Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Bargaining through social networks : a conceptual approach and empirical application in Jambi, Indonesia Lenhardt, Amanda 2016

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

Item Metadata

Download

Media
24-ubc_2017_february_lenhardt_amanda.pdf [ 2.08MB ]
Metadata
JSON: 24-1.0340488.json
JSON-LD: 24-1.0340488-ld.json
RDF/XML (Pretty): 24-1.0340488-rdf.xml
RDF/JSON: 24-1.0340488-rdf.json
Turtle: 24-1.0340488-turtle.txt
N-Triples: 24-1.0340488-rdf-ntriples.txt
Original Record: 24-1.0340488-source.json
Full Text
24-1.0340488-fulltext.txt
Citation
24-1.0340488.ris

Full Text

 Bargaining Through Social Networks: A conceptual approach and empirical application in Jambi, Indonesia   by  Amanda Lenhardt B.A., Carleton University, 2006 M.Sc., University of Birmingham, 2008  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF Doctor of Philosophy in The Faculty of Graduate and Postdoctoral Studies (Integrated Studies in Land and Food Systems) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  December 2016    © Amanda Lenhardt 2016 ii   Abstract Local agricultural markets are deeply socially embedded and strongly governed by local social norms and structures and these affect the degree of bargaining power that a farmer holds when negotiating the sale of their goods. Standard economic measures of bargaining power typically do not capture the effect of social factors on bargaining power and therefore offer a limited understanding of farmers’ opportunities and constraints in the face of imperfect market competition and in the absence of formal contracts. This thesis establishes links between social network analysis concepts common to sociology and well-known economic concepts in order to demonstrate that social network measures can be used to observe the effect of social factors on farmers’ bargaining power. These concepts are used to investigate whether the structure of a market and a farmer's position within it affect the level of competition in the market and whether the strength of the relationships a farmer maintains with their traders affect their bargaining power and, by extension, the prices they receive for their goods. A conceptual model is presented to show the links between the social network concepts of network centralization, degree centrality and strength of ties to the economic concepts of competition, outside options, and informal contract enforcement. The model is tested by an empirical analysis of farmers’ bargaining power in 3 rubber markets in Jambi, Indonesia. The findings show that social network analysis can be used to respond to some of the challenges of conceptualising and measuring social factors in economics. Network centralization is used to show that unequal access to buying options among sellers can lead to monopsonistic competition, allowing buyers to capture greater surplus. Degree centrality shows that an actor whose position in the network is highly central has greater outside options and that these can be leveraged to bargain a higher price. Using a measure of strength of ties, it is also shown that the share of the bargaining surplus going to the farmer is increased when there is trust between them and the buyer.     iii   Preface This dissertation is an original intellectual product of the author, Amanda Lenhardt. The fieldwork reported in Chapters 4-5 was covered by UBC Ethics Certificate number H12-00116.     iv Table of contents Abstract	  ..................................................................................................................................	  ii	  Preface	  ...................................................................................................................................	  iii	  Table	  of	  contents	  ....................................................................................................................	  iv	  List	  of	  tables	  ..........................................................................................................................	  vii	  List	  of	  figures	  ........................................................................................................................	  viii	  List	  of	  equations	  .....................................................................................................................	  ix	  Acknowledgements	  .................................................................................................................	  x	  Chapter	  1:	  Introduction	  ..........................................................................................................	  1	  1.1	  Background	  to	  the	  research	  ........................................................................................................	  1	  1.2	  Research	  problem	  and	  research	  questions	  ..................................................................................	  4	  1.3	  Justification	  for	  the	  research	  ......................................................................................................	  6	  1.4	  Data	  and	  methodology	  ...............................................................................................................	  7	  1.5	  Outline	  of	  the	  thesis	  ...................................................................................................................	  8	  1.6	  Definitions	  ..................................................................................................................................	  9	  1.7	  Delimitations	  of	  scope	  and	  key	  assumptions	  ............................................................................	  11	  1.8	  Conclusion	  ................................................................................................................................	  12	  Chapter	  2:	  Literature	  review	  ..................................................................................................	  13	  2.1	  The	  challenge	  of	  measuring	  social	  factors	  in	  economics	  ............................................................	  13	  2.2	  Approaches	  that	  integrate	  economics	  and	  sociology	  .................................................................	  15	  2.2.1	  The	  field	  of	  social	  economics	  ....................................................................................................	  16	  2.2.2	  Dynamic	  game	  theory	  and	  trust	  ...............................................................................................	  17	  2.2.3	  Social	  capital	  theory	  ..................................................................................................................	  19	  2.2.4	  Social	  network	  analysis	  .............................................................................................................	  21	  2.3	  Empirical	  studies	  of	  social	  networks	  in	  rural	  livelihoods	  and	  agriculture	  ...................................	  23	  2.4	  Empirical	  studies	  of	  social	  capital	  in	  Indonesia	  ..........................................................................	  25	   v 2.4.1	  Social	  economic	  context	  of	  Indonesia	  ......................................................................................	  25	  2.4.2	  Relevant	  studies	  from	  Indonesia	  ..............................................................................................	  27	  2.5	  Conclusion	  ................................................................................................................................	  28	  Chapter	  3:	  Conceptual	  framework	  .........................................................................................	  29	  3.1	  Conceptualizing	  the	  market	  as	  a	  social	  network	  ........................................................................	  29	  3.2	  Market	  competition	  and	  network	  centralization	  .......................................................................	  32	  3.2.1	  Market	  competition	  and	  monopsonistic	  power	  .......................................................................	  32	  3.2.2	  Measures	  of	  market	  power	  ......................................................................................................	  33	  3.2.3	  Network	  centralization	  .............................................................................................................	  34	  3.2.4	  The	  difference	  between	  network	  centralization	  and	  the	  concentration	  ratio	  .........................	  37	  3.3	  Outside	  options	  and	  individual	  degree	  centrality	  ......................................................................	  39	  3.3.1	  Outside	  options	  ........................................................................................................................	  39	  3.3.2	  Degree	  centrality	  ......................................................................................................................	  41	  3.4	  ‘Strength	  of	  ties’	  and	  bargaining	  power	  ....................................................................................	  43	  3.4.1	  Strength	  of	  ties	  .........................................................................................................................	  43	  3.4.2	  Measuring	  strength	  of	  ties	  in	  agriculture	  .................................................................................	  44	  3.5	  Bargaining	  model	  ......................................................................................................................	  46	  3.5.1	  The	  bargaining	  game	  ................................................................................................................	  46	  3.5.2	  Determinants	  of	  E(α)	  ................................................................................................................	  50	  3.6	  Conclusion	  ................................................................................................................................	  53	  Chapter	  4:	  Characteristics	  of	  farm	  trade	  relationships	  in	  Jambi	  ..............................................	  54	  4.1	  Study	  site	  background	  ..............................................................................................................	  54	  4.2	  Survey	  sampling	  design	  ............................................................................................................	  57	  4.3	  Overview	  of	  sampled	  areas	  in	  the	  study	  site	  .............................................................................	  59	  4.4	  Survey	  design	  and	  variables	  used	  in	  the	  empirical	  analysis	  .......................................................	  60	  4.5	  Conclusion	  ................................................................................................................................	  63	  Chapter	  5:	  Empirical	  analysis	  .................................................................................................	  