heads. The agreement follows a commodity model rather than an investment model14: the BioCarbon Fund pays a price of US$4.5/ton of CO2e to the sellers at the time of purchase  14 In the commodity model, buyers purchase CERs and the project developers (sellers) carry all risks, while payments in investment model are upfront (Pearson et al. 2006).  42  of certified emission reductions (CERs). This payment method indicates: [1] the agreed payment level does not differentiate among designated lands that have heterogeneous site qualities and establishment costs; [2] no upfront payment is made to the sellers, which means that the sellers, the local liquidity-constrained companies and the local land users living in the remote areas, carry all potential risks. 2.3.6 Financial Arrangements for the Project Subsidized loans and government financing are two major sources of funding for the project.  Of the total project investment of US$2,822,282, about 40% (US$1,128,913) was financed through a subsidized loan at a rate of 6.21% to the local forest companies, about 40% (US$ 1,128,913) was financed by the Government of Guangxi and about 20% (US$564,465) was funded through commercial loans. The financial arrangements are intended to make the project more appealing to small- scale local forest companies. Large-scale commercial forest companies, who are more financially competitive than the small-scale local forest companies, were reluctant to in- vest in a CDM forest project, which has much longer return period of investment and a lower rate of return than commercial plantation projects. In contrast, small-scale local forest companies, who cannot compete with large-scale commercial forest companies for reforestation of accessible areas, might see this project as a niche market and be moti- vated by the premia associated with investing in reforestation on remote and degraded lands for carbon sequestration.  43  2.4. Data and Sampling Methods Both primary and secondary data are used in this analysis. Primary data were collected using questionnaires in September 2007 through personal interviews with local communi- ty leaders and focus group discussions, which were conducted in 14 villages (i.e., over 50% of all villages involved in the project).  The anecdotal evidence from these interviews is used here to discuss our propositions. We used stratified sampling (Dillman et al. 2008) to choose villages for surveys and village focus group discussions. First, the villages were stratified into three groups ac- cording to project implementation progress: [1] villages with sound (full or almost full) project implementation; [2] villages with a medium level of implementation; [3] villages with no implementation. Second, within each stratum, random sampling was applied to choose villages for the survey. The main purpose of conducting personal interviews with village leaders was to collect primary data to update the field inspection data collected by the local government agencies. These data were supplemented by focus group discussions held with government staff members who were coordinating the project, and managers of the local forest companies.  The discussions were conducted using semi-structured ques- tionnaires. Information collected included socio-economic conditions, social relationships and land tenure issues in surveyed villages and contractual arrangements. Secondary data include government reports, field data previously collected by the lo- cal government regarding the implementation of the project and raw data used for the de- velopment of the PDD submitted to the CDM Executive Board. Secondary data are used  44  to evaluate project implementation outcomes and recalculate the net present value (NPV) of the project under variant scenarios15. 2.5. Results and Discussion The Guangxi CDM project is successful in a number of ways. It provided local land users and local forest companies sufficient financial incentives to invest, and through pooling, it reduced some of the transactions costs and made the project accessible to individual small-holders. However, even though reforestation plan should be nearly completed by September 2007, large areas of the project have not yet been reforested, and unless changes are made, at least 35% of the designated lands will likely remain unforested. The reasons are high costs of regenerating seriously degraded lands, constraining contrac- tual rules and low or weak levels of social capital and property rights disputes.  Each of the successful and unsuccessful features is discussed below. 2.5.1 Project Successes Local Land Users’ Investment Incentives Given the mix of government subsidies and carbon credits, the overall project is poten- tially profitable for all parties. We calculate the net present value (NPV) of the overall project profit, the NPVs of local forest companies and local land users by including land  15 The PDD of the Guangxi CDM project does calculate the NPVs.  We suspect that the original NPV of the overall project calculated in PDD might be misleading because a review of the original analysis data found: [1] on the cost side, the original analysis did not consider the opportunity cost of the lands and the transaction costs; [2] on the benefit side, the original financial analysis used a lower bound of the prices for timber products and pine resin that were prevailing before 2003.   Therefore, we recalculate the NPVs un- der several scenarios.  45  opportunity cost, estimated by local land rent prices, specific transaction costs16, and ad- justed prices of the product outputs. The calculations are based on a combination of two different levels of land rentals and two levels of product prices: [1] land rental prices pre- vailing in 2005, when the project was in its development stage, and those prevailing in 2007, the time after the prices were driven up by competitions in the local land markets; [2] a lower bound17 of the prices of timber and resins, and their average product prices in the last 15 years in the project area. A discount rate of 12%, which is the threshold rate of return (RRR) set by the Government China for CDM forest project to get approved (PDD, 2006), was used.  We find that under most scenarios, the returns are over the 12% needed. The NPVs of the overall project profit is positive at a discount rate of 12%, which is the internal rate of return required by the Government of China to approve the CDM for- est-based carbon sequestration projects. The calculated NPVs for the local forest companies are about US$1,048,211 at a dis- count rate of 6.21%, which is the subsidized rate of interest, and US$126,802 at the dis- count rate of 8%, which approximates the weighted average of the subsidized loan rate and the commercial loan rate for forest investment. From a purely financial point of view, ignoring risk and credit constraints, local land users should be willing to participate in the project, since they can gain a positive NPV of  16 The transaction costs of the Guangxi CDM project include project development cost, validation cost, annual verification cost and due diligence cost, associated with the carbon project. They do not include the cost of establishing property rights, or other search, negotiation and enforcement costs, and are thus a sub- set of the bargaining cost we discuss earlier. 17 The lower bound of the prices is the lowest prices of resins and timber between 1992 and 2007 in the local market in project area.  46  profit by leasing their lands to the CDM project and sharing income with the local forest companies. Table 2-2 presents local land users’ average and lower bound of the NPV from participating in the CDM project. Note that the NPV participating in the project is all positive at different discount rates even at low product prices.  Table 2- 2: Local land Users’ Potential Profit Scenarios NPV at 8% NPV at 12% NPV at 20% (US$) (US$) (US$)  NPVs of profit: low land rent prevailing in 2005; high product prices 2,531,472 1,449,493 650,748  NPVs of profit: high land rent prevailing in 2007; low product prices  1,370,804 619,010 137,424 (Source: Guangxi CDM Project Development Document and Field Survey Data)  In spite of the potential profitability, the project has a shortfall of implementation. As noted above, by the time of surveys in September of 2007, only 55% (about 2,210 ha) of the overall plan (4,000 ha) had been fulfilled. This low level of participation, if not due to lack of direct financial incentives, must be due to high costs of participation or the hold- up problem in participation. We first consider the explicit transaction costs of the project, and then move to other components of participation cost in turn. Transaction Costs As noted above, the Guangxi CDM project’s carbon pooling leads to a relatively low lev- el of explicit transaction costs. Based on the same data used for the PDD, the total trans- action costs of the project are estimated to be US$ 971,817, which includes project de- velopment (US$ 40,000), validation (US$ 20,000), annual verification (US$ 400,000),  47  due diligence by BioCarbon (US$ 120,000) and administrative costs (US$ 391,817). The fixed transaction accounts (US$ 180,000), including project development, validation and due diligence by BioCarbon, account for slightly less than 20% of the total transaction cost of the project. With the total amount of 791,957 tons of CO2e to be produced from the pooled 4,000 ha of lands, the average transaction costs of the project are estimated at about US$1.23/tCO2e if the designated lands are fully implemented; about US$1.64/tCO2e18, assuming that the accomplished 55% of the tree planting plan would proportionally deliver 435,576 tons of CO2e, 55% of the expected 791,957 tons of CO2e. The relatively low level of transaction costs19 can be largely attributable to the roles of pooling in creating economies of scale. Reduced Bargaining Costs and Accessibility to Small-scale Producers The Guangxi CDM project’s share-holding system can reduce the bargaining costs and makes the project accessible to the poor and small-scale local land users. Land users in the project area, especially those in Huanjiang, are among the poorest in China.  Given the remoteness of the designated lands, project establishment costs have to include costs of road building, which would not be either financially or technically feasible for local land users. However, under the share-holding system, the local land users only need to  18 It is only a rough estimate of the average transaction cost by simplifying the assumption that fulfilled 55% of tree planting plan would proportionally deliver 55% of the expected carbon credits.  Given that the un- planted lands are mainly those designated for planting native tree species, the true amount of carbon to be eventually sequestered is more complicated than the roughly estimated.  In addition, with improvement in project implementation, the eventually delivered amount of carbon credits would be more than 55%.  None- theless, our estimate could give a range of the average transaction cost. 19 So far, there has been only one single research conducted by Milne (1999) to estimate transaction costs of a range of forest carbon projects.  Milne (1999) estimated that the average transaction cost of communi- ty-based forest carbon sequestration projects is between US$ 1/tCO2e and US$ 235/tCO2e; the average transaction cost of Actions Implemented Jointly (AIJ) forest carbon projects implemented in other develop- ing countries is about $5.41//tCO2e.  48  allocate their lands to the CDM project while the local forest companies assume costs and take all responsibility for forest regeneration, road building, forest management and sales of the products20. As we note in proposition 2, contracts are one way to reduce bargaining costs. In the case of the Guangxi CDM project, the share-holding system plays a role of enabling local communities to take collective action for joint benefit, even when the social capital to facilitate the collective actions in the local areas is not high. This proposition can be demonstrated by the successful roles played by share-holding system in Xinlong Village and Datong Village in Cangwu County. These villages used to have high frequency of field fires (2-3 times per year), which were largely caused by village members’ traditional practices of burning crop stalk for fertilization for better agricultural productivity. The high risks of fire incidences directly led to lack of reforestation activities undertaken on barren lands, which were deforested in early 1980s. Interviewed village leaders stated that: [1] village members had little in- centive to plant trees, because of the absence of voluntary groups or norms of punishment on those who ruin young trees by whichever means in their villages; [2] almost no com- mercial forest companies had been to their villages to rent the land due to the presence of high fire risks in their villages; [3] weak norms of mutual reciprocity in the villages had prevented attempts made by village members to regenerate communal lands. While some village members have their family members working in cities or earning off-farm in-  20 However, the share-holding system could also create a hold-up problem due to the local land users’ antic- ipated weak bargaining power in their trading relationship with the local forest companies.  Detailed dis- cussions follow shortly.  49  comes and have little use of and low labor supply for regeneration of the barren lands, especially the communal barren lands, some other village members, who have few out- side options, are motivated to regenerate the communal barren lands but they were fre- quently opposed by villagers who saw no benefits, and were afraid of losing their implicit property rights. The law in China specifies that “whoever plants trees can own the trees”. Thus, the fear was that the village members who plant trees on communal lands could possibly claim property rights on planted trees and all future revenues from tree plantings. The village leaders said during the interview: “some members in our villages do not like others to benefit from communal barren land although they do not have much use of the land”. The explicit share-holding system allowed villages to overcome this barrier allowing communities to obtain jointly-held benefit from participation through their contribution of barren lands to the carbon project. The members in these two villages foresaw that they would get a share of income in the future by investing the lands under the share- holding system and thus agreed to participate in the project.  As a matter of fact, all des- ignated barren lands for the carbon project in Xinlong Village and 97% of the designated lands in Datong Village had been successfully regenerated by the time of surveys. Contribution to Equity The increased accessibility of the project to local land users, who are poor communities or small-scale farmer households managing an average land size between 0.3 ha/family to 7.9 ha/family in surveyed villages, has important implications for equity.  Firstly, as dis- cussed earlier on, given the remoteness of lands involved in the CDM project, road con-  50  struction is necessary, which conversely can greatly benefit remote villages from im- proved road access. Secondly, the land users’ participation in the CDM project make it possible for them to generate incomes by contributing their barren lands, which otherwise would continue to be barren or degraded.  Thirdly, over 1,600 households from 4 ethnic minority groups are involved in the CMD project (Guangxi Forestry Bureau 2005).   A good portion of people from these ethnic minority groups can barely speak Mandarin and have few economic opportunities beyond their own communities (Guangxi Forestry Bu- reau 2005).  Therefore, the participation of ethnic minorities in the CDM project brings economic opportunities to the ethnic minority groups. 2.5.2  Project Shortcomings Given the evident success of the Guangxi CDM project, why is so much land still unfo- rested?  Table 2-3 lists the reasons for the lack of tree plantings based on the field inspec- tion data from the local government agencies and primary information from focus group discussions. Table 2- 3: Reasons for Unaccomplished Plans Reasons Cangwu (ha) Huanjiang (ha) Overall (ha) Proportions in total 1. Restrictive contractual rules and insufficient trust causing local land users to withdraw 305 64 369 21% 2. Impossible to regenerate extremely infertile lands  0 257 257 14% 3. High cost of regenerating remote and degraded lands 0 639 639 36% 4. Privately-held lands that were planted to crop trees 14 49 63 4% 5. Land tenure disputes combined with weak social capital 49 134 183 10% 6.Other reasons 253 16 270 15% Total 621 1159 1780 100%   51  As shown in Table 2-3, major reasons for reduced tree planting include: [1]  About 369 ha, 21% of the unplanted portion of the project, of lands (mostly indi- vidually-managed lands in Cangwu), have been withdrawn from the original con- tracts with the local forest companies. Withdraw of these 369 ha of lands from the original contracts could be attributed to the following two reasons.  The first reason is the increase in the opportunity cost of par- ticipation.  After the contract was signed, lands in this area of Cangwu have seen an in- crease in rent due to competition for barren lands from commercial forest companies to develop eucalyptus plantations, indicating that outside options of the barren lands became available with the commercial companies entering the local land rental market.  As a re- sult, the land-users’ opportunity cost of participation has been increased and they have decided to hold up their participation by renegotiating for a larger share of profits from timber products, resins and carbon credits. Perhaps ironically, the higher rental rate offered by the commercial forest companies for eucalyptus plantation development indicate that the commercial companies may well be willing to forest these lands without the carbon money, indicating a shifting baseline for the project. Thus, these lands may no longer be additional. However, the reality is that right now, no one is planting them as local land users continue to negotiate with the local  52  forest companies21.  The second reason is insufficient trust of certain communities in the local forest companies.   Detailed discussion on this point will be followed shortly. [2]  At least 14% (257 ha) of the area designated for reforestation will not be regene- rated due to the extreme infertility of the severely degraded lands that were not identified during the project development stage. [3]  Over one-third of the unplanted lands face particularly high cost and technical dif- ficulties associated with reforestation. These lands are seriously degraded and ex- tremely remote communal lands in Huanjiang. [4]  Other privately-held land was used to plant crop trees, specifically oranges in Cangwu. After the project was signed, orange prices increased, raising the oppor- tunity cost of participating in the CDM project for these producers.  As crop trees themselves sequester carbon, as with those lands in [1], these lands no longer meet the additionality requirement22. [5]  Unresolved land tenure disputes contributes to 10% (183 ha) of unfinished plans. The obstacles set by unresolved land tenure disputes for the local communities for lo- cal communities to reach mutually agreed ratios for income-sharing could be possibly alleviated if the local communities have sufficient social capital to make coordination among themselves, regarding the sharing to the expected income.  Nonetheless, in some  21 One concern is that eucalyptus, being non-native, might not have the ancillary benefits associated with red pine or other native species. 22 If one is concerned about equity, one might see the potential for a carbon contract that paid these produc- ers for the carbon sequestered by their fruit trees.  53  administrative villages, a lack of the social capital makes this potential solution impossi- ble. [6]  About 10% (171 ha) of the unfinished plan is a result of a behind schedule in tree planting because of villages holding out for road-building before allowing plant- ing to proceed (part of reason 6 in the Table 2-3). Although it is certain that at least 14% of the area slated for reforestation will not be regenerated due to the extreme infertility of the severely degraded lands23, according to the information from the focus group discussions, about half of the unfinished tree plant- ing plans, subject to reasons 1, 3 and 5 in Table 2-3, could be planted if more favorable contracts were offered to the local land users, better coordination made for land dispute resolution or stronger technical support provided to regenerate seriously degraded lands. However, as noted above, there is a question as to whether those lands in [1] would still satisfy the additionality requirement, since the alternative use driving up their reservation utility is also forestry-related. Essentially, the major elements leading to current failures can be categorized as: [1] unfavorable income-sharing ratio between local land users and local forest companies; [2] inability of the buyer’s homogeneous payment level to consider the heterogeneity of des- ignated lands and the absence of the upfront payment; [3] low level of some local land users’ trust in their trading partners, the local forest companies; [4] weak social capital combined with the unclear delineation of the land tenures.  23 One option would be for the CDM Executive Board to allow an equivalent area of lands located outside the project boundaries to replace the extremely infertile lands for reforestation with the designated tree spe- cies.  54  Restrictive Contractual Rules The homogeneous payment level and absence of upfront payment are not sufficient to induce participation on the part of the sellers, and largely explain why very remote and marginal lands were left unplanted. For example, about 639 ha of land, all of which are communal lands in Huanjiang, have no road access, and are located at relatively high alti- tude have not been reforested. As discussed in Section 2.3.2, it is these very lands that explicitly demonstrate the project’s “additionality”. Successful reforestation on these lands can reach multiple environmental and developmental objectives, including ecologi- cal rehabilitation, delivery of carbon benefit and local livelihood enhancement. And due to their remote location, these lands would clearly not be reforested without the CDM project. Nonetheless, as reflected by the participants in focus group discussions, tree planting on these lands require road construction and the estimated total road building cost for these lands is about US$ 79,236, which almost accounts for about 3% of the re- quired total investment of the project, while the required investment laid out in PDD for road building, fencing and site preparations in 4,000 ha of the whole project area account for 5% of the total investment. In addition, these lands are susceptible to the risks of low survival and growth rate of young trees, which will directly lead to low expected quantity of carbon credits. Focus group participants revealed that the original establishment costs were estimated based on average costs rather than considering differentiation across different types of lands when the project was designed. However, when it came to the implementation stage, the local forest companies found that the establishment costs of these lands were much higher than expected. Since the commercial banks are reluctant to give loans to the fore-  55  stry sector, the local forest companies have also had difficulty in fully mobilizing the 25% of the required project development and reforestation costs, which were not supplied by the government. As a result, with existing payment methods offered by the BioCarbon Fund, local forest companies, who face liquidity constraints, were inclined to first refor- est relatively accessible lands but leave the 639 ha of very remote and marginal lands un- planted. The above example goes to proposition 2, corollary 1. Although having a standar- dized contract may reduce bargaining costs, the standardized contract may not be flexible enough to facilitate an efficient outcome. If these lands can produce enough carbon to make their reforestation worthwhile, there are several ways a contract could be designed to reforest these lands. First, payment on these higher cost lands could simply be larger. Second, reforestation of these lands could be tied to purchasing carbon credits from more accessible lands. Third, the large upfront costs could have been paid by BioCarbon Fund for these high cost lands. Last, instead of offering a fixed price, one might look to a bid- ding process, where the forestry company offers each parcel at a price that will cover their costs of reforesting that land. Incomplete Contracts and Social Capital When contracts between local land users and local forest companies are incomplete, if local land users, perhaps rightly, do not trust their local forest companies, they may be- lieve that the companies will “act opportunistically” and reduce the land-owner benefit after the contract is signed.  As the individual farmers anticipate a weak bargaining power after they make the specific investment of their lands to the CDM forest project, they  56  would choose to hold up their participation.  In addition, one can think of the lack of the local land users’ trust in their trading partner as evidence of low bridging social capital24, where bridging social capital refers to the networks across groups that enable members to reach outside sources of information, support and resources (Putnam 2000; Narayan 1999). As a result, land-owners decide not to enter the incomplete income-sharing con- tracts with the local forest companies, even though under the terms of the contract, their participation would be profitable. This situation can be illustrated by the anecdotal evi- dence obtained from Dayan Village, where the local land users’ participation rate is ex- tremely low. Dayan Village, located in Shatou Township in Cangwu, is to be planted with a mix- ture of Chinese red pine and oak trees on 86.7 ha of individual lands and tree planting was to be accomplished by spring of 2007. However, by September 2007, no tree plant- ing had occurred in the Village. The local forest companies eventually agreed to provide the Village with a special contract that specifies that the companies are responsible for tree planting, but the local land users can claim all incomes from sales of timber, resins and carbon in the future25.  In spite of this, the local land users in Dayan Village still hesi- tated to enter the contract, even though it was clearly to their financial advantage. Rather, they preferred to plant trees on their own with necessary technical and financial support from local forest companies or local forest agencies. When asked during village-level fo- cus group discussions, village members said that they did not believe that they would get  24 Note that low bridging social capital does not imply low intra-village or bonding social capital.  In fact, by collectively agreeing to forgo these revenues, the village might be thought to evidence a high degree of cohesiveness and thus high bonding social capital. 25 Although the forest companies get even a negative profit from tree plantings in Dayan village, they can still get a positive profit from the overall project through income-share across villages.  