64	  5.1	  Introduction	  .............................................................................................................................	  64	  5.2	  Data	  .........................................................................................................................................	  65	  5.3	  Market	  characteristics	  ..............................................................................................................	  67	  5.3.1	  Price	  received	  for	  rubber	  ..........................................................................................................	  68	   vi 5.3.2	  Product	  quality	  .........................................................................................................................	  70	  5.3.3	  Transportation	  costs	  .................................................................................................................	  72	  5.4	  Farm	  production	  characteristics	  ................................................................................................	  73	  5.4.1	  Size	  of	  farm	  production	  ............................................................................................................	  74	  5.4.2	  Formal	  contracts	  and	  access	  to	  credit	  ......................................................................................	  75	  5.5	  Demographic	  characteristics	  of	  farmers	  ....................................................................................	  76	  5.5.1	  Age	  of	  farmer	  ............................................................................................................................	  76	  5.5.2	  Gender	  ......................................................................................................................................	  76	  5.5.3	  Farmer’s	  origin	  ..........................................................................................................................	  77	  5.6	  Social	  network	  characteristics	  ...................................................................................................	  77	  5.6.1	  Network	  centralization	  .............................................................................................................	  78	  5.6.2	  Degree	  centrality	  ......................................................................................................................	  79	  5.6.3	  Strength	  of	  ties	  .........................................................................................................................	  81	  5.7	  Econometric	  results	  ..................................................................................................................	  83	  5.7.1	  Base	  model	  ...............................................................................................................................	  84	  5.7.2	  Model	  2	  –	  network	  centralization	  ............................................................................................	  86	  5.7.3	  Model	  3	  –	  degree	  centrality	  and	  outside	  options	  .....................................................................	  87	  5.7.4	  Model	  4	  –tie	  strength	  ...............................................................................................................	  88	  5.7.5	  Model	  5	  –	  combined	  effect	  of	  degree	  centrality	  and	  tie	  strength	  ...........................................	  89	  5.8	  Limitations	  and	  suggestions	  for	  further	  inquiry	  .........................................................................	  92	  5.8.1	  Instruments	  ..............................................................................................................................	  92	  5.8.2	  Endogeneity	  ..............................................................................................................................	  92	  5.9	  Conclusions	  ..............................................................................................................................	  93	  Chapter	  6:	  Conclusion	  ............................................................................................................	  95	  6.1	  Research	  aims	  ..........................................................................................................................	  95	  6.2	  Situating	  the	  findings	  in	  the	  field	  ..............................................................................................	  98	  6.3	  Research	  contributions	  .............................................................................................................	  98	  6.4	  Strengths	  and	  limitations	  ........................................................................................................	  100	  6.5	  Research	  applications	  .............................................................................................................	  100	  Bibliography	  ........................................................................................................................	  103	  Appendix	  1:	  Survey	  ..............................................................................................................	  115	   vii  List of tables Table 1: Economic and social network concepts developed in this thesis ................................... 31	  Table 2: Regency selection by area characteristics ....................................................................... 58	  Table 3: Variables used in the empirical model ............................................................................ 61	  Table 4: Number of farmers by crop ............................................................................................. 66	  Table 5: Descriptive statistics for variables included in the model .............................................. 68	  Table 6: Village distances……………………………………………………………………………………... 73 Table 7: Average farm size by village .......................................................................................... 74	  Table 8: Village distance and network centralization…………………………………………………….78 Table 9: Base model and model 2 (network centralization) ......................................................... 85	  Table 10: Model 3 (degree centrality) .......................................................................................... 87	  Table 11: Model 4 (strength of ties) ………………………………………………………………………... 88 Table 12: Model 5 (interaction between degree centrality and tie strength) ................................ 90	    viii List of figures 	  	  Figure 1: Monopsonistic competition: Salop's circle .................................................................... 33	  Figure 2: Complete vs incomplete network density ..................................................................... 36	  Figure 3: The concentration ratio, network density and network centralization .......................... 38	  Figure 4: Degree centrality ........................................................................................................... 42	  Figure 5: Simulation model of relative bargaining power E(α) .................................................... 52	  Figure 6: Jambi Province, Indonesia ............................................................................................. 55	  Figure 7: Structure of Jambi's economy ....................................................................................... 56	  Figure 8: Agricultural sectors in Jambi Province ......................................................................... 57	  Figure 9: Variation in average price between villages .................................................................. 70	  Figure 10: Frequency of harvest and average price received by village ....................................... 72	  Figure 11: Size of production and average price received by village ........................................... 75	  Figure 12: Degree centrality and average price received by village ............................................. 80	  Figure 13: Distribution of tie strength by village .......................................................................... 82	      ix List of equations Equation 1 ..................................................................................................................................... 34	  Equation 2 ..................................................................................................................................... 35	  Equation 3 ..................................................................................................................................... 36	  Equation 4 ..................................................................................................................................... 41	  Equation 5 ..................................................................................................................................... 47	  Equation 6 ..................................................................................................................................... 47	  Equation 7 ..................................................................................................................................... 48	  Equation 8 ..................................................................................................................................... 49	  Equation 9 ..................................................................................................................................... 49	  Equation 10 ................................................................................................................................... 50 Equation 11……………………………………………………………………………………………………….. 51 Equation 12 ................................................................................................................................... 84 Equation 13 ................................................................................................................................... 84	      x Acknowledgements I wish to express my sincere gratitude to my thesis supervisor Jim Vercammen, who’s insightful contributions, support and patience throughout my PhD have made this thesis possible. His guidance helped me to progress my research and pushed me to grow as a researcher, and his constant support through the most challenging times helped to keep me on track. I am also grateful to my thesis committee members Rick Barichello, Kathy Baylis and Sanghoon Lee, who’s feedback and support have helped to guide me throughout my PhD. Their contributions helped me to expand my knowledge and research expertise and helped me to adopt a truly interdisciplinary research perspective. I am also indebted to the faculty and graduate students at the University of Jambi and staff at the Jambi Ministry of Research and Development who provided their time and tireless effort in collecting the data used in this thesis.  