57  the full income from the trees if the local forest companies were involved in tree planting. The villagers believed that if the local forest companies are involved in tree planting, the companies would have a reason in the future to claim any revenues, even if local forest companies promise to stick to the contract. As a result of local land users’ lack of trust in the forest companies, the local land users decided not to participate in the CDM project. This example goes to the converse of our second proposition, indicating that lack of trust raises bargaining costs26. Weak Social Capital and Property Rights Disputes Land tenure disputes combined with weak social capital are obstacles facing some local land users in making successful collective decisions on land investment and income dis- tribution.   As an example, consider the following case of Cuishan Village, where land tenure is in dispute and social capital is low. Cuishan (administrative) Village has weak property rights over its communal lands since the boundaries of these 426 hectares are unclear.  The natural villages initially agreed to participate in the CDM project when the project was still in the development stage; however, once it was seen that these abandoned or grazing lands had value as car- bon storage areas, disagreements emerged on the land use; this was compounded by changes in village leadership during the period. Cuishan Village has relatively weak social capital based on the data collected through village-level surveys.  Cuishan has 20 dispersed natural villages, and only one natural  26 This reticence by the village could also theoretically be solved by having clearly enforceable contracts or a ‘hostage’ given on the part of the company (Williamson 1985).  However, given the lack of trust, it is uncertain whether either of these contractual or institutional changes would be sufficient in this case.  58  village has access to the road.  During the village-level focus group discussions, the vil- lagers said that members from different natural villages had few interactions in their daily life due to the difficult road access and the sparsely distributed natural villages.  There- fore, due to lack of interactions among members from different natural villages, the norms of mutual reciprocity and trust, which are essential bonding social capital within the Village, are relatively weak.  In addition, most of the recently appointed leaders of the Cuishan Village were relatively young, and therefore did not have the respect afforded a village elder. Further, they were previously involved in off-farm activities, suggesting that they spent most of the time outside their village. Therefore, Cuishan’s leaders could not effectively coordinate among the natural villages to resolve land tenure disputes.  Be- cause of the land tenure disputes and weak social capital in the villages, the natural vil- lages in the tenure disputes could not find mutually-acceptable income-sharing contracts. As a result, some natural villages decided to hold up their participation in the CDM project until the disputes are resolved through coordination by the forest government agencies.  Currently in Cuishan Village, about 30% (126 ha) of the communal lands have unresolved property rights disputes. In contrast, other villages in Huanjiang County, such as Beishan and Shangang Vil- lages in Xunle Miao Ethnic Autonomous Township27 also face tenure disputes but they continued to participate in the CDM program. In these villages, the majority of the village members belong to the homogenous Miao ethnic minority group.  Miao ethnic minority  27 In ethnic autonomous townships in China, the majority population belongs to a certain ethnic minority group.  The Miao ethnic minority group people in China are famous for their high bonding social capi- tal.  They are among few minority ethnic groups in China, who still keep their distinct culture and customs, which commonly include strong norms of mutual reciprocity and practicing social sanctions on norm viola- tors.  59  has long tradition of applying norms of mutual reciprocity and social sanction to punish norm violators.  Therefore, village with Miao ethnic minorities as dominant population often have high bonding social capital and are able to coordinate among themselves to solve conflicts such as the tenure problems.  Indeed, all designed tree planting plans in Beishan and Shangang are complete. This example goes to proposition 2, corollary 2. If one views establishing property rights as a subset of bargaining costs, with sufficient so- cial capital, the bargaining costs might be less important in project assessment.  Further, this implies one could have a successful form of Coasean bargain even without clearly established property rights as long as one has sufficient social capital. 2.6. Conclusions A CDM forest project can be thought of as a form of a Coasean bargain, in that people who benefit from a positive externality write a contract with producers of that externality to pay for its provision. Like a Coasean contract, for a CDM forest project to be efficient and engender participation, the surplus it generates needs to be large enough to overcome the costs of bargaining.  The institutional framework in which the CDM operates affects the nature of these bargaining costs, and needs to be taken into account when designing or analyzing PES schemes. The Guangxi CDM project has some innovative attributes that attempt to lower bar- gaining costs.  For example, the share-holding system is an important innovation to at- tempt to make the project accessible to small land users.  However, restrictive contractual rules, property rights disputes combined with weak social capital have led to high bar-  60  gaining costs in some cases, which, in turn, contributed to about 30% of the unfinished designated tree planting plans (Rows 1 and 3 in Table 2-3). We suggest two propositions and two corollary statements about a CDM project and its relationship with property rights, social capital and contracts. First, having a positive surplus is not sufficient to engender participation.  The Guangxi CDM project appears to offer strong financial incentives to participation, yet almost half of the land remains unfo- rested.  Second, we posit that the barrier of high bargaining costs can be reduced by stan- dardized contracts, formal institutions like clearly defined property rights and strong so- cial capital. Our first corollary states that along with the potential benefits of a standar- dized contract, one drawback is that it may be so restrictive as to limit participation by some. We see this in the case of the high-cost lands, which, not surprisingly are used to help demonstrate the additionality of the project. Because of a homogeneous payment coupled with the lack of an upfront payment, these lands remain unplanted. Other lands have remained unplanted because of increased opportunity costs or local land users’ un- certainties of their potential benefits due to their anticipated weak bargaining power after making specific investment of their lands to the CDM forest project. Ironically, some of these lands, which are no longer “additional”, remain unplanted, while their owners nego- tiate with the local forest companies for a larger share of the CDM income. Our second corollary is that high social capital can overcome the lack of clearly- defined property rights often seen as required for a Coasean bargain to be feasible. Social capital plays an important role in project design and implementation through its interac- tion with contractual rules and land tenure arrangements. When trust between local land  61  users and the local forest companies is lacking, local land users tend not to participate in the project under a share-holding system. Further, when land tenure is in dispute and lo- cal land users do not have strong social capital, the pooling arrangement would not suc- cessfully facilitate the project implementation.  Conversely, those communities with high social capital were able to overcome their tenure disputes and are now participating in the program. Thus, if missing formal institutions, such as well-defined property right could be established before a CDM project, or contractual arrangements could be tailored to the degree of local social capital as well as local costs, participation might well be expanded. As a PES involves interactions among agents (Kosoy and Corbera 2010), it inherently is affected by social capital, property rights and contractual rules.  While the project can be made by improving financial incentives, reducing risk of payment and using the con- tracts to strengthen property rights, trust between intermediaries appears to be a critical for a successful PES (Asquith et al. 2008).  Particularly where formal institutions are weak, building trust among service providers is important in achieving successful PES projects.  62  Bibliography Asquith, N.M., Vargas, M.T. & Wunder, S. (2008). Selling two environmental services: in kind payments for bird habitat and watershed protection in Los Negros, Bolivia. Ecological Economics 65 (4), 675–684. Australian Greenhouse Office (2005). Planning Forest Sink Projects: A Guide to Carbon Pooling and Investment Structures.  Retrieved on April 15, 2010, from http://www.greenbiz.com/business/research/report/2005/06/07/planning-forest-sink- projects-guide-carbon-pooling-and-investment-structu. Bhalla, A.S. (1992). Uneven Development in the Third World: A Study of China and In- dia. New York, St. Martin's Press. Bennett, M. (2008). China's sloping land conversion program: institutional innovation or business as usual? Ecological Economics 65 (4), 699–711. Carter, M.R. (1987). Risk sharing and incentives in the decollectivization of agriculture. Oxford Economic Papers 39, 577–595. Cacho, O.J., Marshall, G.R., Milne, M. (2005). Transaction and abatement costs of car- bon sink projects in developing countries. Environment and Development Economics 10 (5), 597–614. Project Development Document (PDD) for AR-AM0001: facilitating reforestation for Guangxi watershed management in Pearl River basin, China. CDM Executive Board.  63  Dillman, D.A., Smyth, J.D., Christian, L.M. (Eds.), 2008. Internet, Mail, and Mixed- Mode Surveys: The Tailored Design Method, Third Edition. Hoboken, New Jersey: John Wiley & Sons, Inc. Engel, S., Pagiola, S., Wunder, S. (2008). Designing payments for environmental services in theory and practice. Ecological Economics 65 (4), 663–674. Gibson, C.C., Williams, J.T., Ostrom, E. (2004). Local enforcement and better forests. World Development 33 (2), 273–284. Grieg-Gran, M., Porras, I., Wunder, S. (2005). How can market mechanisms for forest environmental services help the poor? Preliminary lessons from Latin America. World Development 33 (9), 1511–1527. Guangxi Forestry Bureau (2006). Social Assessment on Guangxi Integrated Forestry De- velopment and Conservation Project. Haupt, F., von Lüpke, H. (2007). Obstacles and opportunities for afforestation and refore- station projects under the Clean Development Mechanism of the Kyoto Protocol. FAO Advisory Committee on Paper and Wood Products. Keenan, R.J., Davey, S., Grieve, A., Moran, B., Donaldson, J. (2004). Market mechan- isms and assessment methods for environmental services from private forests in Aus- tralia. In Proceedings of the workshop forests for poverty reduction: opportunities with clean development mechanism, environmental services and biodiversity, Seoul, Korea. Food and Agriculture Organization of the United Nations (FAO) (edited), Bangkok, pp. 49-60.  64  Kosoy, N., Corbera, E. (2010). Payments for ecosystem services as commodity fetishism. Ecological Economics 69 (6), 1228–1236. Kumar, K.B., Matsusaka, J. (2008). From families to formal contracts: An approach to development. Journal of Development Economics 90 (1), 106–119. Lyon, F. (2003). Community groups and livelihoods in remote rural Areas of Ghana: how small-scale farmers sustain collective action. Community Development Journal 38, 323–331. Milne, M. (1999). Transaction costs of forest carbon projects.  Working Paper CC05. ACIAR project ASEM 1999/093.  University of New England.  Retrieved on April 15, 2010, from  http://www.une.edu.au/carbon/CC05.PDF. Muradian, R., Pascual, U., Corbera, E., Kosoy, N., May, P.H. (2010). Reconciling theory and practice: An alternative conceptual framework for understanding payments for environmental services. Ecological Economics 69 (6), 1202–1208. Narayan, D. (1999). Bonds and bridges: social capital and poverty. Policy Research Working Paper 2167. Washington, D.C: World Bank. North, D.C. (1990).  Institutions, Institutional Change and Economic Performance. Cambridge: Cambridge University Press. Ostrom, E., Nagendra, H. (2007). Tenure is not sufficient: monitoring is essential. Envi- ronmental Economics and Policy Studies 8 (3), 175–199.  65  Ostrom, E. (1991). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press, Cambridge. Pagiola, S. (2008). Payments for environmental services in Costa Rica. Ecological Eco- nomics 65 (4), 712–724. Pagiola, S., Ramírezb, E., Gobbic, J., de Haana, C., Ibrahimc, M., Murgueitiod, E. (2007). Paying for the environmental services of silvopastoral practices in Nicaragua. Eco- logical Economics 64 (2), 374–385. Pagiola, S., Arcenas, A., Platais, G. (2005). Can payments for environmental services help reduce poverty? an exploration of the issues and the evidence to date from Latin America. World Development 33 (2), 237–253. Pearson, T., Walker, S., Brown, S. (2006). Guidebook for the formulation of afforestation and reforestation projects under Clean Development Mechanism. International Trop- ical Timber Organization (ITTO) Technical Series 25. Pretty, J., Ward, H. (2001). Social capital and the environment. World Development 29 (2), 209–227. Reedy, H. (2003). Feasibility assessment of afforestation for carbon sequestration: ex- ploring options for aggregating and selling afforestation carbon credits from small landowners. Canadian Forest Service (Pacific Forest Center). Richardson, G. (2005). The prudent village: risk pooling institutions in medieval English agriculture. Journal of Economic History 65, 386–413.  66  Smith, J., Scherr, S.J. (2003). Capturing the value of forest carbon for local livelihoods. World Development 31 (12), 2143–2160. Sobel, J. (2002). Can we trust social capital? Journal of Economic Literature 40 (1), 139– 154. Tschakert, P., Coomes, O.T., Potvin, C. (2007). Indigenous livelihoods, slash-and-burn agriculture, and carbon stocks in eastern Panama. Ecological Economics 60 (4), 807– 820. UNEP Risoe Center (2009). CDM Pipeline overview. Retrieved April 19, 2009, from http://www.cdmpipeline.org. van Kooten, G.C., Thomsen, R., Hobby, T.G., Eagle, A.J. (2006). Social dilemmas and public range management in Nevada. Ecological Economics 57 (4), 709–723. van Kooten, G.C., Shaikh, S.L., Suchanek, P. (2002). Mitigating climate change by plant- ing trees: the transaction costs trap. Land Economics 78 (4), 559–572. Vatn, A. (2010). An institutional analysis of payments for environmental services. Eco- logical Economics 69 (6), 1245–1252. Williamson, O.E. (1985). The Economic Institutions of Capitalism: firms, Markets, Rela- tional Contracting. The Free Press, New York. Wunder, S. (2005). Payments for environmental services: some nuts and bolts.  CIFOR Occasional Paper No. 42.  Retrieved on April 15, 2010, from http://www.cifor.cgiar.org/publications/pdf_files/OccPapers/OP-42.pdf.  67    Chapter 3  Individual Social Interactions, Village Institutions and Rural Villagers’ Mutual Trust: Survey and Field Expe- rimental Evidence from Western Rural China28  28 A version of this chapter will be submitted for publication.  Gong, Y., Baylis, K., Xu, J. and Bull, G. Individual social interactions, village institutions and rural villagers’ mutual trust: survey and field experi- mental evidence from western rural China.  68  3.1. Introduction Trust is critical for social capital (Bellemare and Kröger 2007; Fukuyama 1995).  It plays an important role for economic growth (Knack and Keefer 1997; La Porta et al. 1997; Zak and Knack 2001), good governance (Knack 20001), individual economic perfor- mance, such as financial loan repayment rates (Karlan 2005), stock purchases (Guiso et al. 2008) and participation in community resource management (Boudma et al. 2008).  It also has important implications for successful policy interventions in developing coun- tries (Boudma et al. 2008).  Trust may  influence  economic success on macro- and mi- cro- levels mainly through its role of reducing transaction costs (Alesina and La Ferrara 2002), lubricating co-operation (Pretty and Ward 2001; Sobel 2002) and facilitating the enforcement of formal or informal contracts (Asquith et al. 2008; Lyon 2003; Ostrom and Nagendra 2007). Given the importance of trust, it is important to have proper measures of trust and a good understanding of the determinants of the trust.  In the past, both survey and experi- mental methods have been used to measure trust.   However, thus far, the relationship be- tween surveyed and the experimental measures of trust is inconclusive (Bellemare and Kröger 2007; Glaeser et al. 2000).  Regarding the determinants of trust, the theory is largely sketchy (Alesina and La Ferrara 2002).  Nonetheless, recent behavioral research has started to posit the view that beliefs could be important economic primitives for trust (Fehr 2009; Naef and Schupp 2009).   From a behavioral point of view, trust is defined as a behavior (Fehr 2009) and the action of trust29 can be characterized as: [1] an individu-  29 Fehr (2009) and Naef & Schupp (2009) followed concept of Coleman (1990) to propose a behavioral definition of trust.  Our paper also follows their behavioral definition of trust.  69  al’s (the trustor’s) voluntary transfer of assets to the other party (trustee), while the trans- ferred assets will be at the trustee’s disposal without any legal commitment from her30; [2] an action motivated by a potential gain (Fehr 2009; Naef and Schupp 2009).  From the behavioral definition, trust essentially involves 2 major motivations for the trustor: [1] his pecuniary self-interest motivation for a monetary return given that there is a potential gain from trust; [2] his expectation regarding the trustee’s trustworthiness.  Since the transferred assets would have to be at the trustee’s disposal, in order for the trustor to place trust in the trustee, he must have an optimistic expectation or belief that she is trustworthy.    Regarding the determinants of trust,  empirical evidence has shown trust is correlated with individual and community characteristics, past experiences, legal institu- tions, repeated interpersonal interactions, social connections and community leadership (Alesina and La Ferrara 2000; Barr 2003; Boudma et al. 2008; Durlauf and Fafchamps 2004; Glaeser et al. 2000).   These correlates of the trust are closely related to the roles of formal and informal institutions31 in determining trust: while formal institutions streng- then the overall environment of trust through the rule of law, enforceable contracts, in- formal institutions engender trust through social mechanisms, rewarding trustworthy be- haviors and imposing social sanctions (Narayan 1999). In this chapter, we aim to indentify the determinants of rural villagers’ mutual trust under the framework of psychological game originally developed by Geanakoplos et al. (1989).  We specifically intend to test the following hypotheses in the specific setting of  30 From here onward, the trustor is referred as “he”, while the trustee is referred as “she”. 31 Legal institutions, including formal contracts, are considered as formal institutions; social connections, networks and the underlying interpersonal interactions can be viewed as informal institutions (Narayan 1999; North 1990).  70  southwestern China’s Yunnan Province, where a number of closely-knit ethnic communi- ties reside. Hypothesis 1.  The formal institution established on village level could reduced  the ru- ral villagers’ mutual trust in Yunnan Province. Hypothesis 2.  Degree of interpersonal interactions can affect rural villagers’ mutual trust. We conducted trust games and questionnaire surveys in 30 administrative villages in 6 counties in Yunnan Province in Southwestern China.   We used both the field experi- mental method and survey method to measure the same dimension of the trust, rural vil- lagers’ trust in their peer villagers32. 3.2. The Measurement of Trust Survey and experimental methods have been used to measure trust in surging empirical research in the past two decades (Alesina and La Ferrara 2002; Berg et al. 1995; Barr 2003; Bouma et al. 2008; Buchan et al. 2008; Croson and Buchan 1999; Eckel and Wil- son 2004; Glaeser et al. 2000; Karlan 2005; Knack and Keefer 1997; La Porta et al. 1997; Naef and Schupp 2009; Schechter 2007; Zak and Knack 2001).  The survey method has been commonly used to measure trust both on aggregated and individual levels (Glaeser et al. 2000; Knack and Keefer 1997; La Porta et al. 1997; Zak and Knack 2001).   The survey measure of trust is thought to largely capture people’s expectation of others’ trustworthiness (Naef and Schutt 2009).  It has been commonly obtained through World Value Survey (WVS) and US General Social Survey (GSS).  However, the survey me-  32 Our research design is similar to the research of Naef and Schupp (2009), who used survey and experi- mental methods to measure the same dimension of trust, i.e. trust in strangers in developed countries.   71  thod is criticized for its lack of behavioral relevance (Naef and Schupp 2009) and intro- duces the divergence between stated versus actual preferences and beliefs (Bouma et al. 2008).   During the past two decades, the experimental method that is applied both in la- boratory and field settings has been used as an additional method to measure trust (Berg et al. 1995; Croson and Buchan 1999; Barr 2003; Boudma et al. 2008; Glaeser et al. 2000; Karlan 2005; Naef and Schupp 2009; Schechter 2007). So far, the relationship between the survey and experimental measures of trust is still inconclusive.  Glaeser et al. (2000) concluded that the survey measure of trust was not correlated with trusting behaviors observed in trust games, while others found a signifi- cant correlation between the survey and experimental measures of trust (Bellemare and Kroeger 2007; Holm and Nystedt 2008).  Naef and Schupp (2009) pointed out that the inconclusive relationship between the survey and experimental measures could be attri- buted to the fact that they were capturing different dimensions of trust, as trust could refer to generalized trust in the general population involving strangers or personalized trust arising from repeated interpersonal interactions (Fafchamps 2006).   Naef and Schupp (2009) therefore used experimental and survey methods to focus on measuring a single dimension of “trust in strangers”.  They found that their experimental measure of trust in strangers was strongly correlated with their survey measure of trust. 3.2.1 The Experimental Measure of Trust The experimental measure of trust has been commonly obtained by conducting trust games in laboratory or field experiments (Barr 2003; Berg et al. 1995; Bourma et al. 2008; Croson and Buchan 1999; Glaeser et al. 2000; Karlan 2005; Naef and Schutt 2009;  72  Schechter 2007).  Trust games have been used to measure trust in strangers, the genera- lized trust (Fehr 2003; Glaeser et al. 1999; Naef and Schupp 2009) or trust in known people (Karlan 2005; Schetcher 2007). The standard trust game was originated by Berg et al. (1995) (hereafter referred as Berg-Dickhaut-McCabe trust game).    In the Berg-Dickhaut-McCabe trust game, the game, which was a one-shot game, was played by two players, player A (he) and Player B (she) with experimenters’ facilitation and under double-blind conditions, under which the individual players’’ decisions were unobservable to all parties involved.  The player A, who played the role of the sender, was given US$ 10 at the beginning of the game.  He was anonymously matched with player B, who played the role of receiver.  The player A could choose to send some positive amount or nothing to the player B.   If the player A decided to send nothing, the game was over; if he decided to send a positive amount to the player B, she then decided the amount to be sent back to him.  Since Berg et al. (1995), the trust games have been commonly used to measure the player A’s trust and player B’s trustworthiness (Barr 2003; Berg et al. 1995; Boudma et al. 2008; Croson and Buchan 1999; Glaeser et al. 2000; Karlan 2005; Schechter 2007). In the past, the Berg-Dickhaut-McCabe trust game has been modified in different ways.  Glaeser et al. (2000) modified the game by having the paired players known to each other to analyze the impact of the players’ social-connectedness on their behaviors in the game.  They also combined survey questions with the trust game to measure the trust and trustworthiness of the student subjects.    Croson and Buchan (1999) conducted the trust game by giving the same endowment to both players at the beginning of the  73  game.  Burks et al. (2003) modified the game to allow the players to play both roles (i.e. the sender and the receiver).  Schechter (2007) conducted the trust game in rural Para- guay by recruiting subjects from the rural villages to play both roles.  Karlan (2005) ap- plied the trust game in rural Peru to recruit members of a certain village banking organi- zation in the local area as the subjects of the game and allowed the subjects to play games face-to-face and the presence of administrators. Karlan (2005) and Schechter (2007) both considered it was crucial for the experimen- ters to be presented in the game when the experiments were conducted in rural settings, where people having varying levels of education.  They believed that the presence of the experimenters could help ensure the rural villagers to understand the game they were playing and the success of the experiments, although there was a risk that the experimen- ter’s presence may influence the players. 3.2.2 The Experimental Measure and Psychological Games The outcomes of trust games have been commonly explained by conventional game theory, which assumes that the players’ utility depends solely on its actions (payoffs) but not on the intentions behind the action.   Nonetheless, the outcomes have also been ex- plained using the framework of psychological games developed by Geanakoplos et al. (1989), in which payoffs are assumed to depend both on the actions taken by players and their beliefs (intentions). One typical application of using psychological games to explain the outcomes of trust games is the reciprocal-trust relationship hypothesis (TR hypothesis) proposed by McCabe et al. (2003).  The TR hypothesis posit that the first-mover (the player A) and  74  the second-mover (the player B) enter a reciprocal-trust relationship if: [1] there are mu- tual gains from their joint actions; [2] the player A takes a risk by trusting the player B; [3] the player B gives up something in order to reciprocate player A’s trust.  According to the TR hypothesis, entering trust-reciprocal relationship can lead to an improvement in group payoffs.  However, the player A has to give up a sure thing with a certain value in ex- change for an anticipated future benefit.  Since A’s opportunity cost of trusting in B is positive, A’s taking the risk to achieve a cooperative outcome can signal A’s intentions to enter a reciprocal-trust relationship.  Moreover, in order for B to reciprocate A’s good intention, she must give up a positive opportunity cost as well. 3.3. Survey and Experimental Design Both questionnaire surveys and trust games were conducted in 30 administrative villages in 6 Counties in different regions of Yunnan Province in Southwest China.  Five adminis- trative villages were randomly selected from each County.  In each administrative village, 2 village clusters, which are sub-units of the administrative village and closely-knit community, were randomly chosen.  In each village cluster, 10 households were random- ly chosen.  A family head from each randomly selected household was asked to partici- pate in surveys and trust games.  In total, 600 respondents were surveyed and participated in trust games. 3.3.1 The Survey Design: Measuring Stated Trust and Beliefs Stated trust and beliefs were measured using survey questions.  Regarding the survey measures of trust, a 5-poin scale was used to elicit the respondents’ trust in their peer vil-  75  lagers.   They were asked to rank the following 3 statements (with 1 indicating “strongly disagree” and 5 “strongly agree). [1]  Your village members are trustworthy. [2]  When you are away, you will ask your neighbor to look after your house. [3]  You will only lend money to good friends or relatives in your village. The above statements were specifically related to the respondents’ opinions of their trust in their peer villagers.   A factor analysis was used to create an indicator of “stated trust”, which was used in the regression analysis to check the correlation between the “stated trust” and trusting behaviors observed in the trust games. In order to measure player 1’s beliefs about the player 2’s trustworthy behaviors, we also elicited his expectation of her behavior by asking him the following question: [4]  How much do you expect your partner to transfer back to you33?  The question was asked to the player 1s who had decided to transfer a positive amount to the player 2. Individual household surveys and village-level surveys were also conducted to get the respondents’/players’ socio-economic and demographic characteristics, and their land use decisions.  The household surveys were mainly conducted before the trust games in order for the respondents to gradually get comfortable with the enumerators, who were also the  33 We also asked the player the following question: How much do you expect to receive from your partner?  76  experimenters of the game.  The village-level surveys were conducted to obtain informa- tion of village characteristics from the (administrative) village leaders. 3.3.2 The Experimental Design: Revealing Trusting Behaviors Two-person trust games were conducted anonymously between 2 randomly selected vil- lagers from a certain village.  The subjects played in the trust games were also the res- pondents of the surveys.  For the trust game, the 2 village clusters selected from the same administrative village were randomly divided into 2 groups: 10 subjects from one natural village playing the role of senders (hereafter referred as the player 1 or he) and 10 sub- jects from another natural village playing the role of receivers (hereafter referred as the player 2 or she).  As a result, 300 subjects from 30 village clusters played the role of senders and 300 subjects from 30 village clusters played the role of receivers. After the questionnaire surveys, the respondents were also asked to play trust games. Each survey enumerator, who was also the experimenter of the game, explained rules of the games to the players and made sure that the players understood the association be- tween their decisions made in the game and the final payoffs they as well as their partner would get from the game. Our trust game mainly followed the game design of Schechter34 (2007) but had the subjects play a single role (either the player 1 or the player 2).  We designed the game to attempt to ensure anonymity among the subjects, who live in closely-knit rural communi- ties and have real interactions in their daily life.  Because the subjects were from the  34 In Schechter (2007), subjects were asked to play both roles as the sender and the receiver.  77  same village and would inevitably interact with each other after the games, if they knew their partners’ identities, their actions would likely be influenced by possible post-game consequences, such as blame from the other party if they decided not to send any amount, or feelings, such as kind feeling for a friend or a relative.  Therefore, ensuring anonymity would help limit these potential impacts.   To keep as much anonymity as possible, we conducted experiments simultaneously with 10 subjects (the player 1s) in one village cluster in the morning and another 10 subjects (the player 2s) in another village cluster in the afternoon.  In rural Yunnan, the village clusters are often far away from each other, finishing experiments within 1 day in 2 different  village clusters  also helped keep the anonymity among the subjects. The trust games were conducted with the following steps: Step 1.  Both players were given RMB 20 yuan at the beginning of the game as their participation fee. Step 2.  After the rules of the game were informed, the player 1 made his decision among 5 choices: to send RMB 0 yuan, RMB 5 yuan,  RMB 10 yuan, RMB 15 yuan or RMB 20 yuan. Step 3.  If the player 1 chose to send 0, then the game was over; if he chose to send some positive amount (X), the experimenter doubled that amount (2X) and would send it to the player 2. In this step, after the player 1 had made his decision, the experimenter asked him the following two open-ended questions:  78  [1]  For all player 1s, they were asked: why did you decide to (not to) send the money? [2]  For those who had sent a positive amount, they were asked: what is the most poss- ible amount of money that you expect to receive from the receiver? Step 4.  After the rules of the game were explained and before the experimenters in- formed the amount received, the player 2 was asked to guess the most possible amount of money she would receive from the player 1.  This question was asked even for those who received nothing.  After she made the guess, the experimenter informed the actual amount that she received. Step 5.  The player 2 decided to send back nothing, a certain amount or all the money out of 2X, which was twice of the amount sent by the player 1, to him35.    If the player 2 decided to send back nothing, the game was over; if she decided to send some amount (Y) back, the experimenter sent back the money to the player 1. Our trust games have the following noteworthy set-ups to fit our purpose of mea- surement and rural Chinese settings: [1]  Our experimental measure of trust also specifically measure people’s trust in their peer members from the same village.  Therefore, in the trust games in our research, the following common information was provided to each player:  35 It was known to both players that the player 2 did not need to send back RMB 20 yuan that she received from the experimenter at the beginning of the game.  79  (i) All paired players in the game were randomly selected from their own (ad- ministrative) villages, which means that anybody from their (administrative) village could be picked up and paired together. (ii) The player 1 could not know who was paired with him, neither could the re- ceiver, but it was certain that the paired ones were from the SAME (adminis- trative) village. [2]  In our game, both players knew that both of them were given RMB 20 yuan at the beginning of the game.   Therefore, they knew that even if the player 1 decided to send nothing to the player 2, both of them would still have RMB 20 yuan at the end of the game. Given this set-up, we hoped to control for the effect of the player 1’s “fairness” con- cern on his decision-making.  Suppose only RMB 20 yuan had been given to the player 1, then he would possibly think that player 2 would have nothing if he did not send any amount to her.  Motivated by this “fairness” concern, the player 1 might have chosen to send “some” amount in order to “be fair” to his co-villager paired with him. [3]  We also tried to control for the possible effect of player 1’s concern for the “riski- ness” of the “investment mechanism” in their decision-makings and to keep the ex- periments in a relatively “sterile” condition that was separated from the daily prac- tices and moral issues. Given the set-up of the trust games, which implicitly involve “investment” element, player 1s may concern with the “riskiness of the investment mechanism” in making their  80  decisions.  In address such concerns, we used an elicitation mechanism that provides very clear information that “no riskiness is involved in the investment mechanism”.  In addi- tion, the experimenters used the words of “transfer to” and “transferred back” instead of “lending” or “borrowing” in order to keep the experiments in a relatively “sterile” condi- tion to control for the effect of daily practices and moral issues on the players’ decision- making.  The following is the elicitation mechanism that we used in our trust game: The investment project will have 100% certainty of generating profit, which is twice of the investment.  Nonetheless, only the player 2 is eligible for investment in the project. In order to be eligible, first the player 2 must deposit RMB 20 yuan, which she received from the experimenter, in the project.  This amount would not generate any profit.  In or- der for the player 2 to earn the profit, she must be able to have some money transferred from the player 1, who has a maximum amount of RMB 20 yuan.  The more amount transferred from the player 1, the more profit would be earned from the investment in the project.  For example, if the player 1 transfers RMB 5yuan to her, she could invest it to the project and would earn a profit of RMB 10 yuan.  If the player 1 decided to transfer RMB 20 yuan to her, she could earn a profit of RMB 40 yuan from her investment.    The earned profit would be divided between two of them with the player 2 having all control of the decision on how to divide the profit.  In case the player 1 decided to transfers noth- ing to her, she would have no money to invest and no profit to earn.  Nonetheless, both of them could still get RMB 20 yuan from the experimenter. [4]  We carefully conducted pretest experiments in order to identify a reasonable stake size for the trust games since Johansson-Stenman et al. (2005) found that the in-  81  crease in stake size could significantly decrease the amount sent in the trust games. In the pretest experiments, we tried to use RMB 5 yuan or RMB 10 yuan as the stake. We noticed, however, that some of the player 1s made very quick decisions by trans- ferring RMB 5 yuan.  Even though the experimenter reminded them that the money was at their disposal, they stated that they did not really feel hurt if they transferred RMB 5 yuan while having no money transferred back to them, because RMB 5 yuan was relatively a trivial amount of money.  After several rounds of the pretests, we decided to use RMB20 yuan.  Based on our observations in the pretests, we decided to use RMB 20 yuan, which is equivalent to the daily wage in the local area, as the stake for the trust games. Some examples were practiced with the experimenters’ facilitation in order for the players to have a good understanding of their own and their partners’ payoffs associated with the decisions they made in the game.  Considering the overall educational level of the subjects in rural Yunnan and complexity of calculation of the players’ own payoffs and their partners’ payoffs, the experimenters used some visual methods to work together with the players through some examples.  In the examples, the experimenters used the tokens marked with different amount of money, including RMB 1 yuan, RMB 5 yuan and RMB 10 yuan.  The players knew that the tokens represented the true amount of money and calculated their own and partners’ payoffs given different strategies they would take. They therefore understood that they need to make a decision that they like most and then the corresponding amount they would earn from the game.   Once the players had a good understanding of the games they were playing through the practices of the examples, the games were formally played.  All players clearly knew that only the decisions they made  82  in the formally played game was counted while the decisions they made in the practices were not counted.   Tokens were used in the examples and the final games to facilitate the players to make calculations of their own and the partners’ payoffs. 3.4. Field Evidence for Player 1’s Motivations and a Theoretical Model 3.4.1. Outcomes of the Game The outcomes of our trust games are inconsistent with the sub-game perfect Nash equili- brium that is predicted based on the individual self-interest payoff maximization.  Overall, 230 (77%) of 300 player 1s sent a positive amount to the player 2s.  Table 3-1 presents some summary statistics for the outcomes.  Table 3- 1: Summary Statistics of Outcomes Variables Mean Std. Dev. Range Obs. Average amount sent by all player 1s 8.38 6.39 0-20 300 Average amount sent by player 1s who sent positive amount 10.93 5.02 5-20 230 Average amount returned by the player 2s 13.03 7.62 0-35 230   The Table 3-1 shows that the average amount sent by the player 1s who sent a posi- tive amount is RMB 10.93 yuan and the player 2s who received a positive amount reci- procated by sending an average amount RMB 13.03 yuan. 3.4.2. Field Evidence of Player 1’s Motivations As discussed earlier, after player 1s had decided the amount to be sent, we asked them the following question: “For what reason do you decide to (or not to) send the money to the other person paired with you?”  For 230 player 1s who sent a positive amount, we were  83  able to obtain answers from 219 of them; for 66 player 1s who sent nothing, we were able to obtain answers from 65 of them. We found that motivations underlying the player 1s’ behaviors of sending a positive amount were quite mixed.  Their reasons can be categorized as follows: [1]   Unconditional altruism: “I simply want to help the other party without consider- ing the amount to be returned”. [2]   Help a peer village: “Peer villager should help each other.  If I send her some amount, she can gain a profit”. [3]  Trust: “I trust a peer member in our village will at least return the amount that I sent to her although I do not know who she is”. [4]  Joint profit: “Sending a positive amount can potentially increase my profit and her profit, so it is a win-win action”. [5]   Investment for return: “Although I know there could be a risk that the other party would not return me the money, I still want to send some amount because sending some amount can help me earn more money than RMB 20 yuan”. The following Table 3-2 is the summary statistics for the reasons explicitly expressed by the player 1s who sent a positive amount to the player 2s.   84   Table 3- 2: Player 1s’ Reasons for Sending a Positive Amount Reasons Frequency Percent Average amount (yuan) Std. Dev. (yuan) 1) Unconditional altruism 16 7 9.69 5.31 2) Help a peer villager 63 29 10.48 4.81 3) Trust 64 29 11.17 4.78 4) Joint profit 28 13 12.14 6.15 5) Investment for return 48 22 11.04 5.04 Overall 219 100 10.96 5.07   The reasons listed in the Table 3-2 represent a spectrum of player 1’s motivations, ranging from pure altruism to pure self-interests. [1]   A small fraction (7%) reported that they sent money for purely altruistic reason (Rea- son 1). [2]  About 29% said that they wanted to help their peer members to get some return al- though he knows that the amount sent is at the disposal of the other party (Reason 2). [3]  About 29% specifically stated that they trusted that the player 2 would at least return them the amount they sent (Reason 3). [4]  About 13% said that by sending a positive amount, they hoped to have a win-win out- come as compared to that each of them earned RMB 20 yuan (Reason 4). [5]  About 22% explicitly stated that they foresaw a positive relationship between the amount they sent and a potential positive return to them.  Therefore, they wanted to “invest” a positive amount of money for certain level of return (Reason 5).  85  Reasons (2)-(4) imply the player 1s’ incentives for investment for a potential return. Reason (3) shows that the player 1s felt optimistic that their partners would return certain amount to them. The joint-profit motivation (Reason 4) implies that the player 1 essen- tially “trusted” that their partners would have a “fair” divide of the profit between two of them.   The “investment for return” implies that the player 1s who sent a certain amount were optimistic that they would get a potential return from their investment.    Therefore, reasons 3, 4 and 5 together (64% in total) indicate the player 1s “trusted” that the player 2s would return a certain amount to them. For Reason (5), although the player 1s did not explicitly express whether or not they were expecting to receive a certain amount from their partners, their expectations could be quite mixed: while some player 1s might expect that their partners would return them the amount at least equivalent to the amount they would send, some others might not. Implicitly, their reason that “peer members should help each other” indicated that they “trusted” their partners would return an amount that would not make them worse-off. The field evidence for player 1s’ reasons of sending nothing is presented below in ta- ble 3-3.  Table 3- 3: Player 1s’ Reasons for Sending Nothing Reasons Frequency Percent 1) Do not trust in the (anonymous) partner at all 53 82 2) Just want to keep RMB 20 yuan at my own disposal 6 9 3) Sending nothing is fair 6 9 Overall 65 100    86  For 65 players who offered reasons for sending nothing, the vast majority (82%) ex- plicitly expressed that they chose to send nothing because they did not trust that their (anonymous) partners would return any amount.  About 9% of them said that they felt more comfortable to keep RMB 20 yuan rather than sending some amount either because they were poor or because they did not want to help the peer villagers, who previously had not helped them when they had been in need. About 9% of them said that it was fair to send nothing because both of them would have RMB 20 yuan. 3.4.3. The Theoretical Model In this section, we present a simple theoretical model to support the empirical model spe- cifications.  As shown by the information in the Table 3-2, except for the motivation of “unconditional altruism, the positive amount sent by player 1s appears to be motivated by player 1s’ incentives for pecuniary payoffs.  Therefore, we construct a simple theoretical model to mainly capture the player 1s’ motivations for pecuniary payoffs rather than their social preferences36.   The model construction is intended to reflect the roles of expecta- tions (beliefs) in determining the player 1s’ “trusting” behaviors in trust games. We made the following assumptions based on our game set-up and field evidence: Assumption 1.  Given that both players have RMB 20 yuan at the beginning of the game, the player 1(the sender) is assumed to experience no “painful guilt” even if he does  36 I can also construct a more complicated model to include player 1’s altruism, which can be described by social preference model, like the model of Charness and Rabin (2002), in which player 1 is assumed to care about both his own pecuniary payoffs from the game and the group payoffs.  In the social preference model, a weight λ can be assigned to player 1’s own pecuniary payoffs and a weight λ−1 can be assigned to the group payoffs.  However, the essence of the roles of expectations (beliefs) is the same for the social prefe- rence model and the model I use.  87  not send any amount to his partner.  Moreover, since both players start at a point of equality, any amount they sent was motivated by something other than equality concerns. Assumption 2.  The player 1 is risk-averse. The following is the model set-up: [1]  There are two players, 1 and 2, with player 2’s type unknown to the player 1.  The player 1 can only make subjective assessment on the player 2’s “trustworthiness” type.  Based on his assessment on the player 2’s type, he makes an estimation on the probability of the Y amount sent back by the player 2, given that he sends X amount to the player 2. [2]  Let player 1’s payoff denoted as Z=20-X+Y and his utility function as 0)(' >Zu and 0)('' <Zu . [3]  Player 1 has an initial wealth of RMB 20 yuan, which can be divided between the two assets, a safe asset (20-X) and a risky asset (X)37.  The risky asset (X) has a total return of Y, while the safe asset (20-X) is held in cash with 1 dollar return per dollar “invested”. [4]  Let the player 1F and the player 1G be two different player 1s.  The player 1F esti- mates that the Y amount to be sent back by his partner (the player 2) to him fol- lows a cumulative probability distribution of F(Y), while the player 1G estimates that the Y amount to be sent back by his partner follows a distribution of G(Y).  37 The concept follows the examples in Mas-Collel et al. (1995, p. 188) and Fishburn and Porter (1976).  88  [5]  In making their decisions, the player 1F and the player 1G estimate that the Y amount to be sent back by their partners to them satisfy the following conditions of ∫ ≥XYdF )(Y  and ∫ ≥XYd )(GY , meaning that their expected returns from partners exceed the X amount sent by them to their partners. The player 1’s utility maximization problem is: )()20(max)]20([max 2000 YdFYXuYXUE X XX ∫≥≥ +−=+−    (3.1) The first-order condition for the optimal choice of *X  is: 0)()20(20 *' =+−∫ YdFYXuX         (3.2) Suppose there is another player 1G and G(Y) has a first-order stochastic dominance (FSD) over F(Y), denoted as )()( 1 YFYG > , then we have the following condition: 0)()20(20 *' >+−∫ YdGYXuX       (3.3) Denoting )*(Xφ as the left-hand side of the equation 3.3, the equation 3.3 means )*(Xφ  is greater than 0 when evaluated at *X .  With the strict concavity of the player 1’s utility function, the optimal amount of X sent by the player 1G is greater than the amount sent by the player 1F. The above models essentially show that under the FSD condition, the player 1G be- lieves that he would have a higher probability of getting a higher amount of Y within a whole range of [0, 2X], which results in a higher expected payoff, than the player 1F.  89  Therefore, given the positive relationship between the X and Y, as the expected utility maximizer, the player 1G would decide to send a higher amount of X than player 1F in the trust game. As the risk-averters, the player 1s also care about the “riskiness” of their “risky asset (X)”.  We therefore further suppose: the distributions F(Y) and G(Y) have the same mean, i.e. )|()|( G XYEXYEF = , but the distribution G(Y) has a second-order stochastic do- minance (SSD) over the distribution F(Y), denoted as )()( 2 YFYG > .  The SSD of G(Y) over F(Y) essentially means: although the player 1F and the player 1G estimate the same amount of mean Y to be sent back by their partners, the player 1F foresees a higher risk in the X amount that he “invests” in the trust game than the player 1G does.  In another word, player 1F  expects to receive the Y amount from his partner as large as 2X or as small as zero, while the player 1G foresees a lower variability in the Y amount to be returned by his partner.   As a risk-averter, the player 1F, who perceives a higher level of riskiness in the X amount sent than the player 1G, tends to send less X to his partner than the player 1G does. 3.5. Empirical Model Specifications In the standard analysis of trust games, the amount sent by the player 1 to the player 2 is commonly used as the behavioral measure of trust (Barr 2003; Bouma et al. 2008; Croson and Buchan 1999; Glaeser et al. 2000; Karlan 2005; Naef and Schupp 2009; Schechter 2007), which is thought to be able to capture the essence of trust better than the survey measure of trust (Fehr 2009).  In the empirical model, we use the behavioral measure of trust as the dependent variable.   In this section, we intend to identify some possible de-  90  terminants of the amount sent by the player 1 to the player 2, also denoted as trusting be- haviors, using our theoretical model as guidance. 3.5.1 Individual and Household Characteristics Given that trusting behaviors are essentially the outcomes of player 1s’ inferences about their own and their partners’ expected payoffs, player 1s’ individual and household cha- racteristics38, including education, age, family size and family income could be important determinants of the trusting behaviors. Education should have a positive effect on the player 1’s trusting behaviors possibly because: [1] the better educated player 1s could have better understanding of the positive relationship between the amount they sent and the potential profit they could gain from the game.   As a result, they could make better prediction based on calculations of their own and partners’ expected payoffs.  In other words, the better educated player 1 could very probably understand the experiment better; [2] the better educated player 1s could have more experience interacting with people and dealing with relational risks similar to those involved in the trust games.  As a result, they could be more optimistic about hav- ing a win-win situation and perceive a lower level of riskiness of the X amount of in- vestment.  Therefore, they might send more money to the player 2s for joint profit. Age could be an important determinant of trusting behaviors.  Similar to education, older player 1s could have more experience interacting with people and dealing with rela-  38 Gender is supposed to be a potential determinant of the trusting behaviors as shown by some experimen- tal studies.  Nonetheless, in Yunnan Province, the male is the most common family head making decisions. Therefore, in our sample, only 4% (12) of the respondents/subjects are female and there are few gender variations.  We then do not include the gender variable in our empirical model.  91  tional risks in the past and thus the older player 1s could be more optimistic about having a situation of win-win outcomes by sending higher amount for joint profit.  Nonetheless, when people get old enough, they tend to be more prudent.  As a result, the relatively older player 1s could be more and conservative for his own guess of their peer members and thus would assign a lower weight to their peer members’ payoffs by sending a lower amount.   Moreover, the relatively older player 1s could be more risk-averse and thus would choose to keep the money for certainty rather than investing them to gain a poten- tial profit with certain level of risks. Family income should have a positive effect on the amount sent by the player 1 through 2 channels.  First, it affects the amount sent through its effect on the player 1’s risk preference.  A player 1 from a wealthier family could be less risk averse and thus may invest a higher amount of money to get a potentially higher profit.  Second, a player 1 from a wealthier family could be more altruistic and willing to help peer members, as- suming helping others is a normal good. Family size could have some effect on the trusting behavior, but the sign of the effect could be undetermined mainly due to: [1] a bigger family could make a player 1 more risk averse and thus would keep the money with certainty rather than invest in “risky” situation for a potential profit; [2] a bigger family might have interacted more with their fellow villagers or received more help from peer villagers, since big families usually need more support in rural Yunnan, where the main family income largely relies on agricultur- al sector.  As a result, a player 1 from a bigger family would be more willing to enter a mutual help situation for joint profit and could send more.  92  3.5.2 Identifying Potential Determinants of Beliefs As shown in the theoretical model, the player 1’s beliefs about the player 2’s trustworthi- ness should be a critical determinant for the trusting behaviors observed in the trust games.  