They also provided invaluable support during my time in Indonesia and made my stay in Jambi an enjoyable one. I am also thankful to my incredible support network of friends and colleagues. I would especially like to recognise my ‘thesis buddies’ – Birte Sniltsveit, Steve Morison and Greg Rekken – who each accompanied me on my PhD journey at different stages and without whom I wouldn’t have made it through. Special thanks are owed to my ‘anti-thesis buddy’ Stephen Nash who provided statistical advice at any hour, and listened to my thesis woes over countless brunches.  I also wish to thank Ryan Flynn for helpful feedback and for copy editing my thesis.  None of this could have been possible without the encouragement and support provided by my family. I would especially like to recognise the encouragement I’ve received from my Grandparents and the constant reminders from my Grandpas to hurry up with my thesis already. I dedicate this thesis to my guiding star, my mother. None of this would have been possible without her love and encouragement. She instilled in me a sense of possibility and perseverance that led me to this PhD and has seen me through to its completion. I will take these qualities with me in everything that I do.   1  Chapter 1: Introduction 1.1 Background to the research Smallholder farmers can face a number of marketing constraints that limit their ability to negotiate competitive prices for their goods: poor rural infrastructure leading to high transport costs; distance from central markets and poor communication infrastructure leading to imperfect information transmission; limited access to credit or formal insurance preventing longer-term investments and inhibiting risk taking; spatial segregation and limited investments in human capital restricting farmers’ awareness of improved crops varieties, technologies, or alternative buyers (Chamberlin & Jayne, 2012; Mendoza & Thelen, 2008; Gulati, Minot, Delgado & Bora, 2005; Salami, Kamara & Brixiova, 2010).  The purpose of this thesis is to understand how smallholder farmers engage their social networks to mitigate the marketing challenges they face. This will be done by comparing the bargaining outcomes experienced by farmers depending on the composition of their social networks. The unique contribution of this thesis is that it integrates a conventional economic framework for examining agricultural marketing with concepts developed in the field of social network analysis. This integrated approach emphasises the effect of limited selling options available to farmers, imperfect information about current market prices, and their reliance on trust in the absence of formal contracts when bargaining with traders over the price they will receive for their goods. It is well known that local agricultural markets are deeply socially embedded and strongly governed by local social norms and structures (Barret & Mutambatsere, 2008; Hinrichs, 2000; Minten & Fafchamps, 2001). However, economists rarely capture these social effects in standard market analyses. Failure to acknowledge the structure of social networks and the exchange of social capital in agricultural markets results in a highly incomplete understanding of the marketing decisions faced by smallholder agricultural producers. Agricultural policies that do not account for these social factors are more likely to be inefficiently designed, particularly in situations where  2 social structures prevent certain farmers from benefiting from them. Uninformed policies may even disrupt existing social structures which protect farmers from market failures, thereby exacerbating the challenges they face.  It is important to keep in mind that smallholder farmers are a heterogeneous group and that the markets they engage in vary considerably by commodity (particularly between perishable and non-perishable goods), location, and institutional setting (Arias, Hallam, Krivonos, & Morrison, 2013). Measures aimed at understanding the impacts of market imperfections on smallholder farmers must therefore account for farmers’ engagement with different types of market actors across different settings. The market failures described above are common in many rural settings, but the ways in which they manifest will depend on economic and social circumstances. This thesis approaches the idea of market imperfection as being affected by the social structures in which these markets operate. While social structures may not necessarily affect the efficiency of markets, this thesis will show that they do affect the distributional outcomes of markets. This view of market imperfection will be demonstrated conceptually and an empirical method will be proposed and tested to examine market imperfections from a combined economic and social perspective. This approach is based on the view that current market analysis methods fail to account for social variations and their effects on farm-level bargaining power (Hammouda & Osakwe, 2008).  The empirical analysis examines the rubber trade in Jambi, on the island of Sumatra in Indonesia. Jambi is an area with relatively poor rural infrastructure, making transport costly and limiting farmers’ opportunities to sell their goods directly to processors. Formal contracts between rubber farmers and traders are rare. These factors mean that most farmers exchange their goods with a limited number of traders and typically rely on informal mechanisms to ensure the outcome of bargaining over the price they receive is in their favour. This setting makes the rubber trade in Jambi an interesting case study to test whether social factors in these negotiations have an effect on the prices received by farmers. Empirical studies in development economics that examine social dynamics and social structures cover an array of sectors, but the agriculture sector has garnered particular attention given its importance in many developing economies. The lenses of social capital and social networks have  3 been used to study price outcomes (Fafchamps, Gabre-madhin, & Minten, 2005; Fafchamps & Hill, 2008; Songsermsawas et al., 2016), risk and vulnerability (Fafchamps & Gubert, 2007; Fafchamps, 2010), rural household welfare (Grootaert, Oh, & Swamy, 2002; Narayan & Pritchett, 1999; Kandpal & Baylis, 2016), and the extension of farming innovations (Fathoni, 2009; Maertens & Barrett, 2012; Rola, Jamias, & Quizon, 2002). The empirical approach developed in this thesis has been informed by this body of literature and builds on the empirical methods that have been tested in these studies.  The evidence put forward by these studies suggests that social networks play a significant role in smallholder marketing decisions. In Madagascar, Minten and Fafchamps (2002) find that farmers that are better connected and make significantly larger sales than farmers that are less well connected. In Uganda, Fafchamps and Hill (2008) show that occasional traders (those trading coffee for three months or less) tour coffee growing regions, inserting themselves between farmers and the traders they have longstanding relationships with to take advantage of farmers’ ignorance about international price fluctuations. In Philippines, Fafchamps and Lund (2003) show that risk sharing and informal credit mostly take place within small networks of family and friends. In Indonesia, Grootaert (1999) finds that households with higher social capital, a product of social networks, have higher per capita expenditure, more assets, higher savings, and better access to credit. The approach adopted in this thesis differs from earlier studies by proposing a new method to directly measure the social relationships between farmers and traders in order to observe emerging patterns that relate to the neoclassical economic concepts of competition and bargaining power. Few studies of trade and social networks measure social network factors directly as they often rely on secondary data and on proxies of social network factors.  Another feature of this thesis is that it combines social network concepts and economic concepts in a unique way. It uses familiar concepts from economic theory alongside familiar concepts from sociology to develop a fully interdisciplinary approach and proposes simple empirical measures for these concepts which are supported by fundamental principles from both disciplines. The literature review for this thesis did not uncover any previous study that has combined these particular economic and social network concepts. The data collected for this thesis, and its  4 application in the models that have been developed to understand the effects of social network factors on trade outcomes, are therefore unique contributions to the literature.  1.2 Research problem and research questions Bargaining power in smallholder agricultural markets has two main components: competition among buyers and trust between the farmer and the trader in the absence of formal contracts. Farmers are more likely to have low bargaining power when there is little competition among buyers, for example when there are few of them and when there is limited trust between the farmer and the trader, this being particularly problematic in systems of exchange where contracts are not used. When farmers have low bargaining power they are more likely to receive low prices for the goods they produce, with negative effects on their overall welfare.  Development economists use a wide variety of economic models to understand smallholder agricultural markets and prescribe welfare enhancing policies to mitigate the market failures that lead to low bargaining power for farmers. For example, spatial equilibrium models are often used to measure market integration and, when applied to real-world settings, can highlight bottle-necks, inefficiencies, and arbitrage (Barrett, 1996; Ravallion, 1986). Computable general equilibrium models are often used to understand the welfare (and other) effects of agricultural trade liberalization more broadly and can be tailored to specific policy interventions (Binswager & Quizon, 1986; Bourguignon, Silva, & Stern, 2002; Gunter, Taylor, & Yeldan, 2005; Kraev & Akolgo, 2005; Stifel & Thorbecke, 2003). Reardon et al. (2009) developed a heuristic model to observe how smallholders decide whether to enter modern market channels or to continue trading in traditional markets. An important component that neoclassical economic evaluation tools lack is a measure of the social environment within which market transactions occur. This leads users of these tools to overlook the social factors that drive micro-level bargaining constraints imposed on farmers resulting from imperfect access to price information, and their reliance on trust in the absence of formal contracts. Although these constraints may be known to applied economists, the discipline of economics has historically avoided measuring social factors and the impact of social interactions (Granovetter, 1985; Manski, 2000).   