In this section, we identify variables that could potentially affect trusting beha- viors through the belief component. We posited that the player 1’s interactions and past experiences with his peer villag- ers could be potentially important determinants of his beliefs about the anonymous part- ner’s trustworthiness in the game because they could affect: [1] the player 1’s mean guess about the anonymous partner’s type; and [2] his confidence of the precision of his mean guess.  In other words, the player 1 would tend to retrospect his past experiences and so- cial interactions with his peer villagers as well as the behaviors he had observed regard- ing his peer members could provide necessary information for him to make his subjective assessment of his anonymous partner and accordingly his best strategies in the trust game. Compared with a player 1 who had less social interactions with his peer members, a play- er 1 who had more social interactions with his peer members could have more informa- tion to make subjective assessment of his anonymous partner’s type.  In addition, he could feel more confident in his mean guess of the partner’s type, which could lead him to assign a higher weight to his partner’s payoff.   Similarly, if a player 1 used to have good experiences with his peer members, such as he was treated nicely by his peer mem- bers before, he would tend to believe that his anonymous partner in the game would also treat him nicely and could be trustworthy.  As a result, he would be willing to send some amount to his anonymous partner in the game.  In contrast, if he had been treated unfairly or badly in his interactions with other peer villagers, he would believe less that his partner  93  would treat him nicely in the game and as a result he might decide not to send a positive amount. Player 1’s social interactions and past experiences with his peer members could be on individual level or group level as discussed by Fehr (2009).  In our empirical analysis, we also tried to identify some proxies reflecting the player 1’s past social interactions and past experiences on both levels. Individual- level Proxies for Past Experiences and Social Interactions We first identify two individual-level proxies for the player 1’s past experiences and so- cial interactions with his peer members: [1] the proportion of gift exchange expenditure in the total family expenditure in 2006; [2] a dummy variable indicating the past help that the player 1’s family got from village members in dealing with wedding or funeral mat- ters. Gift exchange and mutual help in dealing with funeral or wedding matters are impor- tant social interactions among rural villagers in China.  These interactions often occur among neighbors, friends in the same village and relatives.  A common practice for gift exchange is that a person who receives the gift in the first place from another person would return the gift with similar or higher values to that person.   Therefore, a player 1 with more experience with gift exchanges could also have more experience of reciprocity and possibly stronger belief that his partner would return a portion of his investment. Second, wedding and funerals are two important events that could bring a family mone- tary, moral or logistic support from their villagers, friends or relatives.  Compared with a player 1 from a family who had to deal with wedding or funeral matters on their own, a  94  player 1 from a family who got help from villagers that are not their relatives might have had closer social interactions with peer members and a better experience of mutual help and reciprocity.  As a result, he would probably have stronger beliefs about the reciprocal actions to be taken by his anonymous partner in the trust game. We also considered that gift exchange could be endogenous to the amount sent.  For example, on the one hand, the higher proportion of gift exchange expenditure in the fami- ly total expenditure could imply that the player 1’s families could have more relatives or friends within the village.  On the other hand, the player 1, who has more relatives or friends in the village, would also have a higher prior probability that the money sent by him would go to one of his relatives or friends.  As a result, he would tend to send a high- er amount to the anonymous partner with the incentive to benefit the other.  Therefore, the unobserved factor related to the player 1’s social relationship in the village could be correlated with the gift exchange, which potentially could in turn lead to the endogeneity problem. Considering the potential endogeneity problems as discussed above, we then identi- fied two alternative individual-level proxies of the player 1’s social interactions and past experiences to check the robustness of the effect of individual-level of social interactions and past experiences on the player 1’s trusting behaviors. One is the ratio of the average distance of the forest plots to the nearest road on household level to that on village level; the other is the ratio of the average distance of the forest plots to home on household lev- el to that on village level.  95  Regarding the variable related to the distance from the forest plots to the road, it could be possible that the farther the distance of the forest plots to the road, the more possible that the player had observed his peer villagers’ untrustworthy behaviors, such as illegal logging undertaken on his forest plots.  With poor economic conditions, timber from forestland is often valuable assets for rural villagers in Yunnan.  On the one hand, whenever possible or necessary, villagers could have incentives to steal the timbers from others’ forest plots to satisfy some economic needs when other means are not easily available.  On the other hand, by living the same village, the rural villagers often do mu- tual monitoring on stealing activities with consciousness or under consciousness in their daily activities.  Therefore, the nearer the distance of the forest plot is to the road, the more easily that a stealing activity could be detected by villagers and the less stealing cases that the player 1 had observed from his family’s forest plots.   As a result, he could also have stronger beliefs about the trustworthy actions to be taken by his anonymous partner in the game. Further, it is possible that villagers help each other to take care of their forest plots. For example, several neighbors who have forest plots in similar locations could take turns to look after their forest plots, including patrolling and monitoring some possible stealthy activities.  Therefore, the farther the forest plots to home, the more possible that the play- er 1 could have good experiences of mutual help and reciprocity.  Consequently, he could have stronger beliefs about his anonymous partner’s reciprocating or trustworthy actions to be taken in the game.   96  Formal and Informal Institutions and the Potential Crowding-out Effect Formal or informal institutions existing in player 1’s villages were used as a proxy for social interactions and past experiences at group-level, in that institutions can provide a structure of daily life and impose constraints that govern human interactions (North 1990). We specifically used village branches of Chinese Communist Party (CCP) and lineage organizations in villages as key village-level formal and informal institutions, respective- ly.  We hypothesized that with political and institutional changes taken place in rural China in past several decades, the formal CCP institution could have gradually crowded out the informal lineage institution to facilitate villagers’ positive social interactions and mutual trust. Lineages are the major reference groups in village life in China and the existence of lineage halls indicates the strength of informal institutions in Chinese rural villages (Tsai 2002; Xiao 2007).  Like in other rural areas of China, in many rural villages in Yunnan Province, it is quite often that one or several lineages often form a clan and all members of a clan might reside together to form a single community (natural village).  In pre- Communist times in the first half of the last century, in lineage-based villages, lineage members, who commonly have kinship connections based on partilineal descent, used to hold corporate holdings, such as lands.  In each lineage, the lineage elders used to hold formal powers of allocating manpower and property, coordinating among the lineage members, negotiating with officials and outsiders and organizing the defense of the community.  In addition, the lineage members maintained solidarity by feasting together or through worship at the lineage halls.  However, with dramatic socio-economic and ideological changes taken place in post-Communist times (especially during Cultural  97  Revolution from 1966 to 1976), the lineage organizations have been dramatically dis- solved and the solidarity among the lineage members has been reduced in a significant number of rural villages in China (Xiao 2007).  Nonetheless, lineage ties still have certain social significance in ritual contexts, such as weddings, funerals, the lunar New Year, and with respect to various kinds of cooperative relationships among lineage members. In contrast to the weakened influence of lineage institutions in rural villages in post- Communist times, formal grassroots institutions have been strengthened with the estab- lishment of village branches of CCP and village committees.  In rural China’s villages, village branches of Chinese Communist Party (CCP) are the key grassroots local political institutions.  They have overall authority for policy implementation, the guidance of po- litical affairs and promotion of communist ideology in rural villages (Jacka 2009).  While the contemporary formal institutions have gradually replaced informal institutions in rural villages, the formal institutions did not effectively guide the rural villages toward a con- temporary and modern society (Potter and Potter 1990).  In a number of villages, the vil- lage CCP leaders have “alienated” the village members in the process of implementing government policies and CCP affairs (Xiao 2007).  Therefore, it is highly possible that in villages, where the traditional lineage organizations have been dissolved due to the estab- lishment of CCP system, the traditional practices of social sanctions that were largely un- dertaken through traditional informal institutions could have been discouraged  With the ineffectiveness of social sanctions in maintaining cooperation and ineffective policy im- plementations of government policies on grassroots level, it could be possible that coop- eration levels have been reduced in certain villages with village members having percep- tions of largely playing non-cooperative finite games.  98  With institutional changes in rural villages, on the one hand, the informal lineage in- stitutions have been gradually dissolved and its traditional roles of maintaining co- operation and solidarity among villagers have become dysfunctional; on the other hand, the CCP institution has not sufficiently substituted for the traditional roles played by the lineage institution.  For example, when the traditional lineage institutions were still active, many traditional ceremonies, such as worship for the deceased or ancestors, were held in the lineage halls, to unify the lineage members and provide opportunities for the members to interact with each other.  Although the CCP institution has been in place, it has not taken on traditional roles of lineage institutions.  As a result, positive social interactions, cooperation and solidarity among the village members could have been reduced.  With reduced social interactions, cooperation and solidity among the village members, a player 1 may have been inclined to have a lower mean guess for the trustworthiness of his ano- nymous peer villager and felt less confident in the amount to be received from his partner. Village Institutions and Village Characteristics Although the lineage institution used to be criticized for its feudalisms and many lineage halls were destroyed during the period of Cultural Revolution (1966-1976), the lineage institution has been rejuvenated in the past decades (Xiao 2008).  As a consequence, it is possible that the CCP institution could be co-existed with traditional lineage institution in certain villages [1] where the lineage institution has not been destroyed and the CCP in- stitution has few impacts; or [2] where the CCP institution has been well-established and the lineage halls have destroyed before but rejuvenated in recent years.    While the vil- lages falling in the first category could be usually those remote villages, the villages in  99  the second category could be relatively open to the outside world and the market.   We checked some correlations between the above two major indicators of village institutions and some key village characteristics, trying to shed some light on institutional and socio- economic conditions in surveyed villages.  Table 3-4 presents the relevant results.    Table 3-4 shows that the CCP institution is strongly and negatively correlated with the village lineage institution at a 1% significance level, meaning that the lineage institu- tion could be significantly weaker in villages where the ratio of CCP members is higher. This implies that CCP institution could have gradually substituted the traditional lineage institutions. Table 3- 4:Correlations between Village Institutions and Village Characteristics  Ratio of CCP members to total village population If the village has lineage halls (1=Yes; 0=No) Village per capita income Distance to the capital of the county Distance to the nearest credit center Ratio of CCP mem- bers to total vil- lage population 1 If the village has li- neage halls (1=Yes; 0=No) -0.260 1 (0.000) Village per capita income 0.437 0.142 1 (0.000) (0.015) Distance to the capi- tal of the county -0.365 -0.156 -0.549 1 (0.000) (0.007) (0.000) Distance to the near- est credit center -0.140 -0.333 -0.440 0.292 1 (0.016) (0.000) (0.000) (0.000) Percent of families having phones in total no of families 0.247 0.199 0.487 -0.470 -0.085 (0.000) (0.001) (0.000) (0.000) (0.145) Note: Significance of correlations is in parenthesis.  100  Table 3-4 also shows that in villages where ratio of CCP members to the total village population is higher, [1] the village income per capita is also significantly higher and the village is more accessible to the outside world and the market, as shown by the negative correlations between the CCP ratio and the per capita income and the positive correlation between the CCP ratio and the distance from the village to the county capital city or the nearest credit centers; [2] the proportion of families having access to telephones (includ- ing landline and cell phones) in the total number of families in the village is significantly higher.  This information indicates that in terms of number of its members, the CCP sys- tem seems to be better established in villages that are wealthier and have better access to the outside world (the country capital city), the local market (credit centers) and informa- tion (telephones). The significant correlation between the existence of lineage halls and other village characteristics shown in the Table 3-4 also indicates that the lineage institution could have been probably destroyed in the past (especially in Cultural Revolution), but might have been rejuvenated in recent years in those surveyed villages that are wealthier, and have better access to the outside world (the country capital city), the local market (credit centers) and information (telephones). Population Density in Natural Villages The population density in natural villages was used as another proxy for the group-level social interactions because it could reflect the intensity of potential social connections among members in natural villages.  In rural Chinese settings, especially in mountainous regions like Yunnan Province, a large portion of social interactions, such as borrowing  101  farming tools from neighbors, competing for water irrigating their croplands in farming seasons and helping each other in difficulties, occur within natural villages.  A higher population density in a natural village could indicate more potential interactions among its residents and could imply more mutual help or more competitions.  Therefore, the ef- fect of population density in a certain natural village on the player 1’s beliefs about ano- nymous partner’s trustworthiness could be negative or positive effect. 3.5.3 Village Socio-economic Characteristics Besides the player 1’s individual characteristics, village socio-economic characteristics could also affect the player 1’s trusting behaviors through their effect on the villagers’ wealth level, openness to the market and social connections among the villagers.  There- fore, we controlled for the heterogeneity of village characteristics in our empirical models. The proportion of households having phones (mobile or landlines or both) in the total number of villages, village per capita income and ethnic heterogeneity are key variables as proxies for village characteristics included in the regressions. The proportion of households having phones in the total number of households in a certain administrative village can largely reflect the openness of a certain village to the outside world and interaction among villagers.  In a large portion of rural villages in Yunnan, which is a mountainous highland in western China, accessibility to landlines re- flect that a village is relatively near to main roads and outside world.  When a village is more accessible to the outside world, they could have better outside options and have more interactions with outside world.  As a result, their interactions with their peer vil- lagers could be reduced.  Given that villagers living natural villages, which could be quite  102  far away from each other, communications among villagers from different natural villag- es could be rare.  When necessary, the villagers have to walk over to another village to communicate without telephones.  When the villagers have access to the phones, the communication could be made over phones and personal interactions could be relatively reduced.  Because the access of the phones could potentially lead to less face-to-face inte- ractions between a family and other peer villagers, the proportion of the families having telephones in a certain village implies that less face-to-face interactions among villagers could also have “alienated” rural villagers. The (administrative) village per capita income could potentially affect the player 1’s trusting behaviors.  A player 1 from village that has higher per capita income could very probably have higher family income than that from a village with lower per capita in- come.  The arguments of the family income on the player 1’s trusting behaviors are pre- sented in previous sections.  Nonetheless, an administrative village with higher per capita income could also reflect the openness of the village to the outside world.  The argument for the effect of the openness to the outside world on the player 1’s decision is similar to the one as discussed above. Ethnic diversity in a certain administrative village could also affect the player 1’s trusting behaviors.  Ethnically homogeneous villages could imply that people could have more ties and social interactions than ethnically heterogeneous villages.  More social in- teractions and stronger ties could affect the player 1’s beliefs, which in turn would affect his trusting behaviors.  103  3.6. Results and Discussions 3.6.1 Descriptive Statistics Table 3-5 lists some descriptive statistics for the surveyed villages.  Table 3- 5:  Characteristics of Surveyed Administrative Villages Variables Mean Std. Dev. Range Obs. No of natural villages in administrative villages 9 7 2-40 30 Distance between 2 farthest natural villages (km) 10 6 2-30 30 Annual income per capita  (US$) 220 118 82-542 30 Percentage of households having phones in administrative villages (%) 58 25 10-100 30 Distance to the county's capital township      (km) 41 23 0-82 30 Distance to the nearest financial institutions  (km) 9 11 0-45 30 No of ethnic groups in administrative villages 4 2 1-10 30 Average number of households in natural villages (families/natural village) 85 58 4-243 30 Population density of natural villages (persons/per natural village) 361 259 20-1038 30 Percent of CCP Members in the total village population (%) 12 4 6-22 30 Whether or not the village has lineage halls (1=Yes; 0=No) 0.77 0.42 0,1 30   On average, the surveyed villages are relatively remote, poor, sparsely dispersed and ethnically heterogeneous as shown in the Table 3-5.  On average, the distance between the center of the surveyed villages and the nearest financial institutions, where are the local market placed are located, is about 9 km.  The annual income per capita is about US$ 220, which is slightly above the China’s most recently specified poverty line (US$ 175).   The average distance between natural villages on the two ends (the farthest ends) is about 10 km, indicating that the natural villages are quite dispersed.  The natural villages in surveyed administrative villages are quite heterogeneous in terms of size.  An  104  average natural village has a size of 85 households and a population density of about 361 people per (administrative) village.  In some administrative villages, the population den- sity is quite low (20 people) and the village size is quite small with only 4 families39.   On average, the administrative village has four ethnic groups residing together.   In terms of institutions on the village level, on average, 12% of the villagers in surveyed administra- tive villages were CCP members; about 77% of surveyed villages have lineage halls. Some additional correlation analysis show that the income per capita in surveyed vil- lages seems to be highly correlated with the distance to the nearest financial institutions, the distance to the capital townships of the counties and the percentage of the households having phones, which are three main variables to measure the (administrative) villages’ remoteness and their openness to the market and outside world40. In terms of the player 1s’ individual demographic characteristics, they had an average education level of 6 years and were 47 years old on average.  About 96% of the player 1s was male, given that in rural Yunnan men are commonly the household heads making financial decisions. Table 3-6 presents descriptive statistics of the amount sent, the player 1’s expected re- turn and the surveyed trust index.  39 In surveyed villages, only 1village has a natural village having an average household number of less than 10.  The random sampling method we used to select natural village happened not to select the natural vil- lage that has less than 10 households. 40 The correlation coefficient between the income per capita and the telephone rate is positive at 1% signi- ficance level, while the correlation coefficient between income per capita and distance to the county town- ship (or distance to the nearest credit) is negative at 1% significance level.  In addition, the correlation coef- ficient between the telephone rate and the distance is negative at 1% significance level.  105   Table 3- 6: Descriptive Statistics for Trust-related Variables Variable Means  Correlation with Amount Sent Correlation with Surveyed Trust Amount sent   (yuan) 8.38 1 0.20*** Expected return (yuan) 8.83 0.90*** 0.21*** Surveyed trust index -1.80E-09 0.20*** 1 Notes: * **and * denote significance at 1% level and 10% level, respectively.  Table 3-6 shows that: [1] the amount sent by player 1 has a positive and strong corre- lation with both the amount expected and the surveyed trust; [2] the player 1’s expected return has a very strong correlation with the amount sent both in terms of the magnitude of correlation coefficient (0.90) and the significance level (1%). 3.6.2 Main Results We used the amount sent by the player 1 as the dependent variables in the regression models to identify the determinants of the player 1’s trusting behaviors.  First, we ana- lyzed the effect of the player 1’s individual and household characteristics on the amount sent.  Then, we included some individual-level or group-level proxies of the player 1’s social interactions and past experiences with his peer members, which are believed to be able to affect the player 1’s beliefs about his anonymous partners’ reciprocal or trustwor- thiness. Effects of the Player 1’s Individual and Household Characteristics We first ran an OLS model.  Then, considering that the choices of sending RMB 0, 5, 10, 15 and 20 yuan that were available to the player 1 both have cardinal and ordinal proper-  106  ties, we also ran an interval regression model as alternative model specification.   For both model specifications, we included (ADMIN OR NATURAL) village dummies to control for the unobserved village fixed-effect.  Table 3-7 presents the results for OLS and interval regression models.   Table 3- 7: Effect of Individual and Household Characteristics on the Amount Sent Dependent Variable: Amount Sent OLS Interval Regression Age 0.379(0.205) 0.259(0.174) Age-squared -0.004(0.002) -0.003(0.002) Family size 0.319(0.218) 0.296(0.187) Education level 0.314(0.120)** 0.266(0.093)*** Total family income 6.40E-06(2.60E-05) -1.95E-06(1.98E-05) Village dummies YES YES Constant -3.667(5.475) 1.615(4.373) R-squared 0.1726  /lnsigma 1.510(0.034)***   sigma 4.526(0.155) Log pseudolikelihood -402.445 Observations 294   294 Notes:    (1) ***,** and * represent significance level of 1%, 5% and 10%, respectively.               (2) For OLS model, White-heterokedasticiy robust standard errors are in parenthesis.               (3) Village dummies are included in regressions.  The results of the OLS model are essentially consistent with the results of interval re- gression as shown by the Table 3-7.  As expected, education level has positive and signif- icant effect on the amount sent in all model specifications expect for the model, in which the player 1’s expected amount is included.  Age has a concave effect on the player 1’s trusting behaviors in all regressions, indicating player 1s tend to be more “trusting” when they get old but their “trusting” level is decreased after they reach a certain age.  Regard- ing the effect of household characteristics, the player 1’s family size has a positive but  107  insignificant effect on the amount sent and the family income also has a positive but in- significant effect. In the following sections, we analyze the effects of the player 1’s individual characte- ristics and his beliefs about his anonymous partner’s “reciprocal or trustworthy” actions through the lens of social interactions, past experiences and village institutions.  Given that interval regressions could fit our data better given the nature of the dependent varia- ble, the amount sent41 we use interval regressions for the remaining analysis. Effects of Individual-level Proxies for Social Interactions and Past Experiences We ran three different models to check the robustness of the individual-level proxies of social interactions and past experiences on the player 1’s trusting behaviors.   In Model 1, we only included the proxies related to gift exchange and the dummy variable related to help.  As discussed in section 3.5., experience with gift exchange and the level of help obtained from per villagers could be endogenous to the amount sent due to the unob- served family relationship between the player 1’s family and his peer villagers.  There- fore, in Model 2, we used two alternative individual-level proxies, which could make the observation of peer villagers’ behaviors possibly available.  In Model 3, we ran a full model by including all 4 individual-level proxies. In all 3 regression models, we included the player 1’s individual demographic and household characteristics to control for the variations in the individual characteristics and village dummies to control for the unob- served village fixed-effect.  Table 3-8 presents the results for the three regression models  41 We also tried to use OLS model specifications for the analysis in the remaining part of the paper; we found that the results from the OLS regression models and interval regression models are essentially un- changed.  Nonetheless, given the cardinal and ordinal nature of the available choices for the player 1 to send a certain amount to his partner, we think the interval regressions can  fit our data set in a better manner.  108  Table 3- 8: Effect of Individual-level Social Interactions and Past Experiences on Trusting Behaviors Dependent variable: The Amount Sent  (Interval Regressions) (1) (2) (3) Age 0.238 (0.175) 0.318 (0.190) 0.300 (0.190) Age-squared -0.002 (0.002) -0.003 (0.002) -0.003 (0.002) Family size 0.330 (0.188)* 0.213 (0.190) 0.234 (0.189) Education level 0.276 (0.093)*** 0.220 (0.094)** 0.230 (0.094)** Total family income -9.07E-06 (1.81E-05) -3.10E-06 (1.99E-05) -9.53E-06 (1.81E-05) Gift exchange 0.086 (0.030)*** 0.087 (0.032)*** Marital-funeral dummies -0.814 (0.624) -0.692 (0.624) Ratio of average distance from forest plots to the road on household level to village level -5.913 (2.775)** -5.892 (2.666)** Ratio of average distance from forest plots to home on household level to village level 6.104 (4.675) 5.571 (4.608) Village dummies YES YES YES Constant 2.005 (4.335) 0.962 (4.626) 1.191 (4.594)  /lnsigma 1.493 (0.034)*** 1.499 (0.035)*** 1.484 (0.035)***   sigma 4.451 (0.152) 4.479 (0.157) 4.411 (0.153) Log pseudo-likelihood -386.096 -398.034 -386.096 Observations 294   284   284 Notes:    (1) ***, ** and * represent significance level of 1%, 5% and 10%, respectively.               (2) Robust standard errors are in parenthesis.    109  We found strong evidence that the individual-level social interactions and past expe- riences could have significant effect on the player 1’s trusting behaviors.  While positive social interactions and good past experiences have a positive effect on the amount sent by the player 1 in the trust game, the negative social interactions and undesired past expe- riences reduce the amount sent. As shown in the Table 3-8, two individual-level proxies that represent positive social interactions and possible good past experiences, i.e. variables related to the gift exchange and to the distance between the forest plots and the player 1’s home, have positive effect on the amount sent.  The higher proportion of the expenditure on the gift exchange in the total family expenditure has a positive effect on the amount sent at 1% significance level as shown in columns (1) and (3), while the ratio of the average distance between forest plots and the player 1’s home on household level to that on village-level has a positive but insignificant effect on the amount sent as shown in columns (2) and (3). As discussed in the section 3.5, the higher proportion of the gift exchange expenditure in the total fami- ly expenditure could reflect positive social interaction between the player 1’s family and peer villagers, leading to the player 1’s stronger beliefs about reciprocal actions to be tak- en by the anonymous peer villager, who was paired with him, in the trust game.    Simi- larly, compared with the average peer members’ families, the nearer the distance between the average forest plot of the player 1’s forest plots and his home, the more possible that the player 1 and his family could have good past experiences with peer villagers in terms of helping each other to take care of their forest plots.  As a result, the player 1 could have stronger beliefs about reciprocal actions to be taken by his anonymous partner and consequently send a higher amount to the partner in the trust game.  110  In contrast with the effect of the positive social interactions and past experiences, the negative past experiences have negative effect on the amount sent by the player 1 in the trust game.  Table 3-8 shows that compared with a player 1 from a family that only relied on itself to deal with important family event, such as weddings and funerals, a player 1 from a family, which received help from peer villagers in the past, would send a signifi- cant higher amount to the anonymous partner.  Moreover, compared with the average peer members’ families, the nearer the average distance between the player 1’s forest plots and the main road, a significantly less amount of money (at 10% significance level) was sent by the player 1 in the trust game.   As discussed in Section 3.5.2, this could be due to the reason that the nearer the distance between the average forest plot of the player 1’s forest plots and the main road, the less possible that the player 1 and his family had observed the untrustworthy behaviors, such as timber stolen from the forest plots, regard- ing peer villagers.  As a result, the player 1 would have stronger beliefs about his peer members’ trustworthiness and the higher amount he might send in the trust game. It is also important to note that the marginal effect of the player 1’s individual demo- graphic and household characteristics on the amount sent as shown in columns (1)-(3) in the Table 3-8 is largely consistent with that in the Table 3-7.  In all 3 regression models, the player 1’s education level has positive and significant effect on the amount sent, indi- cating that the players having higher education are more optimistic about their anonym- ous partners’ reciprocal or trustworthy actions and about having a win-win situation than those having lower education.  The positive and significant effect of the player 1’s educa- tion level on the amount sent by him could be possibly explained by the reasons dis- cussed in Section 3.5.    Similar to the results presented in the Table 3-7, the amount sent  111  is first increased with age and then decreased with age; the family size and family income both have positive but insignificant effects on the amount sent. The Effect of Village Institutions and Village Characteristics As discussed in Section 3.5, CCP is a major modern formal grassroots institution estab- lished in villages, while a lineage institution serves as a traditional informal institution at the village level.  We posited that the higher ratio of CCP members to the total village population could indicate that the CCP system has been better established, while the exis- tence of the lineage halls in the villages indicates the extent of the influence of the tradi- tional informal institutions.   In order to determine the marginal effect of the village insti- tutions and village characteristics on the player 1’s trusting behaviors in trust games, we ran three regressions.  Model 1 only includes the player 1’s individual characteristics, in- dividual-level proxies for social interactions and past experiences as well as village insti- tutional variables.  Model (2) only includes the player 1’s individual characteristics, indi- vidual-level proxies for social interactions and past experiences as well as village socio- economic variables.  Model (3) is the full model including all variables included in the Models (1) and (2).  We also included county dummies to control for the unobserved he- terogeneity across the surveyed countries.  Table 3-9 presents the results of the three models.  112   Table 3- 9: The Effect of Village Institutions on Trusting Behaviors Dependent variable: The Amount Sent  (Interval Regressions) (1) (2) (3) Age 0.249 (0.175) 0.267 (0.180) 0.139 (0.205) Age-squared -0.002 (0.002) -0.003 (0.002) -0.001 (0.002) Family size 0.298 (0.188)* 0.245 (0.205) 0.266 (0.208) Education level 0.280 (0.093)*** 0.248 (0.101)** 0.204 (0.102)** Total family income 6.50E-06 (1.81E-05) -1.06E-05 (2.07E-05) -2.09E-06 (1.64E-05) Gift exchange 0.072 (0.032)** 0.074 (0.031)** Marital-funeral dummies -0.818 (0.612) -0.756 (0.605) Ratio of average distance from forest plots to the road on household level to village level -5.523 (2.790)** -5.310 (2.764)** Ratio of average distance from forest plots to home on household level to village level 5.145 (4.321) 5.713 (4.212) Ratio of CCP members to total village population -17.229 (9.467)** -16.031 (7.937)** -31.714 (7.346)*** If the village has lineage halls 0.674 (0.628) 0.615 (0.508) 0.871 (0.506)* Village per capita income -1.56E-04 (4.61E-04) Percent of families having phones in total no of families -2.884 (1.031)*** Population density in natural villages -0.005 (0.001)*** Ethnic diversity 0.954 (1.899) County dummies YES YES YES Constant 2.005 (4.335) 3.245 (4.181) 12.992 (5.048)  /lnsigma 1.493 (0.034)*** 1.528 (0.031)*** 1.502 (0.032)***   sigma 4.451 (0.152) 4.609 (0.144) 4.490 (0.143) Log pseudo-likelihood -386.096 -393.473 -385.342 Observations 294   284   284 Notes:    (1) ***, ** and * represent significance level of 1%, 5% and 10%, respectively.               (2) Robust standard errors are in parenthesis.               (3) Standard errors are clustered by villages.  113  Table 3-9 shows that some key individual-level characteristics, such as the player 1’s education level, and the individual-level proxies, including the family gift exchange ex- perience and the potentially observed activities on the family’s forest plots, are consis- tently significant with the same signs as those shown in the Tables 3-7 and 3-8.  This fur- ther shows the robustness of these key variables to the model specifications. Strong evidence is presented in the Table 3-9 that the ratio of CCP members has neg- ative effect on the player 1’s trusting behaviors at a 1% significance level.   The sign and the significance of this variable are robust to the model specifications, in which the vil- lage socio-economic characteristics are controlled for (column 1) or not controlled for (column 3).  Although the lineage institutional variable is not significant for the amount sent, it consistently has positive effect in both models when the village socio-economic characteristics are controlled for (column 1) and not controlled for (column 3).  The nega- tive and significant effect of CCP institutional variable and the positive effect of the exis- tence of the lineage halls in the village indicate that holding other things constant, in vil- lages where the CCP institution has more influence, the players seemed to send a signifi- cantly lower amount to their anonymous partners while in villages where the lineage halls are in existence, the players sent a higher but not significantly higher amount. Because the variable related to the CCP ratio and that related to the lineage halls represent the strength of formal and informal institutions, respectively, the significant and negative effect of the CCP variable and insignificant but positive effect of the lineage hall could shed some light that it could be possible while the formal CCP institution has grad- ually substituted the traditional informal lineage institution, it has not effectively taken  114  the roles that the traditional informal institution, the lineage halls, had been taking.  As a result, it could be possible that the formal CCP institution has crowded out the informal institutions and possibly eroded the mutual trust among the rural villagers in Yunnan Province.   Therefore, our findings provide some field evidence that the formal village institution crowded-out the informal institution in terms of maintaining the mutual trust among the rural villagers.  Nonetheless, in order to make conclusive evidence, some fur- ther rigorous analysis could be conducted in future’s research. The proportion of households having telephones is negatively significant in predicting the player 1’s trusting behaviors as shown in the models (2) and (3) in the Table 3-9. This essentially says that player 1’s from those villages having higher proportion of households that have telephones are less likely to send more to the player 2’s than those from villages that have the lower proportion.  Since the variable of the telephone propor- tion could potentially measure the openness of a certain village to the outside world as indicated by its strong correlation with the distance between the village and the nearest credit center as well as the distance between the village and the county capital city, the negative sign of this variable means that people from closed communities have higher level of trust.   The higher level of trust in closed communities might be due to the reason that members in the closed communities tend to rely on each other more, have more inte- ractions and thus trust each other more. The population density in natural villages has negative effect on the player 1’s trust- ing behaviors in models (2) and (3).  The negative effect of the population density in nat- ural villages could be caused by higher competition among dense populations for re-  115  sources at the stake.  This could also imply that the negative past experiences could affect the player 1’s beliefs about the actions to be taken by his anonymous partner. Ethnic diversity seemed not to be a significant determinant of the player 1’s trusting behaviors in our research, although other studies have found that ethnic diversity has been found to be significantly correlated with the rural villagers’ cooperation level and provisions of public good (Edward and Gugerty 2005). 3.6.3 Robustness Checks for the Effect of Beliefs on Trusting Behaviors As shown in our theoretical model, the player 1’s beliefs about the player 2’s trustworthi- ness should be a critical determinant for the trusting behaviors observed in the trust game. In this section, we attempted to check whether or not beliefs truly played significant roles in predicting player 1’s trusting behaviors. We used the survey measure of trust and the player 1’s expectation of the amount to be sent back by the player 2 as independent variables in the regressions.  Although the survey measure might involve some preference elements (Fehr 2009) 42, it still should be able to capture a large portion of the player 1’s belief about the player 2’s trustworthiness. As discussed in Section 3.3.1, we obtained an explicit measure of player 1’s beliefs (ex- pectation) about the player 2’s actions by asking the amount that he was expecting to re- ceive from the player 2.  If these two measures are strongly correlated with the trusting  42 Fehr (2009) however pointed out that the survey measure of trust could reflect a composite of prefe- rences and beliefs.   We posit that even the survey measure of trust could involve preference component, it still can largely capture the belief component.  116  behaviors, some light can be shed on the importance of beliefs in determining trusting behaviors. Regarding the effect of surveyed measure of trust on the trusting behaviors, there could be a potential endogeneity between the surveyed measure and trusting behaviors, due to potential measurement biases.  For example, one possible bias could be the upward bias between the survey measure and the trusting behaviors (Alberta and La Ferrara 2002): a respondent may feel “good” about himself if he answers affirmatively to the question about trusting others in surveys, while in his actual behaviors he may not be a trusting person.  When this bias is correlated with some other independent variables, such as education and family income, the endogeneity could arise. We attempted to address the potential biases using instrumental variables.  We used the proportion of elderly family members (65 years old and above) to the total family size and village dummies as instrumental variables for surveyed measure of trust.  The elderly family members, who are 65 years old and above, were born before the 1940s.  They grew up in traditional Chinese culture and were deeply influenced by traditional Chinese philosophy and norms, which emphasize the social orders, such as respecting the elderly, caring the young, being cautious, honest and humble in words and behaviors and among others.  We posited that a player 1 (respondent) from a family with the higher proportion of elderly family members than that from a family with lower proportion tend to be more truthful in answering questions related to surveyed trust.  Because some unobserved vil- lage characteristics could affect the way the respondent answering the questions, village dummies are also included as instrumented variables.  117  We ran 5 different models, 3 for the surveyed measure of the trust and 2 for the ex- pected amount to be returned.   For the surveyed measure of trust, we ran OLS, 2SLS and interval regression in order to check the robustness of the correlation between the survey measure and the experimental measure of trust across different model specifications.  For the expected amount to be returned, we ran OLS and interval regression.  For all models, we included village dummies in order to control for the heterogeneity of unobserved cha- racteristics across villages.  In all models, we also included individual and household cha- racteristics in order to control for the effect of other individual components that are not related to beliefs.   Table 3-10 presents the results for the models.   118    Table 3- 10: Surveyed Trust, Expected Return and Trusting Behaviors Dependent Variable: Amount Sent (1) OLS  (2) 2SLS  (3) Interval Regression (4) OLS (5) Interval Regression Surveyed trust 1.829 (0.568)*** 1.872 (0.901)** 1.462 (0.452)*** Amount expected to be returned 0.739 (0.034)*** 0.576 (0.030)*** Age 0.341 (0.219) 0.314 (0.219) 0.228 (0.174) 0.168 (0.086)* 0.099 (0.078) Age-squared -0.003 (0.002) -0.003 (0.203) -0.002 (0.002) -0.002 (0.001)* -0.001 (0.001) Family size 0.300 (0.229) 0.219 (0.229) 0.281 (0.183) 0.086 (0.113) 0.109 (0.109) Education level 0.298 (0.124)** 0.336 (0.104)*** 0.254 (0.090)** 0.063 (0.055) 0.065 (0.049) Total family income 7.08E-06 (2.34E-05) 9.83E-06 (2.33E-05) -1.34E-06 (1.79E-05) -1.66E-05 (7.92E-06)* -1.88E-05 (6.59E-06)*** Village dummies YES YES YES YES Constant -1.884 (5.553) -2.383 (4.825) 3.033 (4.423) -1.494 (2.176) 3.293 (1.971) R-squared 0.2003 0.0773 0.8268  /lnsigma 1.492 (0.034)*** 0.734 (0.089)***   Sigma 4.445 (0.151) 2.083 (0.186) Log pseudo-likelihood -397.706 -207.255 Observations   294 294 294 291 291 Notes:    (1) ***, ** and * represent significance level of 1%, 5% and 10%, respectively.               (2) White-heterokedasticiy robust standard errors for OLS are in parenthesis.               (3) Village dummies are included in regressions.  119  As shown in models (1)-(3) of the Table 3-10, the surveyed measure was significantly and positively correlated with the amount sent, which means that the surveyed measure and the experimental measure are strongly correlated.  In all three models (1-3), the signs, the magnitudes and the significance levels of the included independent variables in OLS model are quite close to those in 2SLS model and interval regression model.    Our find- ing of strong correlation between the survey and experimental measure, which focus on the same dimension of the trust (i.e. the rural villagers’ mutual trust), is largely consis- tent with the findings of Naef and Schuff (2009), which showed that the survey and expe- rimental measures focusing on the same dimension of trust (i.e. trust in strangers) were strongly correlated.  Therefore, it is concluded that with careful designs, the survey and experimental methods could both effectively measure trust. Table 3-10 also shows that the expected amount to be returned is significantly and positively correlated with the amount sent both in OLS model and interval regression model.  This indicates that the strong correlation between the expected return and the amount sent is robust to model specification. The Table 3-10 also indicates beliefs could be an important determinant of the player 1’s trusting behaviors.  The surveyed trust and the expected amount to be returned largely capture the player 1’s beliefs about the trustworthy actions to be taken by his anonymous partner.  Therefore, the strong correlation between the surveyed and the experimental measure of trust and that between the expected return and the trust indicate a strong cor- relation between the trusting behaviors and beliefs.   120  3.7. Conclusions Justified by a simple theoretical model, we attempted to uncover the determinants of rural villagers’ mutual trust through the lens of beliefs in determining the trusting behaviors measured by the amount sent by the senders in the trust games.  We specifically aimed at: [1] examining the effect of individual demographic characteristics, social interactions and past experience, village institutions and village characteristics on the trusting behaviors; [2] the explicit role of beliefs in determining trusting behaviors.   We intended to test two specific hypotheses: [1] the potential crowd-out effect of formal CCP institution on the informal lineage institution on the village level in sustaining rural villagers’ mutual trust; [2] the possible erosion of villagers’ mutual trust in villages where are open to the outside world.  To reach our objectives and test the hypotheses, we conducted household surveys and artefactual field experiments in 30 villages in Yunnan Province in southwest China. We found some initial evidence that the formal institutions established on village lev- el could have crowed out informal institutions and consequently could have eroded mu- tual trust among the villagers.  While the modern formal institutions established on gras- sroots levels, such as CCP institution, could have been gradually substituted the tradition- al informal institution, such as village lineage institution, it probably had not effectively take on the roles played by informal village institutions in maintaining rural villagers’ mutual trust.  Nonetheless, some rigorous and systematic analysis still need to be con- ducted in order to conclude the “crowd-out” effect of the formal village institutions on informal village institutions.  121  We found strong evidence that the individual social interactions and past experiences could critically determine the trusting behaviors in the trust games.  While positive social interactions and good past experiences could contribute to the mutual trust among the ru- ral villagers, the undesired social interactions and unhappy past experiences could have negative impact on the villagers’ mutual trust. We found a negative and significant impact of the village’s openness to the market place on both players’ behaviors in the trust game partially indicates the openness to market could possibly erode the bonding social capital, such as villagers’ mutual trust in China’s western Yunnan Province, which is the most ethnically diverse province, where many closed ethnic communities are reside. We also found strong evidence that the player 1’s beliefs about the reciprocal or trustworthy actions to be undertaken by his anonymous peer villager were a critical de- terminant of the player 1’s trusting behaviors in the trust games.   Besides the beliefs, the player 1’s education level was also a critical determinant for the trusting behaviors. Therefore, trusting behaviors are not only motivated by the players’ beliefs but also are determined by individual demographic characteristics.   122  Bibliography Alesina, A. & La Ferrara, E. (2002).  Who trusts others?  Journal of Public Economics, 85(2), 207-234. Asquith, N.M., Vargas, M.T., & Wunder, S. (2008).  Selling two environmental services: In-kind payments for bird habitat and watershed protection in Los Negros, Bolivia. Ecological Economics, 65(4), 675-684. Barr, A. (2003). Trust and expected trustworthiness: Experimental evidence from Zim- babwean villages. Economic Journal, 113(489), 614-630. Bellemare, C., & Kroger, S. (2007).  On representative social capital.  European Econom- ic Review, 51(1), 183-202. Berg., J., Dickhaut, J., & McCabe, K. (1995). Trust, reciprocity and social history. Jour- nal of Economic Behavior and Organization, 10, 122-142. Bertrand, M., & Mullainathan, S. (2001). Do people mean what they say? implications for subjective survey data.  American Economic Review, 91(2), 67-72. Buchan, N.R., Croson, R. & Solnick, S. (2008).  Trust and gender: An examination of behavior and beliefs in the investment game.  Journal of Economic Behavior & Or- ganization, 68(3-4), 466-476. Burks, S. V., Carpenter, J. P., & Verhoogen, E. (2003). Playing both roles in the trust game. Journal of Economic Behavior and Organization, 51(2), 195-216.  123  Charness, G. & Rabin, M. (2002).  Understanding social preferences with simple tests. The Quarterly Journal of Economics, 117(3), 817-869. Chaudhuri, A., & Gangadharan, L. (2007). An experimental analysis of trust and trust- worthiness. Southern Economic Journal, 73(4), 959-985. Croson, R. & Buchan, N. (1999).  Gender and culture: international experimental evi- dence from trust games.  American Economic Review, 89(2), 386-391. Dasgupta, P. (2005). Economics of social capital.  Economic Record, 81(S1), S2-S21. Durlauf, S.N. & Fafchamps, M. (2004).  Social capital.  NBER Working Papers (10485), National Bureau of Economic Research, Inc. Eckel, C. C., & Wilson, R. K. (2004). Is trust a risky decision? Journal of Economic Be- havior and Organization, 55(4), 447-465. Edward, M., & Gugerty, M. K. (2005). Ethnic diversity, social sanctions, and public goods in Kenya.  Journal of Public Economics, 89, 2325-2368. Fafchamps, M. (2006).  Development and social capital.  Journal of Development Studies, 42(7), 1180-1198. Fehr, E. (2009).  On the economics and biology of trust.  Journal of European Economic Association, 7(2-3), 235-266. Fishburn, P.C., Porter, R.B.  Optimal portfolios with one safe and one risky asset: Effects of changes in rate of return and risk.  Management Science, 22 (10), 1064-1073.  124  Fukuyama, F. (1995). Trust: the social virtues and the creation of prosperity. London: Hamish Hamilton. Geanakoplos, J., Pearce, D., & Stacchetti, E., 1989. Psychological games and sequential rationality.  Games and Economic Behavior, 1, 60–79. Glaeser, E., Laibson, D., Scheinkman, J.A., & Soutter, C.L. (2000). Measuring trust. Quarterly Journal of Economics, 115 (3), 811-846. Guiso, L., Sapienza, P., & Zingales, L. (2008).  Trusting the stock market.  Journal of Finance, 63 (6), 2557-2600. Holm, H., & Nystedt, P. (2008). Trust in surveys and games-A methodological contribu- tion to the influence of money and location. Journal of Economic Psychology, 29, 522-542. Jacka, T. (2009).  Increasing women’s participation in village government in China.  Crit- ical Asian Studies, 40 (4), 499-529. Johansson-Stenman, O., Mahmud, M., & Martinsson, P. (2005). Does stake size matter in trust games? Economics Letters, 88 (3), 365-369. Karlan, D. S. (2005). Using experimental economics to measure social capital and predict financial decisions. American Economic Review, 95(5), 1688-1699. Knack, S. (2001).  Aid dependence and the quality of governance: Cross-country empiri- cal tests.  Southern Economic Journal, 68(2), 310-329.  125  Knach, S., & Keefer, P. (1997). Does social capital have an economic payoff? A cross- country investigation. The Quarterly Journal of Economics, 112(4), 1251-1288. La Porta, R.  Lopez-de-Silane, F., Shleifer, A., & Vishny, R.W. (1997).  Trust in large organizations.  American Economic Review, 87 (2), 333-338. Lyon, F. (2003).  Community groups and livelihoods in remote rural Areas of Ghana: How small-scale farmers sustain collective action.  Community Development Journal, 38, 323-331. Mas-Colell, A., Whinston, M.D., & Green, J.R. (1995).  Mircoeconomic theory.  New York: Oxford oxford University Press. McCabe, K.A., Rigdon, M.L., & Smith, V.L. (2003). Positive reciprocity and intentions in trust games.  Journal of Economic Behavior & Organization, 52 (2), 267-275. Naef, M., & Schupp, J. (2009).  