5  For example, Reardon et al. (2009)’s analysis of the transforming agri-food industry concludes that farmers benefit from modern procurement channels when modern companies pay a higher price for the product “in order to reward (‘lock in’) the farmer for supplying to that channel and thus reduce its risk of inconsistent supply and in search costs for new suppliers”. The authors admit, however, that this evidence is based on cross-sectional studies that do not “confidently estimate what portion of the greater earnings of the farms in the modern channel is due to participation in the channel, versus due to intrinsic characteristics of the farmer that allowed him/her to perhaps have superior earnings before entering the modern channel”. I argue that, beyond the intrinsic characteristics of the farmer referred to by the authors, the market structure into which these companies enter and the existing social relations between ‘traditional’ traders and farmers may also affect the potential gains made by farmers should they choose to ‘lock in’ to a modern supply channel. The potential benefits of being locked in to a longstanding, trusted trade relationship in a traditional market may compete with the benefits of entering a new trade relationship with a large entity.   There has been a growing recognition of the importance of interpersonal interactions and social norms in influencing economic activities in recent years in parallel with a growing interest in behavioural economics (Jackson, 2009). The field of game theory has opened up the discipline to the study of the role of reciprocity in economic exchange (see: Kandori, 1992; Berg, Dickhaut & McCabe, 1994). The role of peer effects in information diffusion has also captured the attention of empirical economists (see for example Montgomery & Casterline, 1996).  In sociology, Mark Granovetter (1985)’s seminal paper outlining the way in which social structures influence economic actions through networks of interpersonal relations captured the attention of this emerging field of social economics and laid the theoretical foundations for a number of lines of inquiry into the nature of social relations in various aspects of the economy.  This thesis takes Granovetter’s paper as a turning point in the literature around the social embeddedness of economic exchange, and anchors the response to this research problem in the discussions that have emerged since its publication. Following on from his line of inquiry related to the ‘strength of weak ties’, it examines whether benefits are better transmitted through a diffuse  6 network of weak ties, as he argued, or whether denser networks of strong ties better transmit benefits. This question is set in the context of agricultural trade (which was not the focus of Granovetter’s research) to understand the extent to which the reach of a farmer’s network, and the strength of the ties that make up that network, affect the prices farmers receive.  Given that smallholder farmers in most developing countries continue to bargain in non-centralized markets and without formal contracts, the effect of socially embedded trading relationships in determining farmers’ welfare outcomes remains an important issue. Yet policies aimed at mitigating market failures which negatively affect farmers’ welfare are currently developed using models that fail to incorporate the effect of social factors. Such policies not only risk being ineffective at improving outcomes, but also risk disrupting existing socially embedded markets and the strategies that farmers use to mitigate market failures.   This thesis poses three research questions to respond to this issue: (1) Can concepts developed in the field of social network analysis be used to improve our understanding of bargaining in agricultural markets? (2) How does the structure of a market and a farmer's position within it affect the level of competition in the market and, by extension, the prices they receive for their goods? (3) Does the strength of the relationships a farmer maintains with their traders affect their bargaining power and, by extension, the prices they receive for their goods? 1.3 Justification for the research Factors such as the increasing fragmentation of land, reduced public investment, competing demands for land and water resources, rising fuel and fertilizer prices, and climate change all contribute to a very precarious environment for smallholder farmers (IFAD, 2010; Lüdeke et al., 2004). Alongside these macro-level factors, market imperfections within the trade of agricultural goods often restrict market competition and leave farmers reliant on informal mechanisms to ensure a beneficial outcome from price negotiations. These factors often have a negative impact  7 on smallholder farmers’ bargaining power, and the lower prices they receive as result of this limited bargaining power has a negative impact on their welfare.  This thesis has two objectives: (1) To establish links between social network analysis concepts common to sociology and well-known economic concepts so as to demonstrate that social network measures can be used to observe the effect of social factors on farmers’ bargaining power; (2) to pilot these social network analysis measures in an econometric model and test hypotheses related to the expected effects of these social factors. This will be done by analysing a unique data set of approximately 200 farmers in Indonesia that includes information on the nature of farmers’ market relationships (who they trade with, how they know their trader, how long they have been trading, whether the farmer trusts the trader), the prices received, and background and production characteristics.  The case of agricultural markets in Indonesia was selected for this study as it demonstrates a market context in transition. Some farmers and some commodities in certain areas have maintained long-term and close-knit trading relationships. For other farmers, the nature of market exchange has shifted towards more short-term and disconnected trading relationships as Indonesia has opened to the world economy and value chains have expanded in areas once relatively closed to outside traders. This diversity will allow for a comparative analysis of different forms of trade relationships and the way in which social capital is exchanged in those different arrangements. The specific case of rubber farmers will be examined, a non-perishable good, thus allowing for abstraction from the pressures imposed by having to sell a good immediately following harvest.  1.4 Data and methodology This thesis has two main components. In the first conceptual component, three standard tools of social network analysis are shown to closely correspond to two three economic constraints that characterize smallholder agricultural markets – imperfect competition, outside options, and informal enforcement mechanisms.  The second component of the thesis is an empirical application of an integrated model to a case study that draws on a unique survey of smallholder farmers in the province of Jambi in Sumatra,  8 Indonesia. This region is particularly relevant to the research problem set out in Section 1.2 because it demonstrates many of the key challenges currently facing smallholder farmers, including poor road infrastructure, competing demands for land and resources, and restricted market access (Feintrenie, Chong, & Levang, 2010). This empirical analysis will test whether adding social network analysis variables to an agricultural price model enhances the explanatory power of the model. The dependent variable in the model is a measure of the bargained price of rubber received by famers. The explanatory variables in this model consist of standard controls that will be highly familiar to economists, as well as a set of social network analysis variables that will be highly familiar to sociologists.  1.5 Outline of the thesis  This thesis is organized as follows: Chapter 2 reviews the literature surrounding the integration of social network analysis into the research of agricultural markets and discusses existing studies that apply similar methods to empirical data. This chapter also compares the social economic approach to agricultural market analysis to the standard economic approach, demonstrating through examples from the Indonesian context that popular approaches like spatial and general equilibrium models fall short by failing to take social mediating factors and social capital into account. Chapter 3 builds on existing theoretical work related to the integration of social network concepts with economic applications and shows that the concepts of network centralization, degree centrality and strength of ties can be used to approximate measures of market competition, outside options, and bargaining power. Chapter 4 introduces the empirical data set collected in Jambi and describes the current social and economic context of the area of study. Chapter 5 applies the data collected for this thesis in an econometric model that controls for standard firm and market characteristics to test whether the addition of social network measures contributes to a better understanding of the bargaining outcomes realised by farmers trading in a market environment that offers choice between longstanding familiar traders and new unfamiliar traders. Chapter 6 concludes by drawing policy conclusions from the conceptual and empirical analysis presented in the thesis and proposes how this model could be applied by agricultural development policymakers.  9 1.6 Definitions A number of terms used in this thesis are not applied uniformly in the literature, therefore is it necessary to define how they will be used here.  Imperfect competition refers to the situation where individual firms have some degree of control over prices in a market. This thesis is concerned with the situation where buyers in a market maintain some control over the price paid for goods. This case of imperfect competition is known as monopsonistic competition.  Informal enforcement mechanisms are used to guarantee beneficial outcomes between parties to an exchange in the absence of formal contracts. These mechanisms may include social norms, customary law, alternative dispute resolution, and mechanisms of reciprocity and collective punishment (Kähkönen & Meagher, 1997). Social Capital has been identified in development economics as a fundamental factor among other forms of capital – physical, human, financial, and environmental (Woolcock, 1998). This thesis employs Putnam’s widely cited definition of social capital, which acknowledges social networks as the operationalization of social capital. Putnam (2000) argues that social capital manifests as “connections among individuals – social networks and norms of reciprocity and trustworthiness that arise from them”. This thesis will investigate how reciprocity and trust manifest in agricultural networks, with social capital being treated as a ‘non-physical tangible asset’. Whereas others have treated social capital as an intangible ‘claim’ (Chambers & Conway, 1991) or ‘residual’ (Hamilton & Liu, 2013), this thesis treats social capital as an exchangeable asset that can be acquired, lost, used as collateral, and which has a quantifiable value (see section 3.1 for further detail on what is meant by a ‘non-physical tangible asset’).  A social network consists of the horizontal social relationships maintained between smallholder farmers and the vertical social relationships maintained between smallholder farmers and traders who purchase from these farmers. Whereas social capital is an asset which is exchanged between two or more actors, an individual’s social network is the sum of their relationships and is defined by some relational boundary (such as market exchange, friendship, or membership in an  10 association), or the total sum of relationships of all actors set by some boundary (such as the whole market, the wider friendship circle, the entire association). Wasserman and Faust have outlined the study of social networks as comprising the following set of assumptions: •   Actors and their actions are viewed as interdependent rather than independent, autonomous units. •   Relational ties (linkages) between actors are channels for transfer or “flow” of resources (either material or non-material). •   Network models focusing on individuals view the network structural environment as providing opportunities for or constraints on individual action. •   Network models conceptualize structure (social, economic, political, and so forth) as lasting patterns of relations among actors.   (Wasserman & Faust, 1994, p.4) Degree centrality is a measure an actor’s position of prominence in a network. It is calculated by comparing the number of ties an actor maintains compared to the total number of possible ties. A prominent actor in a network is one that maintains many ties in a network, meaning that they are more “involved” in the network relative to other actors (Wasserman & Faust, 1994). Actors with higher degree centrality may have more access to, and be able to call upon, more of the network’s resources and have more alternative options than other actors in the network (Hanneman & Riddle, 2005). Throughout this thesis, ‘access to’ and knowledge of’ traders are used interchangeably to describe farmers’ available network of traders.  Network centralization is a measure of the overall connectedness of network, demonstrating the level of inequality among actors within a network. High network centralization indicates that the individual prominence of actors in a network varies substantially, meaning that some actors are better connected than others and are therefore more likely to have access to alternative options (ibid).  Tie strength is a “quantifiable property that characterizes the link between two nodes” (Petroczi, Nepusz & Bazso, 2007, p. 40). Components of tie strength include “the combination of the  11 amount of time, the emotional intensity, the intimacy (mutual confiding), and the reciprocal services which characterize the tie” (Granovetter, 1973, p.1361). Measures of tie strength vary depending on the context being analysed, but may include: reciprocated nominations, frequency of interaction, or through qualitative labels such as family member, friend or acquaintance (Krackhardt, 2003).  1.7 Delimitations of scope and key assumptions This thesis uses social network analysis as one possible solution to the research problem posed in section 1.2. However, there are likely to be other feasible approaches that could be used to conceptualize the social environment as a part of economic exchange. Furthermore, the social network concepts that have been employed by this thesis are foundational concepts from the field of social network analysis rather than more elaborate or unique contributions themselves. They have been explicitly chosen for their simplicity, as this makes them easier to map onto economic concepts and also helps to ensure that they are accessible to practitioners in other disciplines (in keeping with the integration theme of the thesis and the degree programme). Likewise, while the research questions test two particular economic concepts to demonstrate the conceptual links between economic and social elements of market exchange, there are likely to be other economic concepts that could respond to these research questions. It is worth mentioning, however, that other economic concepts were considered for this analysis, but these three concepts appeared to be the most appropriate matches to the social network concepts being explored, while also being among the more relevant but least understood socially influenced economic factors in smallholder marketing (this is explored in further detail in Chapter 3).  The empirical results in Chapter 5 are only demonstrative of the reality of smallholder marketing in Jambi, they are not representative of the entire country nor generalizable to other contexts (see Chapter 4 for further expansion on sampling and the representativeness of the data being analysed). The empirical analysis is intended to demonstrate the explanatory power of these concepts rather than to draw broader conclusions about how agricultural markets are structured or how farmers are situated within them. The policy conclusions that are drawn in Chapter 6 are therefore limited  12 to the social network analysis method’s applicability as a policy analysis tool and possibly the situation only in Jambi, rather than more normative findings on how markets should be structured or farmers’ positions within them improved. Examples are drawn from the empirical results to illustrate the concepts presented in Chapter 3, and lessons specific to the Jambi context are discussed, but these are the delimitations of what can be concluded from the findings of this thesis. 1.8 Conclusion This chapter has presented the foundations of the thesis, including a very brief overview of the literature that both motivates the research and to which this thesis intends to contribute. Chapter 2 builds upon this literature review and further situates this thesis within more specific debates in the field of social economics.  The research problem and research questions were also identified and these will be used to guide the thesis. Chapter 3 addresses these questions from a conceptual perspective while Chapters 4 and 5 provide an empirical case for the impact of social factors on market exchange outcomes.  The research agenda for this thesis was also justified in the context of global poverty reduction, with a particular focus in rural areas where the majority of people are smallholder farmers. The findings of this thesis can hopefully make a small contribution to the policy approaches that are adopted to address smallholder farmers’ marketing constraints and the social factors that impact their outcomes from trading relationships (see Chapter 6 for a full discussion of the policy implications of this research). On these foundations, the report can proceed with the research agenda that has been set.    13  Chapter 2: Literature review This chapter explores the concept of the market as a system of social actors engaged in economic exchange which is, at least in part, mediated by social factors. The aim of this chapter is to situate this approach in the existing literature. It shows how this approach is related to both neoclassical economic concepts – particularly game theory – as well as sociological concepts – particularly social network analysis. The chapter also explores the theoretical literature which has set out to understand the social nature of economic interactions, particularly in the field of social economics, as well as the empirical research that has explored these concepts in agricultural contexts.  The chapter begins with an examination of the challenges faced by economists in incorporating social factors into neoclassical interpretations of market interactions. This is followed by an overview of the literature that has set out to unify sociological and economic approaches to the market, and an exploration of the existing concepts that have been developed to explain the social factors involved in economic exchange. The chapter then proposes two areas of inquiry, one from economics and the other from sociology, which I propose share common understandings of economic relations. The chapter then summarizes existing empirical evidence of how socially embedded markets operate in practice by reviewing recent applied research that has set out to measure the operating mechanisms between social capital, social norms, and economic outcomes. The chapter concludes by highlighting applications of these models in the Indonesian context to contextualize the analysis that is undertaken in the empirical chapters of this thesis.  2.1 The challenge of measuring social factors in economics For most of its history, the discipline of economics has categorically divided forms of exchange as either economic – those which are self-interested and profit-seeking – or non-economic – those which are uninterested and immaterial in form. In other words, non-economic delineates those elements of market exchange which have no obvious monetary value or which are difficult to quantify (Bourdieu, 1986).  14 Until recently, economic models have not been fully equipped to control for the role that social interactions play in market exchanges; while economic models might account for social attributes like individual preferences, they are not designed to accommodate human emotions such as trust (Fafchamps, 2006). Granted, more recently, economists might recognize the role that certain manifestations of social capital play in market exchange, such as reducing transaction costs and informal contract enforcement, but there is not yet a widely accepted measure of social capital that is suited to sit alongside other market measures in economic models (Fukuyama, 2001).  Two important limitations of the neoclassical economic approach have contributed to the separation of economic and non-economic determinants of market exchange. First, the methods developed to observe market interactions have largely been designed with the individual firm as the focal point, and wider market observations are typically made by simply aggregating individual-level factors.  Benhabib et al. (2010) argue that one of the challenges that economics faces in measuring social factors is that the discipline relies on methodological individualism. By focusing on individual-level factors, the discipline tends to overlook structural-level factors that influence market decisions. While there are economic concepts that capture structural-level factors, some of which are explored in this thesis as they share common features to concepts employed in social network analysis, many of these use the aggregation of individual-level factors to observe the structural-level. These could therefore be complemented by building in additional elements such as those developed in the field of social network analysis.  