Measuring trust: experiments and surveys in contrast and combination.  Discussion Paper No. 4087, The Institute for the Study of Labor (IZA), Bonn, Germany. Narayan, D. (1999). Bonds and bridges: social capital and poverty. Policy Research Working Paper 2167. Washington, D.C: World Bank. North, D. (1990).  Institutions, institutional change and economic performance.  Cam- bridge: Cambridge University Press. Ostrom, E., & Nagendra, H. (2007).  Governing the commons in the new millennium: a diversity of institutions for natural resource management. In C. J. Cleveland (Eds.),  126  Encyclopedia of Earth.  Washington, DC: Environmental Information Coalition, Na- tional Council for Science and the Environment. Potter, S.H., & Potter J.M. (1990).  China's peasants: the anthropology of a revolution. Cambridge [England]; New York: Cambridge University Press. Pretty, J., & Ward, H. (2001).  Social capital and the environment.  World Development, 29, 209-227. Schechter, L. (2007). Traditional trust measurement and the risk confound: An experi- ment in rural Paraguay. Journal of Economic Behavior and Organization, 62 (2), 272-292. Sobel, J. (2002). Can we trust social capital?  Journal of Economic Literature, 40 (1), 139-154. Tsai, L.L. (2002).  Cadres, temple and lineage institutions, and governance in rural China. The China Journal, 48, 1-27. Uphoff, N. (2000). Understanding social capital: Learning from the analysis and expe- rience of participation. In P. Dasgupta, & I. Serageldin (Eds.), Social capital: A mul- tifaceted perspective (pp. 215-249). Washington, D.C.: World Bank. Xiao, T. (2006).  Clan and village governance in today's China.  Journal of Literature, History and Philosophy (in Chinese), 4, 156-163. Zak, P.J., & Knack, S. (2001).  Trust and growth.  Economic Journal, 111(470), 295-321.  127    Chapter 4  Risk Aversion and Farm Input Choice: Evidence from Field Experiments in China43  43 A version of this chapter will be submitted for publication.  Gong, Y., Baylis, K., Xu, J., Kozak, R. and Bull, G.  Risk aversion and farm input choice: evidence from field experiments in China.  128  4.1. Introduction Risk plays an important role in the investment decisions44 of individual farmers, for ex- ample, for technology adoption or input use decisions on their farmlands (Babcock and Shogren 1995; Coady 1993; Isik and Khanna 2003; Just and Pope 1978; Knight et al. 2003; Liu 2008; Pope and Kramer 1979).  Risk affects farmers’ decisions differently, de- pending on their capacity to absorb risk, their psychological attitudes or preferences to- wards risk, and the constraints that they are faced with in absorbing risk (Binswanger and Sillers 1983; Eswaran and Kotwal 1990; Knight et al. 2003). Most existing literature has focused on analyzing the effects of risk on farmers’ on- farm investment decisions through the lens of constraints that are faced by farmers to ab- sorb risk, such as limited access to credit markets (Masson 1972; Binswanger and Sillers 1983; Weil 1970; Croppenstedt et al. 2003) and information (Feder 1980), insufficient farm sizes (Coady 1993), and lack of human capital (Knight et al. 2003; Welch 1978). While certain theoretical models have been developed to specifically examine farmers’ risk preferences on their input use decisions (Feder 1980; Just and Zilberman 1983), em- pirical research in this regard has been scant, with only a few exceptions, including Binswanger (1980), Knight et al. (2003), Hill (2009) and Liu (2008). This lack of rele- vant empirical research could be partly due to difficulties in measuring risk preference (Feder et al. 1983). Nonetheless, the development of experimental economics over the last several decades has provided some useful tools and insights for measuring an indi- vidual’s psychological traits, such as risk preference.  44 Throughout the paper, we refer investment decisions generally as technology adoption or input use.  129  The standard experimental method used to measure individuals’ risk-preferences is to provide subjects with binary alternative risky choices that have different expected values and variances, specify the probabilities of being rewarded, and reward them with actual money based on the choices that they make. Because the alternatives with higher ex- pected payoffs are also associated with higher variances (i.e., riskier), an individual is forced to make trade-offs between his/her preference for risk and the expected gain. Ac- cordingly, his/her risk preference is revealed by the decision made between two choices. Two main approaches have been used to provide binary alternative choices to subjects: one is to provide subjects with lotteries by varying the prizes of winning, while keeping the probabilities as constant over lotteries (Bingswanger 1980); the other is to vary the probabilities, while keeping the prizes of winning as constant over lotteries (Holt and Laury 2002). Although risk experiments have been conducted in the past using farmers in field set- tings as subjects (Bingswanger 1980; Grisley 1980; Hill 2009; Knight et al. 2003; Liu 2008; Schechter 2003), empirical research still needs to be conducted to provide further evidence on whether or not risk preferences measured by field experiments can strongly predict risky behaviors of farmers in real-world decision-making. Our paper attempts to provide some concrete field evidence regarding the external validity of the field experi- ments by linking our experimental measure of farmers’ risk preference to their risk beha- vior revealed by their on-farm investment decisions.  130  We collected a unique data set by conducting artefactual field experiments45 and household surveys in 30 villages of Yunnan Province, a region of southwestern China where over 40% of the population was still living under the national poverty line as of 2008. The field experiments were used to elicit farmers’ risk preferences, while the ques- tionnaires were used to collect demographic and socio-economic information. Using our unique data set, we focused our analysis specifically on household decisions regarding the intensity of chemical fertilizer and pesticide use.  We posited that because the appli- cation of these inputs on farmlands always involves a possible profit gain and a potential net loss due to uncertainty in agricultural production, farmers’ decisions on the intensity of chemical fertilizer use and pesticide use could be used to reveal their attitudes towards risk. Therefore, we hypothesize that all other things being equal, chemical fertilizer and pesticide use on farmlands would be more common among the more risk-averse farmers. The logic here is simply that risk-averse farmers would want to stabilize their agricultural production in order to reduce the potential risk of unstable production levels and detri- mental impacts to their livelihoods. We also posited that a family’s social networks would be a determining factor in their ability to pool resources to absorb random shocks in income and set about to analyze this question. Lastly, since farmland size is undoub- tedly a critical constraint that a family is faced with in making farm investment decisions (Just and Zilberman 1985), we also pay close attention to its effects on household input use decisions.  45Typically, conventional experiments employ a standard subject pool of students, abstract framing, and an imposed set of rules. Artefactual field experiments are similar, but use non-student subject pools.  For a detailed discussion on the taxonomy of experiments, please refer to Harrison & List (2004).  131  The paper has two main objectives: [1] to analyze whether or not the experimental measure of risk preferences for the family head – the decision-maker of the household – can strongly predict household decisions regarding the intensity of chemical fertilizer and pesticide use on the family’s farmland; and [2] to analyze the effects of the farmland size and other household socio-economic characteristics, as well as a family’s social networks, on household decisions regarding the intensity of chemical fertilizer and pesticide use on the family’s farmland. The paper has potential to inform both empirical studies within this domain and poli- cy-making. It supports the external validity of field experiments by providing concrete field evidence and sheds some light on the importance of developing means, by which farmers’ risk preferences can be influenced. If China aims to address the current issue of farmers’ overuse of chemical fertilizers and/or pesticides leading to environmental degra- dation and compromised human health (Hu et al. 2007; Huang et al. 2003; Huang et al. 2008), then policy and interventions must serve to induce farmers to reduce use of these chemicals on their farmlands. 4.2. Risk Aversion, Its Measurement and Relations to Farmers’ In- vestment Decisions Risk aversion is often modeled as the curvature of the expected utility function in the standard expected utility approach, where utility is defined with initial wealth. Two measures of risk aversion, absolute risk aversion (Pratt 1964) and relative risk aversion (Arrow 1971), have been commonly used to express the degree of risk aversion.  132  Two main methods have been used to measure individual risk preference. One is the econometric method pioneered by Moscardi and de Janvry (1977) to empirically derive farmers’ risk aversion coefficients from their observed behavior and socio-economic cha- racteristics. The methods applied by Antle (1987, 1988) also fall under this category. The other refers to experimental methods used to observe choices that subjects make over lot- teries with different expected payoffs and variances.  While the econometric method is thought to confound risk behavior with farmers’ resource constraints, as cited by Wik et al. (2004) from Eswaran and Kotwal (1990), the experimental method is criticized for reasons of external validity (Carpenter 2002; Lusk et al. 2006). Nonetheless, the experi- mental method has increasingly been recognized as a more desired method for eliciting and measuring risk. Risk is believed to play an important role in the investment decisions of individual farmers (Babcock and Shogren 1995; Coady 1993; Isik and Khanna, 2003; Just and Pope 1978; Knight et al. 2003; Liu 2008; Pope and Kramer 1979).  Risk in production is a strong characteristic of agricultural production. Although risk is closely associated with agricultural production, it can be controlled, to some degree, by farmers through the use of modern inputs, such as chemical fertilizers and pesticides (Feder 1980; Just and Zil- berman 1983). Risk in agricultural production can be exogenously-caused or endoge- nously-induced. While exogenous risk, which may arise from extreme weather conditions or threats of disease and pest outbreaks, is independent of farmers’ production decisions, endogenous risk is incurred solely by such production decisions. In other words, while the pest outbreaks can be categorized as exogenously-caused, the use of pesticide to con-  133  trol for pest outbreaks is endogenously-induced by farmers’ decisions (Knight et al. 2003). The specific relationship between farmers’ risk aversion and their decisions regarding input use on farmlands has been analyzed with two theoretical models developed under the framework of technology adoption with production uncertainty (Feder 1980; Just and Zilberman 1983). Feder’s (1980) model assumes that modern technology can lead to a higher variability in outputs (uncertainty) than traditional technology and that no fixed costs are required for the adoption of the technology. His model predicts that the degree of farmers’ risk aversion has no bearing on farmers’ decisions regarding the intensity of fertilizer applied per acre of land. Just and Zilberman (1983) extended Feder’s (1980) model by assuming a fixed cost of the adoption, which could interact with the stochastic structure of production in determining the relationship between farm size and adoption decisions. Their model shows that the intensity of the modern input use depends on: [1] the nature of the risk associated with the input, and specifically, whether it is risk- increasing or risk-increasing; and [2] the properties of relative risk-aversion, i.e., whether it is decreasing, constant, or increasing. Assuming that chemical fertilizer use is a risk- increasing activity, their model predicts that farm size ought to be negatively correlated with chemical fertilizer use if the relative risk-aversion is decreasing. On the other hand, if the relative risk-aversion is increasing, farm size should be positively correlated with the chemical fertilizer use.  134  4.3. Methods 4.3.1. Survey Design and Data Description Survey Design In the fall of 2007, household surveys and risk experiments were conducted in six coun- ties located in different regions of Yunnan Province in southwest China.  Five administra- tive villages were randomly selected from each county for a total of 30 villages in the sample frame. In each administrative village, two natural villages, which are sub-units of the administrative villages, were randomly chosen and within each natural village, 10 households were randomly chosen to take part in the study. Specifically, the family head from each randomly selected household was asked to participate in surveys and games. In total, 300 respondents were surveyed and participated in risk experiments. Household surveys were conducted prior to the risk experiments and were used to collect the following information: [1] household demographic and socio-economic cha- racteristics; [2] information on the total area of farmland operated by the farmers’ house- holds; [3] detailed information on the types of crops planted in 2006 and the correspond- ing sown area for each type of crop; [4] detailed information on the expenditures for var- ious types of input for each crop, including chemical fertilizers, pesticides, seeds, labor, and machinery; and [5] information on families’ social relationships and participation in social networks. Survey Data Description Among the 300 households that were surveyed, 290 were able to provide information on the costs of chemical fertilizers and pesticides. We noticed that, in our original data set,  135  two of the respondents’ ages were 80 and 81 years old and uneducated and suspected that they could be potential serious outliers that could affect our ability to make statistical in- ferences. This was verified using box plot graphs and these two outliers were eliminated. We opted for this approach, as opposed to transforming the data set, for a number of rea- sons, including: the complexity of the survey and experiment for those over 80 years in age; the already high proportion of older respondents (60 years or more) in our sample; and the lack of skewness in our age distributions without the outliers46. Table 4-1 pro- vides the summary statistics for the key variables of interest from the household surveys (excluding the outliers described above).  Table 4- 1: Summary Statistics Variable Mean Standard Deviation Range Number of Observations Age (years) 46.39 11.42 23-77 290 Male 0.97 0.18 0,1 290 Family size (persons) 4.72 1.52 1-12 290 Family head's education level (years) 5.39 3.30 0-15 290 Total value of assets (yuan) 36775.48 51232.24 30-309897 286 Total farmland operated (ha) 0.52 0.43 0.01-2.67 290 Total sowing area (ha) 0.61 0.44 0.03-2.67 290 Proportion of non-agricultural income in total family income 0.34 0.32 0-0.95 290 Unit cost of chemical fertilizer (yuan/ha) 2048.23 1486.56 115-12500 290 Unit cost of pesticide (yuan/ha) 468.86 505.77 0-3409 290 Belonging to big family names 0.45 0.50 0,1 290 Belonging to big ethnic groups 0.90 0.30 0,1 290 Participation in village-wide networks 0.17 0.37 0,1 290 Family has members in Chinese Com- munist Party 0.20 0.40 0,1 290   46 Interestingly, our decision seemed justified after the analysis in that there were statistically significant differences in our econometric models with and without the outliers.  136  On average, surveyed respondents had an average family size about 5 persons and had acquired elementary school education (5.39 years). The average age of surveyed res- pondents was about 47 years old. As discussed earlier on, the family head of a randomly selected family was asked to participate in our surveys, and in our set, about 97% of family heads were male. The extremely high portion of male family heads could be attri- buted to two reasons. First, in rural Yunnan, there is a strong tradition of men dominating family decisions. Second, in many instances, females may be the true household heads, but they are not used to participating in interviews. All that being the case, the reader should approach results on the head of households with some caution. For the most part, families were operating on a very small farm, but their family in- comes were mainly derived from agricultural activities. In 2006, the average size of farm- lands operated by farmers’ families was only about 0.52 hectares (ha) with an average sown area of about 0.61 ha per family due to inter-planting of crops. Given that the aver- age family size was 5 persons, this translates into a per capita average farm size of 0.08 ha per person. Table 4-1 shows that, despite small farmland sizes, families’ non- agricultural income accounted for only about 34% of their total incomes.  This implies that the respondent families in our data relied heavily on small farmlands to satisfy many of their subsistence needs. The value of total assets in 2006, including the estimated value of durable goods and productive assets that a family owns, was used as a proxy for family wealth levels. Res- pondents were asked to provide information on the year that they purchased/acquired as- sets (for both durable goods and productive assets) and an estimate of the value of assets  137  that they held in 2006. In this year, the average total asset value for responding families was estimated to be RMB (yuan) 36,775 (equivalent to US$ 5,384 in 2009). Correlation analysis indicated that family wealth level was significantly and positively correlated with the proportion of non-agricultural income relative to total family income, indicating that the wealthier families relied less on agricultural activities. The intensity of pesticide use was approximated by the unit cost of pesticide, calcu- lated as the total costs of pesticide applied to crops by a family divided by the size of the farmland operated by the family in 2006. Similarly, the intensity of the chemical fertilizer was approximated by the unit cost of chemical fertilizer, calculated as the total cost of fertilizers applied to crops by the family divided by the size of the farmland operated by the family in 2006. In 2006, the respondent families invested an average of RMB 469 per ha (equivalent to US$ 69 per ha in 2009) for pesticides and RMB 2,048 per ha (equiva- lent to US$ 300 per ha in 2009) for chemical fertilizers. We noticed that the distributions of both the unit pesticide costs and unit chemical fertilizer costs were highly right-skewed (skewness coefficients of -0.23 and -0.47, re- spectively). As such, we attempted log transformations on both variables to normalize the data. Since all of the surveyed families used chemical fertilizers in 2006, taking log trans- formations for this cost data did not cause any problems with discarding observations from the analysis. However, 26 households (accounting for slightly less than 9% of the total of 290 valid observations) did not apply pesticides on their crops in 2006. This re- sulted in this portion of data being dropped from the analysis with the log transformation. That said, both variables followed approximately normal distributions after the log trans-  138  formation and we, therefore, opted to use the transformed data in our statistical inference at the expense of including only 259 observations in our final econometric analysis. Finally, in our data set, about 17% of the surveyed respondents’ families participated in some form of village-wide social network in 2006 and about 20% of families had members belonging to Chinese Communist Parties (CCP) members. 4.3.2. Field Experimental Design and Data Description Field Experimental Design The risk experiments were conducted after the surveys with the survey respondents also being the participants in the experiments. Since no respondents refused to participate, there was no selection bias for the participants. In our experiment, we followed the methods of Holt and Laury (2002) with justifica- tions that are similar to those put forward by Harrison et al. (2007) 47. The lottery-choice decisions offered to subjects in our experiments, seen in Table 4-2, were similar to the low-payoff treatment in Holt and Laury’s (2002) experiment with minor modifications to the convenience of making payments48.   47 Harrison et al. (2007) justified that they adopted the method of Holt and Laury (2002) rather than that of Binswanger (1980) because the former has been widely implemented in recent laboratory experiments and requires a relatively transparent task. 48 If we converted the lottery choices expressed in US $ in Holt and Laury’s (2002) experiment into RMB, it would have resulted in a need to carry small coins. Therefore, we modified the payoffs for the sake of convenience such that only paper money was required.  139   Table 4- 2: The Ten Pair- wise Lotteries Decision Option A  (Probabilities and Payoffs) Option B (Probabilities and Payoffs) Difference in Expected Payoff 1 1/10 of ¥20,   9/10 of ¥16  1/10 of ¥35,   9/10 of ¥5  ¥ 8.4 2 2/10 of ¥20,   8/10 of ¥16  2/10 of ¥35,   8/10 of ¥5  ¥ 5.8 3 3/10 of ¥20,   7/10 of ¥16  3/10 of ¥35,   7/10 of ¥5  ¥ 3.2 4 4/10 of ¥20,   6/10 of ¥16  4/10 of ¥35,   6/10 of ¥5  ¥ 0.6 5 5/10 of ¥20,   5/10 of ¥16  5/10 of ¥35,   5/10 of ¥5  ¥ -2.0 6 6/10 of ¥20,   4/10 of ¥16  6/10 of ¥35,   4/10 of ¥5  ¥ -4.6 7 7/10 of ¥20,   3/10 of ¥16  7/10 of ¥35,   3/10 of ¥5  ¥ -7.2 8 8/10 of ¥20,   2/10 of ¥16  8/10 of ¥35,   2/10 of ¥5  ¥ -9.8 9 9/10 of ¥20,   1/10 of ¥16  9/10 of ¥35,   1/10 of ¥5  ¥ -12.4 10 10/10 of ¥20,   0/10 of ¥16   10/10 of ¥35,   0/10 of ¥5   ¥ -15.0 Note: ¥ = Chinese Yuan (RMB)   Both options in Table 4-2, A and B, had high and low payoffs, but option B is more “risky” than option A since its payoffs (RMB 35 and RMB 5) are more variable than the payoffs for option A (RMB 20 and RMB 16).  In the experiments that we conducted, a risk neutral subject, as an expected payoff maximize, would be expected to switch to op- tion B once the probability of getting high payoff reaches 0.5 and the expected payoff from option B is increased to be greater than that from the option A.   A risk-averse sub- ject would be expected to choose option A with low probabilities of gaining high payoffs and switch to option B at a certain probability level, depending on his/her utility function. While an extremely risk-averse person would only switch to option B in the bottom line of Table 4-2 (Decision 10), an extreme risk-taker person would always choose option B. The subjects were told that they would receive a real payment from the experimenters, but that the amount would depend exclusively on the choices that they would make in each of the 10 questions. Specifically, the subjects were told that, once they finished  140  answering all 10 questions, one question of these 10 would be selected at random and the choice they had made in the selected question would be used to determine their payment. Therefore, subjects understood that each question that they were going to answer would have an equal chance of being selected. The subjects were also told that that, irrespective of the choice that they would make between the two options in each question, there were always corresponding probabilities of receiving a high payoff and a low payoff. The experimenters used visual means to assist the subjects, typically rural farmers with relatively low education levels, in gaining a clear understanding of the probabilities of receiving a high payoff and a low payoff. The experimenter showed subjects 10 white balls marked with 1 to 10, respectively, and then put all of the balls in an empty opaque bag to make sure that neither the subjects nor the experimenters could see the marked balls. Next, the experimenters explained the concept of probabilities by means of the balls. For example, for the probability of 1/10 of getting high payoff (RMB 20 yuan) for option A, it was stated that: “if you draw a ball marked with 1 from the bag containing the 10 marked balls, then you will be paid RMB 20; if you draw any other balls, then you will be paid RMB 16 yuan.”  Similarly, for a probability of 3/10 of getting a high payoff (RMB 20 yuan) for option A, it was stated that: “if you draw a ball marked with 1, 2, or 3 from the bag, then you will be paid RMB 20; if you draw a ball marked with any number from 4-10, then you will be paid RMB 16 yuan.”  The experimenters worked through a number of such examples with subjects to help them understand the differences between the two options.  141  The game was formally started when the subjects had obtained a good understanding of the association between the choice that they made in each question and the final pay- ment that they would earn from the game.  The game was played in 3 steps: Step 1.  The subject was presented a list of questions as shown in the Table 4-2. He/she answered 10 questions one by one by making choices between the option A and the option B. A consistent subject would be expected to keep selecting Option B once he/she switched from A to B, while an inconsistent subject would possibly switch back and forth. However, it is also possible that subjects could mistakenly switch back and forth between the two options due to cognitive or non-cognitive reasons, such as a lack of motivation, impatience, or confusion over the game. We allowed for this possibility in the experiment. Step 2. After the subject finished answering the 10 questions, he/she was asked to randomly draw 1 out of 10 playing cards marked with numbers from 1 to 10.  Each card represented a certain question that the subject had answered; the card with number 1 cor- responded to question 1, the card number 2 represented question 2, and so on.  For exam- ple, if he/she happened to draw a card numbered 5, then the answer he/she had given in question 5 was used by the experimenter to make the payment. Step 3.  After question number was randomly selected, he/she was paid according to the answer he/she had given in the selected question. In order to determine the payoff, high or low, the subject was asked to draw one ball from the opaque bag containing 10 balls marked with numbers from 1-10. The experimenter then paid him/her according to  142  the probability of getting high or low payoff that was specified in the selected question. For example, the Decision 5 was randomly selected as illustrated by the example in the Step 2.  For the Decision 5, the subject had chosen option B.  In this step, if he/she drew a ball marked with the number 7, then the experimenter paid him/her RMB 5 yuan. Field Experiment Data Description In total, we were able to elicit risk-preference data for 292 subjects since we discarded two subjects who were 80 or more years old and an additional six subjects were invali- dated for various reasons49. Of these 292 subjects, 23 switched back from B to A, while the other 269 subjects switched over to option B only once. We considered these latter 269 subjects to be consistent subjects and decided to exclude the remaining 23 inconsis- tent subjects (8%) from our econometric analysis to minimize noise in the models. As discussed earlier, a risk-neutral subject would switch from option A to option B when the expected payoff from option B was bigger than from option A (corresponding to the fifth question in our experiment). Therefore, a risk-neutral subject could be ex- pected to choose four “safe choices” (option A). A risk-taker, on the other hand, would tend to make less than four safe choices, while a risk-averse subject would be inclined to make more than four. Table 4-3 presents the summary statistics for the outcomes of the risk experiments both for all 294 subjects and the 269 consistent ones.   49 For example, some subjects obviously did not understand the experiments even the experimenters tried to guide through them and showed some examples.  These subjects had to be discarded.  143    Table 4- 3: Outcomes of the Risk Experiments Number of Safe Choices All Subjects  Consistent Subjects Frequency Percent   Frequency Percent 1-3 90 31  71 26 4 38 13  36 13 5-9 164 56   162 60 Average number of safe choices 5.24 (2.66)   5.47 (2.59) Note: standard errors of the mean are given in parenthesis.   Our experimental outcomes confirm the statement of Young (1979) that farmers in developing countries can be reasonably assumed to be risk-averse and that their expected utility function is quadratic. This risk-averse behavior was certainly the case for our expe- rimental subjects. As shown in Table 4-3, the average number of safe choices chosen by the subjects was 5.24, over one more than the four safe choices that would typify a risk- neutral subject. Moreover, in terms of the relative frequencies, the Table 4-3 shows that more than half of the subjects in our risk experiments chose 5-9 safe choices, while roughly one-third of the subjects chose 1-3 safe choices. 4.4. Some Theory and Empirical Model Specifications Just and Zilberman (1983) incorporate the stochastic production model of Just and Pope (1978) into expected utility models to analyze the effect of farm size and farmers’ risk preferences on farmers’ decisions on the intensity of chemical fertilizer use and pesticide use.   Assuming a farmer having a total area of L ha of land, which is in full application to plant crops.  The land can be applied with traditional input (such as manure), or mod-  144  ern input (such as chemical fertilizer), the structural model for farmers’ decisions on the modern input use was modeled as50: ])([ 01211 1 WLLLEUMax L +−∏+∏          (4.1) where  1L  =land planted with crops that are applied with modern input  111 )( εμ Xh+=∏  is the net return per ha having a variance of 212 )( σXh  222 εμ +=∏  is the net return per ha having a variance of  22σ  X is the modern input, such as chemical fertilizer or pesticide 0(.),0(.) ''' <> UU , indicating risk-aversion As shown in Zilberman (2006), which has the same essence of the model solution in Just and Zilberman (1983), the first-order condition generates: L XhXh XhXh WRXhXh L A 12 2 2 2 1 2 12 2 2 2 12 2 2 2 1 2 21* 1 )(2)( ])()([ )])(2)([ σσσ σσ σσσ μμ −+ −+−+ −= (   (4.2) where  )WRA( is the Arrow-Pratt measure of absolute risk aversion  )12110 ( LLLWW −++= μμ  is the mean wealth As shown in Zilberman (2006) or Just and Zilberman (1983), differentiation of the equation 4.2 with respect to farmers’ total land size, L , generates: 12 2 2 2 1 2 12 2 2 2 12 2 2 2 1 2 211 )(2)( ])()([1... )( ) . )])(2)([ σσσ σσ σσσ μμ XhXh XhXh WLd Wd R W Wd WdR WRXhXhLd dL A A A −+ −+−+ −−= ( (     (4.3)    50 I also refer to Zilberman’s lecture note in 2006 in addition to Just and Zilberman (1983) to present the model.  145  It is noteworthy that ][)(2)( 2112 2 2 2 1 2 ∏−∏=−+ VarXhXh σσσ in the equation 4.2 and 4.3 represents the degree of risk, which is a function of modern input.  Equation 4.3 indi- cates that riskiness of modern input as well as farmers’ risk aversion could play roles in farmers’ land allocation decisions on the modern input. Just and Zilberman (1983) further showed in their paper whether or not the farm size, L , has the effect on farmer’s rate of modern input use depends on: [1] the effect of mod- ern input on the degree of riskiness, i.e. whether the modern input is risk increasing, risk decreasing or has effect on risk; or [2] the characteristics of farmers’ relative (or absolute) risk aversion.  The above analysis of Just and Zilberman’s (1983) was based on the as- sumption that the risk was the only consideration that farmers took when making deci- sions on the rate of modern input use. Knight et al. (2003), however, consider a situation where farmers have to make tra- deoff between the income (Y) and security (smaller variance of income), denoted as S, to maximize the utility of profits under conditions of risk.   Therefore, they added the secu- rity argument, which was defined as the inverse of the variance in the income. Similarly, their argument could be applied to our specific sample collected in rural Yunnan, where most of farmers’ household income was mainly from agricultural income and was used for home consumption.  As shown in the previous Table 4-1, in the specific setting in our sampled area, farmers’ were cultivating on small farmland size and their family income was from agricultural production 1.  It is highly possible that even though farmers might know that there is a risk in investment because using chemical fertilizer,  146  which incurs cost but does not necessarily lead to an increase in crop yields, could result in a net loss of investment, they still decided to use chemical fertilizer due to their “per- ceivable risk caused to their food security or livelihood” without using chemical fertilizer. Accordingly, such risk perception held by farmers could have significant impact on their decisions on the intensity of modern input use on their lands.  Although we did not spe- cifically measure farmers’ perceptions of “risk caused to their food security or livelih- ood”, we believed their perception could potentially affect their on-farm investment deci- sions through its interaction with farmers’ risk preference. Given the above specific setting of our sample, farmers’ utility model in our specific sample could reasonably include two arguments: [1] net profit; [2] stability of their in- come to secure their livelihood.  Although higher profit resulting from more intensive modern input use, especially chemical fertilizer, could be associated with higher variance (more risk), it also could largely contribute to help secure their livelihood.   Therefore, we can model the security term (S) in our case as an inverse of the variance in income.  The following structural model is the extension of the models of Just and Zilberman (1983) with the incorporation of insights from Knight et al. (2003) to fit our specific setting: } ][ )1(])([{ 21 01211 1 ∏−∏ −++−∏+∏ Var WLLLEUMax L αα    (4.4) where α  is the weight that farmers assign to the profit associated with risk  1-α is the weight assigned by farmers to security of their food and livelihood 12 2 2 2 1 2 21 )(2)(][ σσσ XhXhVar −+=∏−∏  147  We do not provide further detailed mathematical solution or conduct comparative static on our model, because we mainly intend to use the model in (4.4) for explorative purpose to justify our empirical model specifications.  With similar spirits to the model solution of Just and Zilberman (1983), the model in (4.4) includes the security (S) argu- ment in addition to those presented in the model of Just and Zilberman (1983). In our empirical analysis, besides farmland size and farmers’ risk preference and wealth level, we also included farmers’ demographic information as well as some va- riables related to external factors (such as farmers’ family’s social connections and its participation in social networks) that could affect farmers’ capacity to pool resources to deal with random income shocks .  Our justifications are as follows: while the model of Just and Zilberman (1983) mainly focused on the role of farmers’ risk attitude and showed the crucial role of the marginal risk effects of modern inputs in determining far- mers’ decisions on the intensity of input use, Eswaran and Kotwal (1990) recognized that given identical risk preference, farmers would have different levels of capacity to bear risk due to external factors, such as credit constraints.  Indeed, some (Knight et al. 2003) considered human capital, like education, could play important role in determining far- mers’ adoption of modern inputs. We use 4 different types of model specifications by including different sets of inde- pendent variables in the main regression models in our analysis.  Model 1 only included the family’s demographic and socio-economic variables; model 2 included additional in- formation on family’s social connections and its participation in social networks within its village to model 1; model 3 included additional experimental measure of farmer’s risk  148  preference and its interaction with the family’s wealth level to model 1; model 4 included a complete set of independent variables used in previous 3 models.  In all 4 models, we controlled for heterogeneous variations across villages by including village dummies. 4.5. Results and Discussions 4.5.1. Descriptive Evidence Before formally linking our survey data to the experimental data in econometric analysis, we briefly investigated correlations among some key variables of interest contained in survey and experimental data sets, specifically including the intensity of chemical ferti- lizer use or pesticide use on farmland, total area of farmland that were operated by the surveyed farmers’ families, and the number of safe choices that the farmers chosen when they participated in the risk experiments (Table 4-4).  Table 4- 4: Correlations among Key Variables of Interest Variables Unit chemical fertilizer cost (yuan/ha) Unit chemical fertilizer cost (yuan/ha) Total area of farmland (ha/family) Unit chemical fertilizer cost (yuan/ha) 1 Unit chemical fertilizer cost (yuan/ha) 0.5745 1 (0.0000) Total area of farmland (ha/family) -0.2945 -0.1999 1 (0.0000) (0.0006) Number of safe choices 0.0150 0.1192 -0.0647 (0.7994) (0.0430) (0.2686) Note: the significance of the correlation is in parentheses.    149  We found significant and strong correlations between the size of farmland and the in- tensity of chemical fertilizer and pesticide use per unit of land but significant and positive correlations among the intensity of input use, chemical fertilizer, pesticide and labor  as shown in the Table 4-4. The experimental measure of family heads’ risk preference, measured by the number of safe choices chosen by them in the risk experiments, seemed to be positively and sig- nificantly correlated with the intensity of pesticide use on farmland, which indicates that families with more risk-averse heads could be very probably using a significant higher amount of pesticide on their farmland.   We found a positive but insignificant correlation between the family heads’ risk aversion level and the intensity of chemical fertilizer use. 4.5.2. Determinants of Household Investment Decisions on Input Use Table 4-5 and Table 4-6 report results for household investment decisions on chemical fertilizer use and pesticide use, respectively, for all subjects, including consistent ones and those switching back and forth between two options, involved in the risk experiments.   150   Table 4- 5: Household Decisions on Chemical Fertilizer Use (Consistent Subjects) Dependent variable: intensity of chemical fertilizer use   (1) (2) (3) (4) Family size -0.0177 -0.0212 -0.0126 -0.0156 (0.0235) (0.0237) (0.0238) (0.0239) Family head's age -0.0003 0.0002 -0.0011 -0.0004 (0.0033) (0.0034) (0.0033) (0.0033) Family head's education level -0.0019 -0.0021 -0.0021 -0.0006 (0.0110) (0.0119) (0.0110) (0.0111) Total farmland size -0.0351 -0.0347 -0.0349 -0.0346 (0.0069)*** (0.0068)*** (0.0068)*** (0.0068)*** Total asset 3.03E-07 2.17E-07 9.47E-07 6.30E-07 (7.32E-07) (7.30E-07) (1.88E-06) (1.88E-06) Number of safe choices 0.0289 0.02805 (0.0170)* (0.0170)* Interaction of number of safe choices with wealth -1.16E-07 -7.70E-08 (3.01E-07) (3.03E-07) Family belonging to big family name 0.0154 0.0034 (0.0735) (0.0739) Family belonging to big ethnic groups 0.0396 0.0407 (0.1304) (0.1293) Participation in village-wide so- cial networks 0.1976 0.1994 (0.1043) (0.1039)* Family has CCP members -0.0293 -0.0338   (0.0906)   (0.0910) Village dummies YES YES YES YES Constant 5.2132 5.1289 5.0915 4.9931 (0.2625)*** (0.2985)*** (0.2147)*** (0.2953)*** R-squared 0.4218 0.4307 0.4302 0.4391 Observations 238 238 238 238 Note: (1) Standard errors are in parentheses.            (2) ***, ** and * represent significance level of 1%, 5% and 10%, respectively.    151   Table 4- 6: Household Decisions on Pesticide Use (Consistent Subjects) Dependent variable: intensity of pesticide use   (1) (2) (3) (4) Family size -0.0317 -0.025 -0.0313 -0.0238 (0.0315) (0.0317) (0.0313) (0.0316) Family head's age -0.0075 -0.0069 -0.0088 -0.0082 (0.0045)* (0.0045) (0.0045)** (0.0044)* Family head's education level 0.0051 0.0056 0.0028 0.0070 (0.0147) (0.0158) (0.0146) (0.0148) Total farmland size -0.0477 -0.0460 -0.0470 -0.0456 (0.0092)*** (0.0091)*** (0.0091)*** (0.0090)*** Total asset -6.39E-07 -3.89E-07 3.23E-06 2.60E-06 (9.78E-07) (9.69E-07) (2.50E-06) (2.49E-06) Number of safe choices 0.04911 0.0460 (0.0227)** (0.0225)** Interaction of number of safe choices with wealth -6.76E-07 -5.34E-07 (4.01E-07)* (4.01E-07) Family belonging to big family names 0.0063 -0.0065 (0.0975) (0.0979) Family belonging to big ethnic groups 0.1890 0.1939 (0.1730) (0.1713) Participation in village-wide social networks 0.1150 0.1139 (0.1388) (0.1377) Family has CCP members -0.3050 -0.2923   (0.1201)**   (0.1206)** Village dummies YES YES YES YES Constant 4.2314 4.0363 4.1040 3.8917 (0.3509)*** (0.3960)*** (0.3584)*** (0.3911)*** R-squared 0.5011 0.5161 0.5112 0.5250 Observations 238 238 238 238 Note: (1) Standard errors are in parentheses.            (2) ***, ** and * represent significance level of 1%, 5% and 10%, respectively.   152  Family head’s age is negatively correlated both with the intensity of chemical fertiliz- er use and pesticide use as shown in Table 4-5 and Table 4-6.  While it was never signifi- cant in predicting the household decision on fertilizer use,  it appears to be significant, especially when the experimental measure of risk preference was included in regression models as presented in columns (3) and (4) in the Table 4-6.  The significance of the age in the presence of risk aversion measure implies that a family’s head could be interacting with risk aversion term to affect the household decision on pesticide use.  Nonetheless, the robustness of the effect of the family head’s age still needs to be further checked. Farmland size significantly predicted household decisions on the intensity of chemi- cal fertilizer and pesticide applications on farmland.  It is negatively and significantly correlated with the intensity of chemical fertilizer use (Table 4-5) and the intensity of pesticide use (Table 4-6), indicating that households having smaller farmland size were both applying chemical fertilizer and pesticide on their farmland more intensively in sig- nificant terms than those having larger size of farmland.  Both the significance level and the sign of the farmland size in determining household decisions were robust to model specifications as shown by the Table 4-5 and 4-6. Our finding of the significant and negative effect of farmland size on household deci- sions on chemical fertilizer use and pesticide use is interesting and justifiable, given spe- cific settings of Yunnan Province.  As shown in the Table 4-1 and discussed in previous sections, farmers’ families in our sample were cultivating very small-size farmland, while their family income was mainly from on-farm activities.  Consequently, they could have strong inclinations to stabilize crop yields on their farmland by applying pesticide to re-  153  duce potential damage caused by pests to their crops and chemical fertilizers to increase their crop yield.  Naturally, families with smaller size of farmland were inclined to use the pesticide chemical fertilizer more intensively to improve the productivity to ensure sufficient production to satisfy their subsistence needs. Family wealth, approximated by values of total asset that a family was holding in 2006, has insignificant effect on the household decisions on the intensity of chemical fer- tilizer use and pesticide use in all model specifications, including models with risk- aversion term and without risk-aversion terms, as shown in the Table 4-5 and the Table 4- 6.  The positive sign of the family wealth in regression models for chemical fertilizer use is reasonable since chemical fertilizer application often accounts for a relatively good portion of total investment made by a farmer’s family on its farmland51.   Naturally, a wealthier family was able to purchase chemical fertilizer to be applied on its farmland to increase its crop yields. The number of safe choices is positively and significantly correlated with the house- hold decision on the intensity of chemical fertilizer use and pesticide use as shown in the Table 4-5 and the Table 4-6.   The positive and significant coefficient of the number of safe choices chosen by the family head in the risk experiments indicates that compared with a family with a less risk-averse family head, a family having a more risk-averse fam- ily head was applying a significantly higher amount of pesticide and chemical fertilizer on per ha of its farmland possibly with the main reason to stabilize the crop yields.  51 Indeed, in our data set, fertilizer cost accounted for 17% of a family’s total investment, including cost of labors (including home labors and hired labors), fertilizers, seeds, pesticide and machineries.  154  Regarding the relation between the risk-aversion and chemical fertilizer use, it could be negative, given that chemical fertilizer could be a risk-increasing investment decision (Just and Zilberman 1983) and thus risk-averse farmers tend to use chemical fertilizer less intensively; it could be also positive because with uncertainty in agricultural produc- tion, risk-averse farmers would be driven to apply more chemical fertilizers on their land (Isik and Khanna 2003).  Our finding of a positive and significant correlation between the number of safe choices and the intensity of chemical fertilizer use could be explained by the fact that in a rural Yunnan, where farmers’ households’ income were highly depen- dent on agricultural production, farmers’ families were willing to pay for a high risk pre- mium by applying chemical fertilizer more intensively to their farmland. The interaction term of the family head’s risk preference and family wealth indicates is negative but insignificant for the decision on the intensity of chemical fertilizer use.  It is also negatively correlated with the intensity of pesticide use but the significance level of this term is consistent across model specifications. One potentially interesting finding in our paper is that a family’s social connections or social networks could have potentially significant impact on its decision on pesticide use but not necessarily on chemical fertilizer use decision.  We found a negative and sig- nificant correlation between the household decision on the pesticide use and a dummy variable indicating whether or not a family had its members who had participated in Chi- nese Community Party.  If a family had a member belonging to Chinese Community Par- ty, it used a significantly less amount of pesticide on per ha of its farmland while it did not use a significantly less amount of chemical fertilizer on per ha of farmland.  This  155  could be largely true in Yunnan Province, where the credit market and the crop insurance market are still very underdeveloped.  Having a family member belonging to a Chinese Community Party often indicates that a family has some access to potential political net- works beyond the village.  Therefore, when faced with negative shocks such as pest out- breaks, the family could have more resources to be pooled from outside networks to deal with the shocks.  As a result, the family could be investing less on pesticide.  Or probably, a family having a member belonging to Chinese Community Party, which is the most im- portant grassroots institutions to promote new government policies and disseminate in- formation to village members, could mean that a family get more access to information on the hazards of pesticide on human health and as a result it used much less pesticide on its farmland.  In spite of above speculations, if more conclusive evidence is to be sought regarding the effect of a family’s social connection on its pesticide use decision, more specific and rigorous research needs to be conducted .  Our finding only sheds some light on the potential role in this regard. No strong evidence was found on the significant effect of a family’s social connec- tions, such as whether or not a family belongs to a big family name or a big ethnic group in the village, and participation in village-wide social networks on household decisions on chemical fertilizer use or pesticide use.  The insignificance of these relevant variables might be due to: [1] we only collected some brief information rather than detailed infor- mation on family’s participation in networks and social connections; or [2] the networks within the villages might not have been effective in helping a family to deal with risk.  156  4.5.3. Robustness Checks We checked the robustness of our estimation from 3 different aspects to check the ro- bustness of our estimates to model specifications by including some additional variables. We first did some robustness check by: [1] including the interaction term of the farm- land size and the risk variable to both equations of the chemical fertilizer use and pesti- cide use; [2] only including the interaction term of the family head’s age and the risk va- riable to the pesticide use equation.  Our justifications of adding the interaction term of the farmland size and the risk variable to both equations are: [1] our previous models all indicated that the farmland size were significantly correlated with the decisions on both of the input use and the number of safe choices seemed to be highly correlated with deci- sions both input use; [2] the theoretical model of Just and Zilberman (1983) shows that farm size could affect farmers’ decisions on “risky” input use through its interaction with farmers’ attitudes toward risk.  Our inclusion of the interaction term of age and the risk variable in the equation of pesticide use decision was also based on the significant corre- lation between the age and the input use as shown in the previous models. We found that the 2 interaction terms could neither significantly predict household decision on chemical fertilizer use nor its pesticide use decision.  The significance level and sign of other variables remain the same as those presented in the Table 4-5 and Table 4-6.  Since adding the 2 interactions did not significantly change the results in the Table 4-5 and Table 4-6, for parsimony, we did not report the detailed information on the re- sults.  157  4.6. Conclusions In this chapter, we combined household survey data and experimental data collected in Yunnan province, China to try and uncover both a farmer’s personal and household cha- racteristics, as well as their social connections. Taken together the data could explain their ability to absorb risk and the risk behavior found in their on-farm decisions. First, we found that farmland size is significantly and negatively correlated with household decisions both on the pesticide use and chemical fertilizer use, other things being equal.  This indicates that with limited farmland to support their family main in- come, farmers in the Province of Yunnan tend to use pesticide and chemical fertilizer on their farmland intensively. Second, we found some evidence for the potential role played by social capital in de- termining household on-farm investment decisions when faced with risk.  We found that when a family has CCP members, it appears to use pesticide significantly less intensively on its farmland.   If a family has members who have participated in village branch of Chinese Communist Party (CCP), one of the most important social networks at grassroots level, it potentially has the resources to be pooled to deal with unexpected shocks, such as reductions in crop yields due to unexpected outbreaks of pests.   To some degree, our finding is closely related to the concept of social capital, which includes bonding social capital, defined as networks within groups that bring members closer together, and bridg- ing social capital, referring to the networks across groups that enable members to reach outside sources of information, support and resources (Putnam 2000; Narayan 1998).  158  Third, and importantly we found that: [1] the degree of a family head’s risk aversion level can significantly and positively predict the household decision on the intensity of chemical fertilizer use on the family’s farmland; and [2] it is positively correlated with the household decision on pesticide use.  However, while the significance level of the correlation between the risk variable and the household decision on chemical fertilizer use is robust to all model specifications, the significant level of correlation between the risk variable and the household decision on the intensity of pesticide use is not robust in all model specifications.  Therefore, we must be cautious in explaining the power of a farmers’ risk aversion in predicting the household decision on pesticide use. .Our findings support the notion that individual behavior in our artefactual field expe- riment seemed “generalizable” to behavior in the field, indicating that the experimental results could be informative about the real-world behavior. Huang et al. (2008) attributes the overuse of chemical fertilizer to insufficient train- ings provided to farmers regarding using fertilizers more efficiently and Huang et al. (2003) maintains that a Chinese farmers’ intensive use of pesticide results from their per- ceptions of yield loss.  Based on these and our findings we suggest that if China attempts to address the current problem of farmers’ overuse of chemical fertilizer and pesticide, the government policies could be formulated to: [1]  change the farmers’ risk preference over time; [2] reduce a farmers’ external constraints, such as constraints on credit and income sources, to mitigate the constraint of their small farmland size; or [3] provide bet- ter training, information and extension services to change farmers’ risk perception and improving their knowledge of using the modern inputs.  159  Bibliography  Antle, J.M. (1987).  Econometric estimation of producers’ risk attitudes.  American Jour- nal of Agricultural Economics, 69(3), 509-522.  Antle, J.M. (1988).  Nonstructural risk estimation.  American Journal of Agricultural Economics, 71 (3), 774-784. Arrow, K.J. (1971).  Essays in the theory of risk bearing.  Amsterdam: North Holland. Babcock, B.A. & Shogren, J.F. (1995).  The cost of agricultural production risk.  Agricul- tural Economics, 12(2), 141-150. Binswanger, H.P. (1980).  Attitudes toward risk: experimental measurement in rural India. American Journal of Agricultural Economics, 62 (3), 395-407. Binswanger, H.P., & Sillers, D.A. (1983).  Risk aversion and credit constraints in farmers’ decision-making: a reinterpretation.  The Journal of Development Studies, 20 (1), 5- 21. Carpenter, J.P. (2002).   Measuring social capital: adding field experimental methods to the analytical toolbox.  In E. Elgar (Eds.) Social capital, economic development and the environment.  Sunder Ramaswamy: Jonathan Isham and Thomas Kelly. Coady, D. P. (1993).  An empirical analysis of fertilizer use in Pakistan.  Economica, 62, 213-234.  160  Croppenstedt, A., Demeke, M., & Meschi, M.M. (2003).  Technology adoption in the presence of constraints: the case of fertilizer demand in Ethiopia.  