For example, the concentration ratio (CR) is often used to measure the degree of competition in a market. It is calculated by summing the market share of the top four firms (CR4) or the top six firms (CR6) in a market. The overall market share captured by these top firms is then used to determine the extent to which they could collude to control the market. The measure is therefore a simple aggregation of the market power of the top 4 or 6 firms in a market. The CR does not account for the number of market actors that each firm actually interacts with (this is assumed by the firm’s size rather than the actual number of market relationships the firm maintains), the balance between the number of buyers and and the number of sellers in the market, or the past market behaviour of these firms and their propensity to collude. These structural factors require measures that look beyond the individual firm level towards the composition of the group of firms that make-up the market. This issue will be revisited in Chapter 3 where the concenrtation ratio is  15 compared with the social network concept of network centralization – a measure of network structural composition  The second limitation of the neoclassical approach relates to the assumption held by most theorists that social factors, while potentially influential in determining market outcomes, do not have a direct measureable effect and are therefore of lesser concern. Becker and Murphy (2000) suggest that economists tend to view social influences as having only an indirect effect on the behaviour of actors in a market, and argue that economists will not incorporate social factors into their analyses until other fields (such as sociology and anthropology) have developed effective techniques for analysing social influences on behaviour.  Until very recently, the examination of how social factors affect economic interactions has been largely absent from the discipline of economics, yet they have been at the core of the disciplines of sociology, anthropology, political science, and history for some time (Granovetter, 1985). The categorical division that economists have drawn between ‘economic’ and ‘non-economic’ aspects of market exchange has therefore led to economists and other social scientists using different analytical tools. It was not until the 1980s that the role of social context and social relations earned any traction in economics, largely owing to a few highly regarded theoretical papers and the emerging field of behavioural economics (Jackson, 2009). And it has only been in the last two decades that methodological tools to measure social interactions in an empirical way have been developed (Woolcock, 1998).  2.2 Approaches that integrate economics and sociology Despite the historical divergence between the disciplines of economics and sociology, there are areas in which economists and sociologists have found common ground. This section begins by examining the field of social economics, a discipline that strives for the integration of economics and sociology as its guiding aim. This is followed by an examination of one area of economics where social interactions have been built in – dynamic game theory – and one area of sociology where economic principles have been built in – social capital theory. Highlighting the commonalities between these two disciplines will help to build the foundations for the conceptual model that is presented in Chapter 3 of this thesis.  16 2.2.1 The field of social economics Many regard Karl Polanyi’s 1944 book The Great Transformation as the first attempt to bridge this divide between the economic precept of rationally self-interested economic actions and the sociological view of ‘socially embedded’ economic actions. Though Polanyi himself rarely used the term ‘social embeddedness’, and it has been argued that his work only briefly touched on the theme, his introduction of the notion of the ‘moral economy’ paved the way for the reunification of ‘economic’ and ‘non-economic’ concepts of market exchange (Beckert, 2007).  The term ‘moral economy’ was used by Polanyi to describe market relations as embedded in a social structure or set of social norms, emphasizing that market transactions do not occur in isolation of perceptions of morality (Granovetter, 1985). This assumes that “an actor is not an atomized individual” and “immediate utility cannot explain the full meaning of social relations” (Ghezzi & Mingione, 2007).  Polanyi presents the moral economy in a historical perspective as he sets out to track its influence on market interactions throughout human history. He observes that: “In the vast ancient systems of redistribution, acts of barter as well as local markets were a usual but not more than a subordinate trait. The same is true where reciprocity rules: Acts of barter are here usually embedded in long-range relations implying trust and confidence, a situation which tends to obliterate the bilateral character of the transaction.”      (Polanyi [1944] 1957: 61; quoted in Becket, 2007) The ‘great transformation’ referred to by Polanyi in the title of his seminal work was that which saw this moral economy becoming increasingly shaped, and eventually controlled, by markets, and in doing so becoming abstracted from the social norms and institutions that had once mediated it.  It wasn’t until Mark Granovetter elaborated on the social embeddedness of economic actions that the concept came into the mainstream, likely due to the tempered approach he adopts which comfortably situates his theory between the divide of economics and sociology. Unlike Polanyi, Granovetter does not view economic actions as becoming ruled by something separate from social norms (as Polanyi argued that market hierarchies were doing) but as being ‘embedded’ in social  17 relations. Change has occurred over time in the kind of social relations that economic actions are embedded in, but not the degree to which they are embedded in social relations (Woolcock, 1998). In his widely cited 1985 paper, Granovetter proposes that both ‘undersocialized’ constructs of the market, namely those put forward by economists, as well as ‘oversocialized’ constructs of the market, those adopted by other social sciences, fail to satisfactorily capture the agency and decision-making power of economic actors. He argues that both undersocialized and oversocialized definitions of market interaction assume that some structural determinant, either rational self-interest or social and cultural normative structures, is the driving motivation for all economic actions. His argument follows that: “Actors do not behave or decide as atoms outside a social context, nor do they adhere slavishly to a script written for them by a particular intersection of social categories that they happen to occupy. Their attempts at purposive action are instead embedded in concrete, ongoing systems of social relations”           (Granovetter, 1985, p. 487) Granovetter cautions both economists and sociologists against viewing the behaviour of actors in a particular trading environment as being wholly determined by some structural force, when in fact there are degrees of influence of both self-interest and social norms.  2.2.2 Dynamic game theory and trust Game theory is perhaps the approach best suited to examining social factors in markets from an economic perspective. Dating back to the writings of Antoine Cournot in 1838, and becoming a field of economic inquiry in its own rights with von Neumann and Morgenstern’s 1944 publication Theory of Games and Economic Behavior, game theory seeks to explain conflict and cooperation between individuals, firms, or groups (Turcoy & von Stengel, 2001). The theory is built on the premise that actors seek to maximise utility, while taking into account that other actors will also seek to maximise their utility when interacting with them, and this awareness shapes their behaviour (Buskens, 2002).  Manski (2000) suggests that a ‘radical consequence’ of the rise of game theory, particularly the approach adopted by theorists in the 1970s and 1980s, is that it enables economic theorists to study  18 phenomena that has traditionally fallen outside of economics. This includes the study of social norms and their impacts on economic outcomes as well as the use of trust to ensure mutually beneficial outcomes in the absence of formal contracts. Trust has been recognised as a critical factor in economic exchange by a number of economics; particularly those working in game theory. Arrow (1974) argued that trust is “ubiquitous to almost every economic transaction” (quoted in Berg, Dickhaut & McCabe, 1994, p.123) and Maclaulay (1963) suggested that “social pressure” and reputation” may be used more often than formal contracts” in economic exchange(quoted in Kandori, 1992).Theorists studying the role of trust in economic exchange typically point to the role of social reciprocity – both positive reward from good behaviour as well as retaliation for bad behaviour – as an informal mechanism to guarantee cooperation  Kandori (1992) distinguishes between two forms of reciprocity that can be used to enforce mutual benefits through informal social channels. The first is personal reinforcement, whereby mutually beneficial outcomes can be sustained in equilibrium if the same set of actors engage frequently through repeated interactions. In this repeated game scenario, bad behaviour by one actor party to the exchange triggers retaliation by the victim.   The second form of enforcement Kandori identifies is at the community level, where dishonest behaviour against any actor in the community leads to sanctions from other members of the community. Kandori goes on to construct a formal game theoretic model to show that “any efficient and rational outcome can be sustained when there are frequent interactions among the community members” and where minimal information transfer occurs to between community members (p. 64).  These concepts of personal and community-level reciprocity will be revisited in Chapter 3.  Game theory has created an opening for economists to observe the direct effect of social factors on decision-making – one of the barriers Becker and Murphy (2000) argued was preventing economists from accounting for social factors. However, the empirical methods needed to measure social phenomena – the second limitation of the neoclassical approach identified above – have not yet been developed.   19 Manski (2000) puts forward two reasons for the outstanding gap between game theoretic modelling and empirical research to explain the effects of social factors on economic outcomes. First, economic models tend to adopt sociological concepts and terminologies (such as social capital, which is explored below) without defining them with any precision. This makes it difficult to interpret the findings of these empirical analyses and to compare findings across studies. Second, the data used by economists to observe game theoretic behaviour are often poorly suited to observing social interactions. Approximations of social factors are typically relied upon, most of which are open to a wide range of interpretation.  This thesis seeks to overcome these two limitations. First, the integrated approach of this thesis, which is equally informed by economic theory and sociological theory, will carefully define the concepts being used and anchor these in the existing literature from both disciplines. Second, a unique data set has been collected to measure both economic and social factors directly. The survey used to collect this data was also designed with methodological considerations from both disciplines.  2.2.3 Social capital theory Alongside Granovetter’s influential 1985 paper attempting to broker diverging views of market interaction, the discipline of economics, particularly the field of development economics, began to recognise the role of social capital among other forms of capital. While physical and human capital had been well established determinants in the production and exchange of goods, the importance of social capital has only recently been recognized as an input into the production and marketing of goods.  From an economic perspective, social capital has been seen as both cost saving, where it reduces search costs (Fafchamps & Minten, 2002; Gabre-madhin, 2001; Minten & Fafchamps, 2001), and also cost inducing, where the maintenance of trusted relationships requires investment of time and costly favours (Coleman, 1990; Jackson, 2010). Alternatively, social capital has been conceived of as a ‘credit slip’, in that if a person does some favour for another, this person can later redeem the investment in that favour through a reciprocal exchange (Coleman, 1990).  20 The transmission of social capital though market exchange has been explored in a number of empirical studies in the last 20 years: as an exchangeable asset used to disperse risk between two (or more) market actors or leveraged for informal credit (Fafchamps & Gubert, 2007; Fafchamps, 2010; Knack & Keefer, 1997; Woolcock & Narayan, 2000); as an investment in market relationships that engenders trust and establishes reputations of good conduct in the market (Andriani & Karyampas, 2009; Fafchamps & Minten, 2002); and as a mediating structure through which market information transfers (Cox & Fafchamps, 2008; Fafchamps & Gabre-madhin, 2006; Minten & Fafchamps, 2001). These studies are some examples of the ways in which the divide between neoclassical economic and sociological approaches have been bridged through the application of social capital measures to economic questions.   In recent years, however, the term social capital has been one of the most misappropriated terms in the social sciences and some have questioned whether it is merely a fashionable term without substance (Burt, 2000; Durlauf, 2002; Lin, 1999). Inconsistency in the use of the term contributes to the challenge posed by Becker and Murphy above; without a clear demonstration of the direct impacts of social capital on market actors, economists will be discouraged from adopting measures of social capital into their own approaches.  This inconsistency stems partly from the fact that the concept of social capital lacks a universally applied definition, which results in confusion around how to measure it. Data constraints and the need for methodological efficiency have often meant that social capital is narrowly conceived in economic terms and imprecisely approximated (Andriani & Karyampas, 2009). This ambiguity of definition has damaged the reputability of the approach and led to studies that have applied social capital measures very loosely (Durlauf, 2002).  Social networks have increasingly been recognised as the mediating structure through which social capital flows. For example, Putnam’s widely cited definition of social capital as “connections among individuals – social networks and norms of reciprocity and trustworthiness that arise from them” employs social networks as the operationalization of social capital (Putnam, 2000). Similarly, Ronald Burt (2000) suggests that social networks are the ‘technology’ through which social capital is channelled.   21 Kenneth Arrow (1999), Ronald Burt (2000), and others (see Robinson et al., 2002 for a review) have suggested that, if we are interested in the variable outcomes that result from different endowments of social capital, then we should drop the ‘metaphor’ of social capital altogether and instead look at the way it is transferred. This suggestion is useful in responding to the challenge posed by Becker and Murphy above. Rather than using social capital as a catch-all term to include the various direct and indirect effects that social factors have on economic actors, economic applications of social capital should instead focus on the direct transmission of social capital. I propose that the field of social network analysis is best equipped to measure these direct social effects on market exchange and therefore an approach that can be used to bridge the gap between neoclassical economics and sociology.  2.2.4 Social network analysis The field of social network analysis shares similar roots and proponents to those of social embeddedness. Social network analysis seeks to understand the structural properties of networks of social actors by analysing both the nature of relations between individuals within a network, as well as the broader social structure in which those individuals act (Scott, 1988). It has been applied across a number of disciplines, though its applications in economics are only recently emerging and its uses in development economics have been relatively limited.  The discipline of social network analysis emerged in the 1930s as a graphing method to understand the web of social relations of small groups. In the 1970s it was taken up by a group of American sociologists who transformed social network analysis into a mathematical and ethnographic method used “to explain the importance of relationships among interacting units” with applications in disciplines such as economics, social psychology, and geography (Wasserman & Faust, 1994, p.3) Social network analysis has been easily adapted to the study of social factors in markets since the concept of a network very closely resembles that of a market, where individuals (or firms) interact in a particular environment and patterns of behaviour can be observed. The flow of social capital, like the flow of resources, can be observed at the macro level – the flow of social capital through network structures and normative systems such as through government institutions and legal  22 structures – and at the meso and micro levels, through the networks and norms that mediate interaction between individuals, firms, and other market actors (Grootaert, 1999). In addition to the flow of social capital, social networks are understood to be conduits for information diffusion, a prerequisite for market competition given that “knowledge of prevailing prices allows farmers to reap the gains from broader market search” (Goyal, 2010). Without information on the spatial distribution of prices, farmers may sell their goods to traders that offer lower than average prices (Aker & Ksoll, 2012).   The study of information diffusion through networks is one of the most common applications of social network theory. Studies of information diffusion have examined the spread of technological innovation through social networks (Bandiera & Rasul, 2006; Maertens & Barrett, 2012), of the distinct roles of information transfer and influence in disseminating information about microfinance opportunities and women’s empowerment programs (Banerjee, Chandrasekhar, Duflo, & Jackson, 2013; Kandpal & Baylis, 2016) and the study of information flows through online social networks (Lerman & Ghosh, 2010; Mochalova & Nanopoulos, 2013).  Information transfer is a principal function of agricultural markets as it is necessary to relay price signals, the demand and supply of goods and the means by which technology transfer occurs (Frenzen & Nakamoto, 1993). Although arms-length transactions translated entirely through price signals might appear unbiased, in practice they are not immune from informational gaps and inconsistencies (Uzzi, 1997). Where information asymmetries exist, as is often the case in agricultural markets in developing countries, gains from arbitrage are possible (Barrett, 2005; Rapsomanikis, Hallam, & Conforti, 2003; Shepherd, 1997).  Actors who behave as information brokers or who restrict information from others are positioned to take advantage of these bargaining deficits and therefore capture surplus value from those they maintain trade relationships with. Where price variability can be predicted through market information (knowledge of planned policy interventions, climate forecasts, updates on international prices etc.) we would expect farmers to be able to make production and marketing decisions accordingly and adapt to price fluctuations (Nerlove, 1956). However, in most developing countries, market information is restricted by poor communications and poor rural infrastructure (Barrett & Mutambatsere, 2005; Fafchamps, 2004).   23 The recent surge in mobile phone subscriptions has been seen as a potential solution to the problem of information asymmetry. As mobile phones subscriptions have become affordable to even the poorest farmers, various systems have been developed to distribute up-to-date information to farmers on commodity prices, weather, and crop advice. However, recent studies measuring the benefit of these services for smallholder farmers have found limited effects.  Fafchamps and Minten (2012) conducted a randomized control trial by randomly providing an SMS-based farmer information service to 1000 farmers across 100 villages in India. They find that although treated farmers are more likely to seek gains from arbitrage (specifically they were more likely to change the market at which they sell), there was no significant difference between the price received by treated farmers and the price received by non-treated farmers.   In a similar study conducted by Aker & Ksoll (2012) in Niger, the authors found that farmers who gained access to mobile phones diversified their production and sold more cash crops than farmers without access to mobile phones, but these farmers did not receive higher prices for their goods. The authors conclude that one explanation for the price of goods remaining unchanged with the introduction of mobile phones could be that despite greater access to price information, farmers’ bargaining power remained unchanged.   Studies of information transfer, and social network studies more generally, are by their nature context specific (as will be shown in greater detail in chapter 3) therefore these studies are only indicative of the limitations of mobile phones to improve price outcomes for smallholder farmers and may not reflect the situation in Jambi.  2.3 Empirical studies of social networks in rural livelihoods and agriculture Accounting for the role of social capital in agricultural development is still relatively new and the lack of clarity of definition has meant that only a few studies have taken this on empirically. But despite its loose definition, social capital has, since the late 1990s, been understood to contribute to agricultural incomes, productivity, welfare, and sustainable development (Sorensen, 2000). This thesis intends to address the need for a coherently defined conceptual and methodological framework for the analysis of social networks in agricultural trade, and the empirical research  24 presented in Chapters 4 and 5 is intended to contribute to the still emerging evidence-base on the role of social capital in economic exchange.  Narayan and Pritchett (1997; 1999) published some of the earliest empirical work which integrated social capital into an economic analysis of rural welfare. They show that the level of social capital and different social norms in villages in rural Tanzania led to differentiated household incomes across villages. They find that variation in incomes are more pronounced between villages than between households within villages, and that a one standard deviation increase in social capital (approximated by membership in a community association) led to a 20-50% increase in average household expenditure (Narayan & Pritchett, 1999).  