Review of Devel- opment Economics, 7 (1), 58-70. Eswaran, M., & Kotwal, A. (1990),  Implications of credit constraints for risk behavior in less Developed economies.  Oxford Economic Papers, 42(2), 473-482. Feder, G. (1980).  Farm Size, Risk Aversion and the Adoption of New Technology under Uncertainty.  Oxford Economic Papers, 32, 263-283. Feder, G., Just, R.E., & Zilberman, D. (1985).  Adoption of agricultural innovations in developing countries: a survey.  Economic Development and Cultural Change, 33 (2), 255-298. Harrison, G., List, J.A., & Towe, C. (2007).  Naturally occurring preferences and ex- ogenous laboratory experiments: a case study of risk aversion.  Econometrica, 75 (2), 433-458. Hill, R.V. (2009).  Using stated preferences and beliefs to identify the impact of risk on poor households.  The Journal of Development Studies, 45(2), 151-171. Holt, C., & Laury, S. (2002).  Risk aversion and incentive effects.  American Economic Review, 925, 1644-1655. Hu, R., Cao, J., Huang, J., Peng, S., Huang, J., Zhong, X., Zou, Y., Yang, J., & Beresh, R.J. (2007).  Farmers participatory testing of standard and modified site-specific ni- trogen management for irrigated rice in China.  Agricultural Systems, 94, 331-340.  161  Huang, J., Rozelle, & Lu, F. (2003).  Farm pesticide use, rice production and human health.  In T.W. Mew, D.S. Brar, S. Peng, D. Dawe & B. Hardy (Eds.), Rice Science, innovations and impact for livelihood (pp. 901-918).  International Rice Research In- stitute. Huang, J., Hu, R., Cao, J., & Rozelle, S. (2008).  Training programs and in-the-field guidance to reduce China’s overuse of fertilizer without hurting profitability.  Jour- nal of Soil and Water Conservation, 63 (5), 165A-167A. Isik, M., & Khanna, M. (2003).  Stochastic technology, risk preferences, and adoption of site-specific technologies.   American Journal of Agricultural Economics, 85(2), 305-317. Just, R.E., & Pope, R. (1978).  Stochastic specification of production functions and eco- nomic implications.  Journal of Econometrics, 7, 67-86. Just, R.E., & Zilberman, D. (1983).  Stochastic structure, farm size, and technology adop- tion in developing agriculture.  Oxford Economic Papers, 35 (2), 307-328. Knight, J., Weir, S., & Woldehanna, T. (2003).  The role of education in facilitating risk- taking and innovation in agriculture.  The Journal of Development Studies, 39 (6), 1- 22. Liu, E.M. (2008). Time to change what to sow: risk preferences and technology adoption decisions of cotton farmers in China.  Working Papers 1064, Department of Eco- nomics, Princeton University.  162  Lusk, J.L., Pruitt, J.R., & Norwood B. (2006).  External validity of a framed field expe- riment.  Economics Letter, 93, 285-290. Masson, R.T. (1972). The creation of risk aversion by imperfect capital markets.  Ameri- can Economic Review, 62 (1), 77-86. Moscardi, E., & de Janvry, A. (1977).  Attitudes towards risk among peasants: an econo- metric approach.  American Journal of Agricultural Economics, 59(4), 710-716. Pope, R., & Kramer, R. (1979).  Production uncertainty and factor demands for the com- petitive firm. Southern Economic Journal, 46, 489-501. Pratt, J. (1964).  Risk aversion in the small and in the large, Econometrica, 32 (1-2), 122- 136. Putnam, R. D. (2000).  Bowling alone: the collapse and revival of American community. New York: Simon & Schuster. Schechter, L. (2007). Traditional trust measurement and the risk confound: An experi- ment in rural Paraguay. Journal of Economic Behavior and Organization, 62 (2), 272-292. Welch, F. (1978).  The role of investment in human capital in agriculture.  In T.W. Schultz (Eds.), Distortion of agricultural incentives.  Bloomington: Indiana Univer- sity Press. Wik, M., Kebede, T.A., Bergland, O., & Holden, S.T. (2004).  On the measurement of risk aversion from experimental data.  Applied Economics, 36, 2443-2451.  163  Young, D.L. (1979).  Risk preferences of agricultural producers: their use in extension and research.  American Journal of Agricultural Economics, 61 (5), 1063-1070. Zilberman, D. (2006).  Lecture 5 of ARE 241: economic analysis of behaviors under un- certainty.  Retrieved on February 10, 2010, from http://are.berkeley.edu/~zilber/ARE241/fall2006/Risk.pdf    164    Chapter 5  Conclusions   165  This dissertation examines the role of trust and risk preferences in Chinese farmers’ investment de- cisions on their lands.  The dissertation has three primary objectives: [1] To analyze the roles of property rights, contractual rules and social capital in farmers’ decision to participate in a CDM for- est project; [2] To measure and analyze the determinants of rural villagers’ mutual trust. [3] To ana- lyze the effect of risk preference on a farmer’s on-farm investment decisions. Key contextual elements for this dissertation were that: [1] afforestation is increasing in impor- tance to the government of China for its climate change mitigation. It is being conducted under a specific land tenure system in China, where, in many cases, land use decisions are still made on community level. This implies that farmers’ are collective in their decision-making; therefore the role of social capital and institutions is important.  [2] although trust has been found to play an im- portant role in defining economic success at both the macro-levels (Knack and Keefer 1997; La Porta et al. 1997; Zak and Knack 2001) and micro-level (Guiso et al. 2008; Karlan 2005), the mea- surement of trust and its determinants have empirical challenges; [3] risk preferences affect a far- mer’s decisions to invest in agricultural production or forestry activities.  In the past, although the experimental method was used to measure a farmer’s risk preferences (Bingswanger 1980; Grisley 1980; Hill 2009; Knight et al. 2003; Liu 2008; Schechter 2003), the external validity of the experi- mental method was under debate (Carpenter 2002; Lusk et al. 2006); therefore further empirical evidence is still required.  In addition, overuse of chemical fertilizer and pesticide has become an environmental problem in China (Huang et al., 2008), but little empirical research has been con- ducted in China to analyze the role of farmers’ risk preference on their input use on their farmland. In this context, the dissertation is both empirically importance and policy relevancy for rural devel- opment.  166  In the second chapter, I asked: What factors affect a farmer’s participation in the world’s first CDM forest project, established in Guangxi Province, China?   I used a Coasean framework to ar- gue that efficient participation requires: [1] the combination of social surpluses evoked by a change in behavior has to be sufficiently large; and, [2] the bargaining costs have to be sufficiently low, so that the residual surplus after bargaining is still positive (or at least not negative).   Regarding the second argument, I postulated that social capital could interact with property rights and contractual rules to affect the bargaining cost, which in turn can affect a farmer’s participation decision on CDM forest project.  Based on both village-level field data and personal interview evidence, I found that the project facilitated farmers’ participation through innovative arrangements in carbon pooling and a design of share-holding system, both of which are intended to address high risk and high transaction cost that could prevent farmers’ from participating.   I also found that in spite of these two innovative arrangements, a high proportion of the project land remains unforested mainly due to constrained contractual rules, property rights allocation disputes and low levels of social cap- ital in some villages. In the third chapter, I focused on analyzing the determinants of rural villagers’ mutual trust, based on a unique data set collected through survey and field experiment methods, with which household surveys and trust games were conducted in 30 administrative villages in Yunnan Prov- ince in southwestern China.  I found evidence that: [1] the positive individual social interactions and good historic relations could have significant and positive effect on the villagers’ mutual trust; [2] the formal village institutions might be gradually substituting for the traditional informal institu- tions, and that the new formal institutions had not effectively taken on the roles of informal village institutions in maintaining villagers’ mutual trust; [3] the openness of the villages to the outside world and the market could have eroded mutual trust among rural villagers Yunnan Province, the   167 most ethnically diverse province in Western China.   I also found that when designed properly, the survey and experimental measures could be strongly correlated. In fourth chapter, I focused on using the artefactual field experimental method to measure far- mers’ risk preference and linking it to real-world decisions on the input choice for their farmlands, with the aim of better understanding whether or not the experimental measure of a farmers’ risk preference can strongly predict the risky behavior revealed in their real-world decisions.   I also in- tended to reveal whether or not a farmer’s personal and household characteristics, as well as their social connections, can predict their behavior in on-farm decisions.   I find that risk preference and farmland size both affect a farmers’ decisions.  There are both empirical and policy implications. The empirical implication is that field evidence of behavior can be informative in real-world deci- sions. The dissertation makes the following contributions: [1] it is one of the few research projects to demonstrate that people’s interaction and people’s trust are highly correlated; [2] it is one of the few research projects to demonstrate that formal institutions are negatively correlated with trust; [3] it used both survey and experimental methods to identify the determinants of the farmer’s trust in an ethnically diverse area; [4] it is one of very few papers to use a risk game to get a measure of risk preferences and relate them to a farmer’s investment decisions. The dissertation, naturally, has some limitations.  First, although the dissertation identified the roles of property rights, contractual rules and trust in a farmers’ participation in CDM forest project, the findings were made based on village-level survey data and personal interviews.  The causal rela- tionship between trust (property rights or contractual rules) and farmers’ land use investment deci- sions, however, requires further rigorous econometric analysis.  Second, in the third chapter, despite the strong evidence regarding the strong and negative correlation between formal institutions and   168 rural villagers’ mutual trust, the relationship between villagers’ mutual trust and the openness of the villages to outside world or market is not conclusive.  Therefore, more systematic and rigorous ana- lyses still need to be conducted.   Third, the econometric analysis employed in the fourth chapter does not aim to address the “endogeneity” problems.  Rather, the focus was on analyzing the corre- lation between the experimental measure of risk preference and the risk preference revealed by far- mers in their real-world investment decisions.   Therefore, it is possible that the endogeneity did arise in the econometric model.  Similar problems could also exist in the third chapter. In its totality, the dissertation is only a first step in: integrating social capital into the institution- al analysis, understanding the role of collective action and applying experimental method to re- source management problems in rural development in China.  Despite the limitations, preparing this dissertation has given me the opportunity to learn creative, practical and original empirical research techniques in the field of resource and developmental economics.   169  Bibliography Binswanger, H.P. (1980).  Attitudes toward risk: experimental measurement in rural India.  Ameri- can Journal of Agricultural Economics, 62 (3), 395-407. Carpenter, J.P. (2002).   Measuring social capital: adding field experimental methods to the analyti- cal toolbox.  In E. Elgar (Eds.)  Social capital, economic development and the environment. Sunder Ramaswamy: Jonathan Isham and Thomas Kelly. Fehr, E. (2009).  On the economics and biology of trust.  Journal of European Economic Associa- tion, 7 (2-3), 235-266. Geanakoplos, J., Pearce, D., & Stacchetti, E. (1989). Psychological games and sequential rationality. Games and Economic Behavior, 1, 60–79. Grisley, W., & Kellog, E. (1987).  Risk-taking preferences of farmers in Northern Thailand: mea- surements and implications.  Agricultural Economics, 1, 127-142. Guiso, L., Sapienza, P., & Zingales, L. (2008).  Trusting the stock market.  Journal of Finance, 63 (6), 2557-2600. Hill, R.V. (2009).  Using stated preferences and beliefs to identify the impact of risk on poor households.  The Journal of Development Studies, 45 (2), 151-171. Huang, J., Hu, R., Cao, J., & Rozelle, S. (2008).  Training programs and in-the-field guidance to reduce China’s overuse of fertilizer without hurting profitability.  Journal of Soil and Water Conservation, 63 (5), 165A-167A. Karlan, D. S. (2005). Using experimental economics to measure social capital and predict financial decisions.  American Economic Review, 95 (5), 1688-1699.  170  Knack, S., & Keefer, P. (1997). Does social capital have an economic payoff? A cross-country in- vestigation. The Quarterly Journal of Economics, 112 (4), 1251-1288. Knight, J., Weir, S. & Woldehanna, T. (2003).  The role of education in facilitating risk-taking and innovation in agriculture.  The Journal of Development Studies, 39 (6), 1-22. La Porta, R.  Lopez-de-Silane, F., Shleifer, A., Vishny, R.W. (1997).  Trust in large organizations. American Economic Review, 87 (2), 333-338. Liu, E.M. (2008). Time to change what to sow: risk preferences and technology adoption decisions of cotton farmers in China.  Working Papers 1064, Department of Economics, Princeton Uni- versity. Lusk, J.L., Pruitt, J.R., & Norwood B. (2006).  External validity of a framed field experiment.  Eco- nomics Letter, 93, 285-290. Schechter, L. (2007). Traditional trust measurement and the risk confound: An experiment in rural Paraguay. Journal of Economic Behavior and Organization, 62 (2), 272-292. Zak, P.J., & Knack, S. (2001).  Trust and growth.  Economic Journal, 111(470), 295-321.   171       Appendix 1 - Survey52       52 The survey in appendix was used by the research team led by Professor Jintao Xu at Environmental Economics Program in China (EEPC) at Peking University in China to collect data in 8 provinces in China.  This dissertation only used the data collected by the research from Yunnan Province.  172     云南省林权改革调查问卷(农户表) 省: 市(地区) 、 县: 乡(镇): 村: 农户姓名: 民族: 电话号码:  173    174  A. 家庭基本情况 表 A1  家庭基本特征 (2006 年) 0.1 你家现在有几口人?          人(见注释)  表 A.2 . 家庭人口状况(2006) 编 码 01 与户主的关 系 02 性别 03 户口类型 04 年龄 05 上过几年学 06 是否党员 07 是否是村干部? 08.是否在林业部 门工作? 09 是否常年在家? 代码 1=男;2=女 1=农; 2=非农; 3=没户口 周岁 年 1=是;2=否 1=是;2=否 1=是;2=否 1=是;2=否 1 2 3 4 5 6 7 8 9 10 11 12 13 14 与户主关系代码:1=户主;2=配偶;3=孩子;4=孙辈;5=父母;6=兄弟姐妹;7=女婿,儿媳,姐夫,嫂子;8=公婆,岳父母;9=亲戚;10=无亲戚关系   175   受访者姓名                               与户主关系                               性别  表 A.3  .家庭人口状况(2000) 编 码 该成员是增 还是减的 变化原因 01 与户主的 关系 02 性别 03 户口类型 04 年龄 05 上过几年学 06 是否党员 07 是否是村干 部? 08.是否在林业 部门工作? 09 是否常年在 家? 1=增 2=减 3=没变 代码 代码 1=男;2=女 1=农; 2=非农; 3=没户口 周岁 年 1=是;2=否 1=是;2=否 1=是;2=否 1=是;2=否 1 2 3 4 5 6 7 8 9 10 11 12 13 14 注:人口变化的增加与实际变化情况相符 与户主关系代码:1=户主;2=配偶;3=孩子;4=孙辈;5=父母;6=兄弟姐妹;7=女婿,儿媳,姐夫,嫂子;8=公婆,岳父母;9=亲戚;10=无亲戚关系 变化编码:1=出生;2=死亡;3=迁移;4=婚嫁;5=就学或返家;6=服役或复员;7=丧偶或孤儿;8=分家;9=其它   176  B 非农就业 B.1.  2006非农工作情况  个  人  编  码  2006 年 是否务 农或从 事非农 工作?    1 是 2 否  没参加的原 因 1 年老 2 身体不好 3 待业 4 只做家务 5.上学 6.其他(请注 明) =>下一人 收入最高的非农工作 收入第二的非农工作 收入第三的非农工作 地 点 工 作 类 别  工 作 性 质  请 说 明 具 体 工 作   全年 干了 多少 天? (毛收入,自办企业则是毛利润)  全年现金 和实物收入?  地 点 工 作 类 别  工 作 性 质  请 说 明 具 体 工 作   全年 干了 多少 天? (毛收入,自办企业则是毛利润)  全年现金 和实物收入?  地 点 工 作 类 别  工 作 性 质  请 说 明 具 体 工 作   全年 干了 多少 天? (毛收入,自办企业则是毛利润)  全年现金 和实物收入?  全年 一共 给家 里多 少 钱?    编码 编 码 编 码  天 元 编 码 编 码 编 码  天 元 编 码 编 码 编 码  天 元 元 1 2 3 4 5 6 7 8 9 10 地点编码: 1=本村, 2=本乡, 3=本县, 4=本省, 5=外省, 6.其他请注明 工作类别编码: 1=工业, 2=服务业, 3=农业生产, 4=工程队, 5=行政, 事业单位, 6=手艺人, 7=其他    工作性质编码:1=拿工资, 2=自己做   177  B.2.  2000 非农工作情况  个  人  编  码  2000 年 是否务 农或从 事非农 工作?    1 是 2 否  没参加的原 因 1 年老 2 身体不好 3 待业 4 只做家务 5.上学 6.其他(请注 明) =>下一人 非农工作一 非农工作二 非农工作三 地 点 工 作 类 别  工 作 性 质  请 说 明 具 体 工 作  全年 干了 多少 天? 全年现金收入和实物收入?  地 点 工 作 类 别  工 作 性 质  请 说 明 具 体 工 作  全年 干了 多少 天? 全年现金收入和实物收入?  地 点 工 作 类 别  工 作 性 质  请 说 明 具 体 工 作  全年 干了 多少 天? (毛收入,自办企业则是毛利润)  全年现金 和实物收入?  全年一 共给家 里多少 钱?    编码 编 码 编 码  天 元 编 码 编 码 编 码  天 元 编 码 编 码 编 码  天 元 元 1 2 3 4 5 6 7 8 9 10 地点编码: 1=本村, 2=本乡, 3=本县, 4=本省, 5=外省,  作类别编码: 1=工业, 2=服务业, 3=农业生产, 4=工程队, 5=行政, 事业单位, 6=手艺人, 7=其他         工作性质编码:1=拿工资, 2=自己做 C 林权改革 情况   178   01. 村里哪一年开始林权改革? 年 02. 在林权改革的过程你家的林地经营权是否发生变化?(包括换发 林权证;林地的增加或减少) 1 有=>03     2 没有=>04 03. 林权改革过程中, 你家是否有自由选择承包地块的权利? 1 是=05         2  否 04. 你家不参加的原因是? (多选) 1 家里没有劳动力; 2 家里主要从事其他经营 3 没有资金去承包林地 4 风险太大; 5 没有技术 6.其他(说明)  05. 在林权改革过程中, 是否有人向你家征求有关林权改革的意见? 1. 是         2. 否 林权改革过程 村民代表大会 户主会 06. 林权改革决策过程——村民代表大会,户主会 a. 关于这次林权改革, 村里开过多少次村民代表大会/户主会? 次 b. 你家有人是村民代表或党员吗? 1. 是村民代表        2. 否 =>07  3. 是党员但不是代表 c. 你家参加过几次? 次 07. 林权改革决策过程——村民大会 a. 关于这次林权改革, 村里开过多少次村民大会? 次 b. 你家参加过几次? 次 08. 林权改革的方案最终由谁通过? 1 村里说了算; 2 村民代表大会投票通过 3 村民大会投票通过; 4 其他(说明) 09. 你对这次林权改革有些看法或建议?   179  D1 地块投入(2006) 样 本 地 块 编 码 2006 年 施了多 少有机 肥 2006 年施了多少化肥?(需注明化肥名称) 打农 药、 杀虫 剂、 除草 剂共 花了 多少 钱 浇 水 花 了 多 少 钱 机 械 服 务支出 自投多 少工  雇工数量 雇 工 日 工 资 畜工费 用 其 他 生 产 资 料 花 了 多 少钱  针对 2006 年新增造林情况 化肥 1: 化肥 2: 化肥 3: 化肥总 支出: 帮 工 换 工 造林树 种(同 林产品 编码) 造林 面积 株数 种苗费  公斤 公斤 公斤 公斤 元 元 元 元 工日 工日 元/工日 元 元 编码 亩 株 元 1  2  3  4  5  6  7  8     180  D3 地块投入(2000) 样 本 地 块 编 码 2000 年 施了多 少有机 肥 2000 年施了多少化肥?(需注明化肥名称) 打农 药、 杀虫 剂、 除草 剂共 花了 多少 钱 浇 水 花 了 多 少 钱 其 他 生 产 资 料 花 了 多 少钱 自投多 少工 雇工数量 (包括帮 工和换 工) 雇 工 日 工 资 畜工费 用 机械服 务支出 针对 2000 年新增造林情况 化肥 1: 化肥 2: 化肥 3: 化肥总 支出: 帮 工 换 工 造林树 种(同 林产品 编码) 造林 面积 株数 种苗费  公斤 公斤 公斤 公斤 元 元 元 元 工日 工日 元/工日 元 元 编码 亩 株 元 1  2  3  4  5  6  7  8    181  D4 地块投入(其他年份) 2000 年到 2006 年之间你家是否有造林或者高于正常年份的林业生产投入行为? 是——1  否——2   >>回答 E 部分 (与前面一致)  样本地块编码  投资年份    针对新增造林情况 当年施 了多少 有机肥 当年施了多少化肥 打农药 花了多 少钱 浇水花 了多少 钱 机械服 务支出 自投工 雇工数 量 雇工日 工资 畜工费 用 其他生 产资料 花了多 少钱 下一年 的投入 是否与 2006 年 相似 造林树 种(同 林产品 编码) 造林面 积 株树 种苗费 化肥 1: 化肥 2: 化肥 3: 化肥总 支出 帮工 编码 亩 株 元 公斤 公斤 公斤 公斤 元 元 元 元 工日 工日 元/工日 元 元 1=是 2= 否〉〉 继续问 下一年 的投入               182  E 林地经营中的合同安排       E1. 合同安排   类型一 类型二 类型三 01.林地经营类型 编码(参照第 5 页)  对应地块编码(与前面一致) 02.你和谁订的合同? 1.村 2.小组 3.其他 03.合同形式 1.口头 2.书面 04.何时定的合同? 年 月 05.是否已经拿到林权证? 1. 是    2. 否=>07 06 何时获得林权证 年 月 07. 如果没拿到林权证, 原因是 1. 还未发放.  2. 没听说过.  3 其他 08.需要抵押吗? 1.需要 =>09 2.不需要 =>10 09.需要多少? 元 10.合同年限? 11.期限如何确定? 1. 固定 =>13; 2. 轮伐期 =>12 ;3 采伐权的年限=>13 12 一个轮伐期约为多少年? 年 13 合同是否需要付费 1 是=>14      2 否=>15 14. a.付费方式   1.每年      2.一次付清  b.支付目标? 1 林地   2. 林木  3. 林地和林木 4.其他(说明)      c.何时付费? 年, 月  d.应付多少 元  e.已付多少 元 15 是否与合同签订方分成 1 是  =>15           2 否=> 表 D2 16       a.分成对象 1 销售净收入  2 销售毛收入   3 木材          b.何时分成 年,月   183  E 2 农户对林地的使用决策情况   类型一 类型二 类型三 01. 林地经营类型 编码(参照第 5 页)  对应地块编码(与前面一 致) 02. 你有权将林地转作农地吗? 1 可以 2 可以, 但需由村里批准 3 不可以 4 其他  03. 你有权将承包来的林地转种其他林种吗? 如在 用材林上种经济林或竹林,  04. 在不改变林种的情况下, 你有权决定用什么树 种吗?  05. 你有林下资源经营利用的权利吗? 06. 你有权将林地转给本村的其他村民吗? 07. 你有权将林地转给外村的其他村民吗? 08. 你是否能将林地抛荒? 1 是   2 否 09. 如不能抛荒, 原因是什么(请农户详述原因) 注:1. 此处的经营类型应该涵盖在 P5 中所出现过的所有林地经营类型(下表同)   184  2. 如果同种经营类型林地的使用决策情况有一项不不一致都请分开问;如果使用决策情况完全一致则可以合并(下表同)   林权证情况 类型一 类型二 类型三 林地经营类型 编码(参照第 5 页)  对应地块编码(与前面一 致) 01. 你有权将林权证作为抵押吗? 1 可以 2 可以, 但需由村里批准 3 不可以 4 没有林权证=>02 5 其他  02. 如果没有林权证, 你有权将林地或林木作为抵 押吗? 1 可以 2 可以, 但需由村里批准 3 不可以 4 其他  03. 你是否试图通过抵押林地或林木(或林权证) 来获取贷款? 1 是=>04    2 否=>表 E3 04. 你最终获得贷款了吗? 1 是 (贷款情况需在表 W 体 现) 2 否=>05  05. 如果没有获得贷款, 请详述原因    185   )   转入者1 转入者2 转入者3  转出者1 转出者2 转出者3 01.地块编码 注意补问 P7 的地块基本信息 01.地块编码 02.林地经营类型  编码 02.林地经营类型 编码同左 03.需要得到谁的批准? 1.村  2.小组 3.不需要 4.其他 03.需要得到谁的批准? 编码同左 04.与转入土地者的关系 1.亲戚=>06   2.熟人=>05   3.其他=>06 04.与转出土地者的关系 编码同左 05.你认识这个熟人多久了? 年 05.你认识这个人多久了? 编码同左 06.他们住在什么地方? 1.本村 =>07 2.外村 =>08 06.他们在什么地方? 编码同左 07.他是本村的人吗? 1.是 =>09  2.不是 07.他是本村的人吗? 编码同左 08.他们是从哪里来的? 1.本乡外村   2.本县外乡   3.本省外县   4.外省 08.他们是从哪里来的? 编码同左 09.合同形式 1.口头  2.书面 09.合同形式 编码同左 10.是否有担保? 1.有 =>11  2.没有 =>12 10.是否有担保? 编码同左 11.担保人是? 1.亲戚   2.村干部   3.熟人  4.其他 11.担保人是? 编码同左 12.何时定的合同? (年/月) 12.何时定的合同? 编码同左 13.期限是固定的吗? 1.是 =>14  2.否 =>15 13.期限是固定的吗? 编码同左 14.期限是多少年? 年 14.期限是多少年? 编码同左 15你转包地给他已经有多少年了? 年 15你转包他的地已经有多少年 了? 编码同左 16.需要抵押吗? 1.需要 =>17 2.不需要 =>18 16.需要抵押吗? 编码同左 17.需要多少钱抵押? 元 17.需要多少钱抵押? 编码同左 18 是否需要付费 1=是 2=否 18 是否需要付费 编码同左 19. a.第一种付费方式   1.每年 2.一次性 19. a.第一种付费方式 编码同左  b.应付多少?   b.应付多少? 编码同左   E3. 转出地(农户自主行为)(包括转交给村组集体的) E4. 转入地(农户自主行为)(不包括从村组集体获取的)  186  F 木材采伐  F.1.  2006 木材采伐情况(对用材林和竹林) 地块编 码 第一树种 第二树种 树种 采伐量 根数 自投工 雇工 雇工 工资 机械支出 (采伐、 集材、运 输) 采伐设 计费 树种 采伐量 根数 自投工 (包括帮 工、换 工) 雇工 雇工 工资 机械支出 (采伐、 集材、运 输) 采伐设计 费 编码 (用材林) (竹林) 帮工、 换工 编码 (用材林) (竹林) (同林产 品    (同林 产品编 码)  编码) 立方米 根 工日 工日 元/工日 元 元 立方米 根 工日 工日 元/工 日 元 元 1   2   3   4    5       187  F.2.  2000 年木材采伐情况(对用材林和竹林)  地块编 码 第一树种 第二树种 树种 采伐量 根数 自投工 雇工 雇工工 资 机械支出 (采伐、 集材、运 输) 采伐设计 费 树种 采伐量 根数 自投工 (包括帮 工、换 工) 雇工 雇工工 资 机械支出 (采伐、 集材、运 输) 采伐 设计 费 编码 (用材林) (竹林) 帮工、换 工 编码 (用材林) (竹林) (同林产 品    (同林产 品编码) 编码) 立方米 根 工日 工日 元/工日 元 元 立方米 根 工日 工日 元/工日 元 元 1   2   3   4   5    6       188  F 3.  其他年份木材采伐情况(针对用材林和竹林)  2000 年到 2006 年之间是否有商品性(除了自用)采伐    是——1 否——2   >> 回答 G 部分  地 块 编 码 第一树种 第二树种 树种 采伐量 根数 自投工 雇工 雇工工资 机械支出 (采伐、 集材、运 输) 采伐设 计费 树种 采伐量 根数 自投工 (包括帮 工、换 工) 雇工 雇工工资 机械支出 (采伐、 集材、运 输) 采伐设计 费 编码 (用材林) (竹林) 帮工、换 工 编码 (用材林) (竹林) (同林产 品    (同林 产品编 码)  编码) 立方米 根 工日 工日 元/工日 元 元 立方米 根 工日 工日 元/工日 元 元 1   2   3   4   5   6     189    G  木材销售情况 G1 用材林  (编码: 成片出售=1,  单独出售=2, 请在选择项里填写编码) 年 2000 2001 2002 2003 2004 2005 2006 成片 出售   1 出售面积(亩) 单独 出售   2 出售量 (立方米) 选择 (     ) 出售总价款(元) 出售价格 (元/立方米) 出售地点 1.本村公路边, 2. 本乡, 3. 本县, 4. 本省 出售对象来自于 1 本村人, 2 本村企业, 3.本乡人, 4 本乡企业, 5 本县人, 6 本县企业, 7 省内,  8 外省 出售对象是 1 收购木材的企业或个人(单户出售), 2. 收购木 材的企业或个人(联户出售) 3. 与农户有协议的 公司,  4.其他(说明).    (单户和联户指农户)  家里自用 立方米  G2 竹林及薪柴 年 2000 2001 2002 2003 2004 2005 2006 种类 1=竹林   2=薪柴 出售量 根/公斤/百斤(注明单位) 出售价格 元/根,元/公斤, 元/百斤 出售地点 1.本村公路边, 2. 本乡, 3. 本县, 4. 本省 出售对象来自于 1 本村人, 2 本村企业, 3.本乡人, 4 本乡企业, 5 本县人, 6 本县企业, 7 省内,  8 外省  家里自用 根/公斤/百斤(注明单位)    190  H -1 联户经营情况  联户经营的类型 1 强制 2、自愿组合 3、其他(说 明) 01. 参与联户经营的总户数 户 02. 参与联户经营的总人数 人 03. 你们签定协议的形式? 1. 口头     2. 书面 04. 合同期间, 你可以退出吗 1. 可以     2. 不可以 05. 各户之间的投入比例的确定方式 1. 按户数   2. 按人口  3 其他 06. 各户之间的收入分成方式 1. 按户数   2. 按人口  3 其他   H -2 “分股不分山”经营情况    本次林改 最近一次自发林改 分股不分山经营的类型 1 强制 2、自愿组合 3、其他(说 明) 参与股份经营的林地来源 1.自留山 2.责任山 3.集体山林   4.其 他(请说明) 01. 参与股份经营的总户数 户 02. 参与股份经营的总人数 人 03. 你们签定协议的形式? 1. 口头     2. 书面 04. 合同期间, 你可以退出吗 1. 可以     2. 不可以 05. 各户之间的投入比例的确定方式 1. 按户数   2. 按人口  3 其他 06. 各户之间的收入分成方式 1. 按户数   2. 按人口  3 其他   191  I 林业税费(生产性支出)    2006 2000 林业税费种类 征收标准 (填代码)  征收比例 应交金额(元) 实交金额(元) 征收标准 (填代码) 征收比例 应交金额(元) 实交金额(元)     合计  林地税费编码 1 育林费,  2 维检费,  3 采伐设计费,  4 检疫费,  5 检尺费,  6 林价返还款,  7 农林特产税,  8 其他(说明) 征收标准:1 每年按亩收取 2 按采伐根数收取 3 按采伐立方米收取 4 按销售收入收取 5 其他(请说明)   192   J 非自家地块上的其它林副产品利用情况  单位 2006 2000  采集量 销售量 价格(元/公斤) 采集量 销售量 价格(元/公斤) 薪材:(包括枝桠, 剩余物) 立方米/年,公斤/年 (注明单位) 木耳(干量) 公斤 蘑菇(野生食用菌)(干量) 公斤 其他(请说明)  1              2    193   L. 农户对耕地和林地的使用决策情况  L1 土地调整情况 你村哪一年分田到户?(实行联产承包制) 你家现在有几口人分到地? 人 从分田到户至今, 你们村有过几次土地大调整? 次 最近一次土地大调整是在哪一年? 年 从分田到户至今, 你家参加过几次土地小调整? 次 最近一次土地小调整在哪一年? 土地大调整是由谁决定的? 1.村决定 2.小组决定 3.有时由村决定, 有时由小组决定, 4 其他(说明) 土地小调整是由谁决定的? 1.村决定 2.小组决定 3.有时由村决定, 有时由小组决定, 4 其他(说明) 你估计以后还会有土地大调整吗? 1 会有,  2 不会有,   3 不知道 你估计以后还会有土地小调整吗? 1 会有,  2 不会有,   3 不知道   L2 农户对土地的使用决策情况 你能决定种植什么作物吗? 1 可以  2 可以, 但需村里批准  3. 不可以.  你能将土地转作其他用途吗, 如鱼塘, 果园等? 你能将土地转给本村人吗? 你能将土地转给外村人吗?    194  M 林地和农地面积的变化情况 你的耕地面积在最近六年发生变化的情况 (填下表) (农地转林地不受时间限制) a .调整年份 b.增减情况 1.增加 2.减少 c.变化数量(亩) d.原因编码         变化原因编码: 1.家庭人口增加  2.家庭人口减少  3.自愿退还给村或组里  4.没有完成税费和 订购任务  5.耕地撂荒  6.违背了种植计划  7.因孩子外出而被收回  8.分家  9.转作其他 农业用途,如耕地改林地或果园,林地或果园改耕地,等等   10.其他(请说明)   你的林地面积在最近六年内发生的变化(农地转林地不受时间限制) a. 林地经营类 型编码 b.年份 c.增减情况   1.增加 2.减少 d.变化数量(亩) e.原因编码        变化原因编码 家庭人 增加 家庭人 减少 自愿退还给村或组里 林权改革  195  N 采伐限额  单位 用材林 竹林 你家砍伐木材或竹材需要申请采伐指标吗? 1 是        2 否 最近 5 年你多少次申请采伐指标? 次 实际多少次获得了采伐指标? 次 最近三次申请采伐指标的情况(从现在往回推算) 第一次在哪一年? 年 你申请了多少采伐指标? 立方米 (根) 实际获得了多少采伐指标? 立方米 (根) 第二次在哪一年? 年 你申请了多少采伐指标? 立方米 (根) 实际获得了多少采伐指标? 立方米 (根) 第三次在哪一年? 年 你申请了多少采伐指标? 立方米 (根) 实际获得了多少采伐指标? 立方米 (根) 你从哪里获取采伐指标? 1 村 2 乡 3 林业站 4 林业局 5 购买 其他  你对采伐限额的看法怎样, 有什么意见或建议?  >>>>>跳至最后一页作 RISK GAME(上午); 跳至社会资本表格第二页作 TRUST GAME STEP 2(下午)   196   O 农业生产 O.1.  2006 年农业生产情况 农地面积:              亩 单位 作物 1 作物 2 作物 3 作物 4 作物 5 作物 6 作物 7 作物 8 作物 9 作物 10 作物 11  实际面积 亩 01 种植面积 亩 02 总产量 公斤 03 销售量 公斤    其中当年生产的数量 公斤 04 销售价格 (如果没有销售问当地平均销售价格) 元/公 斤  05 购买种籽秧苗费用 元 06 农家肥投入量 公斤 07 农家肥估价 元/公斤 08 化肥支出 元 09 农药费用 元 10 畜力费用 元 11 机械服务支出 元 12 排灌费用 元 13 雇工数 工日 14 雇工日工资 元/ 工日  15 其他生产性支出 元   197  O.2.  2000 年农业生产情况 农地面积:              亩 单位 作物 1 作物 2 作物 3 作物 4 作物 5 作物 6 作物 7 作物 8 作物 9 作物 10 作物 11  实际面积 亩 01 种植面积 亩 02 总产量 公斤 03 销售量 公斤    其中当年生产的数量 公斤 04 销售价格 (如果没有销售问当地平均销售价格) 元/公斤 05 购买种籽秧苗费用 元 06 农家肥投入量 公斤 07 农家肥估价 元/公斤 08 化肥支出 元 09 农药费用 元 10 畜力费用 元 11 机械服务支出 元 12 排灌费用 元 13 雇工数 工日 14 雇工日工资 元/ 工日  15 其他生产性支出 元    198   P.畜牧业   家畜禽 编码 年初存栏数 购进 仔畜数量 购进仔畜单 价 出生数 死亡数(包 括被偷数) 出栏数 (=出售数+自 家食用) 出售数 出售价格 单位产品 净赚多少? 销售收入 (包括卖蛋的 收入) 年底存栏数 送他人的 数量 在林间散 养的数量 (数量) (数量) (元/单位产品) (数量) (数量) (数量) (数量) (元/单位产 品) (元) (元) (数量) (数量) (数量)           注:年底存栏数=年初存栏数+购进子畜数+出生数-死亡数-出栏数 家畜禽编码:1=牛;  2=奶牛;  3=种公牛;  4=马;  5=驴;  6=骡;  7=肉猪;  8=母猪;  9=仔猪;  10=种公猪;  11=绵羊;  12=山羊;13=鸡;  14=鸭;  15=鹅;  16=其它禽类; 17=兔子;  18=蚕;  19=蜂;  20=野鸡; 21=其它(注明):               ;  P.1.  畜牧业生产情况 ( 2006 年)  199     家畜禽 编码 年初存栏数 购进 仔畜数量 购进仔畜单 价 出生数 死亡数(包 括被偷数) 出栏数 (=出售数+自 家食用) 出售数 出售价格 单位产品 净赚多少? 销售收入 (包括卖蛋的 收入) 年底存栏 数 送他人的数量 在林间散养 的数量 (数量) (数量) (元/单位产品) (数量) (数量) (数量) (数量) (元/单位产 品) (元) (元) (数量) (数量) (数量)           注:年底存栏数=年初存栏数+购进子畜数+出生数-死亡数-出栏数  家畜禽编码:1=牛;  2=奶牛;  3=种公牛;  4=马;  5=驴;  6=骡;  7=肉猪;  8=母猪;  9=仔猪;  10=种公猪;  11=绵羊;  12=山羊;13=鸡;  14=鸭;  15=鹅;  16=其它禽类; 17=兔子;  18=蚕;  19=蜂;  20=野鸡; 21=其它(注明):               ;  P.2.  畜牧业生产情况 ( 2000 年)  200   Q 其它现金收入  收入来源 2006 2000 收入来源 2006 2000 收入来源 2006 2000 填纯收入 (元) 填纯收入 (元) 填纯收入 (元) 01 渔业收入(净收入)   07 奖券收入  13.村分红收入 02 退休金   08 出租土地、设备或房屋的收入  14 农业补贴 03 医疗补助   09.出售交通工具的收入(如汽车、卡车和自行车等)  15 退耕补助 04 政府的其它补助收 入   10.出售房屋所得的收入 其他收入 1(注明)   05 银行存款利息收入   11.出售耐用品的收入  其他收入 2(注明) 06 贷款利息收入   12.接受的馈赠  其他收入 3(注明)    201   表 X 1  2006 2000 1.如果你家需要借 500 元钱急用,你能在一 个星期内借到吗? 1 是 2 否 2. 你能从哪儿借到这 500 元钱? 填代码(可多选) 3. 你家如果发生红白喜事一般如何操办? 填代码(可多选) 4. 你们村有哪些农民自发组成的社团 填代码并注明具体社团名称(可多 选)  5 你家参加了哪类社团? 填代码(可多选) 针对有参加社团的村民(如果参加多个社团需要分别问) 5-1 你们什么时候加入该社团?  注明年月 5-2 社团每个月活动几次 次(如果一年仅一次则为 1/12) 5-3 该社团有多少成员  人 5-4 该社团的成员是否属于同一个民族 1= 是 2= 否 5-5 该社团的成员是否都是同一个姓氏 1= 是 2= 否 6. 你们村是否有宗祠? 1= 是 2= 否 7. 你们家是否参加宗祠活动 1=是  2=否  借钱途经:1 本村人借(免息) 2 向本村人借(有利息) 3 向外村人借(免息)4.向外村人借(有利息) 5 信用社或银行 6 其他(说明) 红白喜事操办者:1.自家独立操办 2. 亲戚朋友一起帮忙办 3. 由本村专门的人办 4. 花钱雇外面的人办 5.其他(请说明) 民间社团:1. 生产组织 2.技术组织 3.销售组织 4.信贷组织 5.法律咨询/服务组织 6.宗教组织 7.婚丧事务组织 8.运动团体 9.学校事务组织 10.妇女组织 11. 文艺团体 秧歌队等) 12..其他组织(请注明)  X  社会资本  202  表 X.2 下面我们想了解一下你对一些问题的看法. 我们希望从几个方面来了解你对村干部,乡干部和县干部的看法.  请打分,1-10 分 分数越高表示越认同该种提法,不清楚请直接注明  村干部 乡干部 县干部 01. 干部可信赖 02 干部做事公平 03 干部能从农民的利益出发处理相关问题    203  游戏一、把农户的选择在下表中圈出并记录  甲 乙 1 20元 如果抽到 1 16元 如果抽到 2 3 4 5 6 7 8 9 10 35 元如果抽到 1 5元 如果抽到 2 3 4 5 6 7 8 9 10 2 20元 如果抽到 1 2 16元 如果抽到 3 4 5 6 7 8 9 10 35元如果抽到 1 2 5元 如果抽到 3 4 5 6 7 8 9 10 3 20元 如果抽到 1 2 3 16元 如果抽到 4 5 6 7 8 9 10 35元如果抽到 1 2 3 5元 如果抽到 4 5 6 7 8 9 10 4 20元 如果抽到 1 2 3 4 16元 如果抽到 5 6 7 8 9 10 35元如果抽到 1 2 3 4 5元 如果抽到 5 6 7 8 9 10 5 20元 如果抽到 1 2 3 4 5 16元 如果抽到 6 7 8 9 10 35 元如果抽到 1 2 3 4 5 5元 如果抽到  6 7 8 9 10 6 20元 如果抽到 1 2 3 4 5 6 16元 如果抽到 7 8 9 10 35元如果抽到 1 2 3 4 5 6 5元 如果抽到 7 8 9 10 7 20元 如果抽到 1 2 3 4 5 6 7 16元 如果抽到 8 9 10 35元如果抽到 1 2 3 4 5 6 7 5元 如果抽到    8 9 10 8 20元 如果抽到 1 2 3 4 5 6 7 8 16元 如果抽到 9 10 35元如果抽到 1 2 3 4 5 6 7 8 5元 如果抽到 9 10 9 20元 如果抽到 1 2 3 4 5 6 7 8 9 16元 如果抽到 10 35 元如果抽到 1 2 3 4 5 6 7 8 9 5元 如果抽到 10 10 20元 如果抽到 1 2 3 4 5 6 7 8 9 10  35 元如果抽到 1 2 3 4 5 6 7 8 9 10  农户代码:_______________ 你农闲时一个月打几次牌或麻将?_________     你觉得自己是一个爱冒险的人吗? (1=是  2=否 3=不知道(不读出))______

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0069996/manifest

Comment

Related Items