Out of concern that group membership could be endogenous to household consumption (e.g. high incomes leading households to join more associations), they also employ an instrumental variable of ‘trust in others’ to determine whether trust is correlated with household income. Their results show that social capital is indeed an exogenous determinant of income as they find that trust is correlated with social capital but not with income.  In a study of social capital impacts on rural livelihoods in South Africa, Maluccio, Haddad, and May (2000) examine the causal relationship between social capital (also measured through associational membership) on rural consumption, using a panel data set collected in 1993 and 1998. They find that social capital did not have a significant relationship with per capita expenditure in 1993 but did in 1998 (Maluccio, Haddad, & May, 2000). The authors attribute the change observed in the influence of social capital on household welfare to the changing social and economic environment in South Africa during this time.  They employ a series of instrumental variables (a set of associational group characteristics) and village fixed effects, following the example of Narayan and Pritchett. One important difference in their results compared to the Tanzania study is the finding that household-level differences were more significant in determining the variation observed in per capita expenditure than village-level differences.  There have been further studies of social capital and social network influences in rural development, including: Knack and Keefer (1997) who use measures of trust and civic norms from the world values survey to understand how these contribute to economic growth; Grootaert, Oh,  25 and Swany (2002) who study how membership in local associations and network impacts rural household welfare in Burkina Faso; and Grootaert and Narayan (2004) who use the approximation of group and associational membership to look at different social capital endowments and across different socio-economic groups and locations across Guatemala.  One important element that each of these studies shares is the approximate measure of social capital using associational or group membership. As will be further developed in Chapter 3, this thesis aims to better approximate the flow of social capital through the use of social network analysis, rather than through a proxy measure as these studies have done.   However, one body of work does stand out as being well adapted to responding to the role that social capital plays in market interaction: a collection of empirical studies carried out by Marcel Fafchamps and co-authors over the past 20 years (Cox & Fafchamps, 2008; Fafchamps et al., 2005; Fafchamps & Gabre-madhin, 2006; Fafchamps & Gubert, 2007; Fafchamps & Hill, 2008; Fafchamps & Lund, 2003; Fafchamps & Minten, 2002; Fafchamps, 2010; Minten & Fafchamps, 2001). The empirical methods that have been adopted by Fafchamps and co-authors have been instrumental in the design of the model presented in Chapter 5.  2.4 Empirical studies of social capital in Indonesia A number of studies have explored the role of social capital in rural markets in Indonesia, and these provide important context to inform the approach of this thesis. Any study of social factors in economic exchange must be informed by a sound understanding of the context in which market actors are engaging. The context of both the economy as well as social norms is needed to interpret the network structures observed through social network analysis and also to understand the motivating factors behind actors’ decision-making.  This section begins with an overview of of the context of rubber farming in Indonesia before turning to the empirical studies of social capital.  2.4.1 Social economic context of Indonesia Cribb and Brown (1995) suggest that, beginning in the late 1960s, the rapid industrialization of the country led to a “corrosion of the long-standing social and moral ties which bound agricultural  26 communities together” (148-149, quoted in: Miguel, Gertler, & Levine, 2003). An empirical study of the impacts of migration during this industrialisation period finds that social capital gains were made in rapidly industrialising areas, but that outmigration from non-industrial, primarily rural areas led to social capital losses in those areas. (Miguel et al., 2003).  The 1997 Asian Financial crisis also had a major impact on Indonesia. It brought the impressive economic growth trajectory the country had been on to a halt, ended the authoritarian Suharto government, and sparked mounting social unrest. In the wake of the financial crisis, Indonesia witnessed significant short term increases in unemployment and poverty, along with rising prices for basic goods, including food and fuel. During this period, associational ties (the proxy most often used to measure social capital) declined among all income groups, especially among the poorest, while informal network ties increased (Wetterberg, 2007).  Indonesia is the world’s second largest producer of natural rubber and the province of Jambi is one of the largest contributors to the country’s total output (Peramune & Budiman, 2007). Rubber was introduced to Indonesia in the late 19th century from South America and the demand for rubber surged in the early 20th century due to the growing automobile industry (Bissonnette & De Koninck, 2015). Rubber became Indonesia’s largest export commodity between 1920s and 1950s and has maintained an important share of the country’s agricultural exports and garnered a great deal of political attention (Drakeley, 2005).  Rubber is a labour intensive crop and requires minimal capital inputs. Once trees are tapped, it is not time sensitive and storage does not require significant investments in infrastructure (Bissonnette & De Koninck, 2015). It is predominantly produced by smallholder farmers in Jambi and the sale of rubber makes up a significant share of household incomes in the province with nearly 40 % of household depending on rubber production. (Arifin, 2005; Kopp et. al, 2014).  Farmers typically trade rubber at the farmgate, selling to small village traders who are often medium-size famers themselves, and these village traders then typically the rubber on to district traders, while some trade directly to processing centres (Peramune & Budiman, 2007; Kopp et. al, 2014). There were nine processing centres in Jambi as of 2014 and five of these were based in the provincial capital, Jambi city.  27 2.4.2 Relevant studies from Indonesia One of the earlier attempts to measure the economic benefits of social capital in rural development, and the most relevant to this thesis, is Grootaert (1999) who sets out to measure the role of social capital in household welfare in Indonesia at the micro (household) and meso (community) levels. Grootaert borrows Putnam’s measure of social capital – membership in community groups and associations – as an approximation of social capital and adds this to a ‘conventional model of household economic behaviour’ composed of direct household expenditure and exogenous asset endowments (Grootaert, 1999). The study finds that households with greater social capital have higher household expenditure per capita, better access to credit, are more likely to save, and that these benefits are greatest for poorer households. Grooteart’s results also find that household background characteristics are highly significant in the model (ibid). A longitudinal empirical study by Wetterberg (2007) finds that those households who maintained local social connections through organisational memberships fared better during the period of the 1997 financial crisis than those that either did not hold such connections before the crisis or those that did not maintain them over the period. The study also shows that, during the crisis period, maintaining more local connections had a positive impact on household welfare, but that more distant connections had a negative impact on household welfare.  Levels of trust and associational membership, which are both measures of social capital, have also been found to vary across regions and by gender. Mavridis (2014), using nationally representative perceptions data, finds that levels of perceived trust and the prevalence of group membership varies according to ethnic diversity within the country’s regions, with more ethnically homogeneous areas having higher rates of social capital by both measures. Silvy and Elmherst (2003) show, using two case studies covering the time of the 1997-crisis, that social norms restrict women’s participation in certain networks, and that the socio-economic status of women within households can restrict their potential to be empowered or assisted by the social capital of their peer networks.  The evidence presented in these studies substantiates that social capital, transmitted through social networks, plays a significant role in rural market interactions in Indonesia. It has been shown that household welfare is at least partly determined by the amount of social capital held by a household as well as the structure of the environment in which households interact. Closer social connections  28 have been found to yield greater benefits, particularly for poorer households, and the maintenance of social capital has been found to ease the negative impacts of external crisis.  However, as with the studies highlighted in the previous section, the majority of studies which have sought to understand the influence of social capital and social networks on economic welfare in Indonesia have relied on approximate measures, such as associational membership, rather than directly tracing relationships between economic actors. The flow of social capital through the economy has largely been measured indirectly using general indicators of household welfare rather than by directly isolating economic activities, such as trade via markets, to understand the channels by which social capital is transmitted.  2.5 Conclusion This chapter has shown that despite the divide between the disciplines of economics and sociology in their approaches to the effects that social factors have in economic exchange, there are foundations on both sides from which to construct a more integrated approach. Social network analysis has been proposed as the bridging approach to achieve this integration. The next chapter will build on this theoretical foundation by developing a formal conceptual model that uses integrated concepts from both disciplines.  This thesis also builds upon the existing empirical research that has set out to observe the effect of social capital on rural welfare. What is unique about this thesis is that it goes beyond indirect measures of social capital (which has been the focus of the majority of studies identified in this chapter) by measuring the actual flow of social capital through networks, with a focus on interactions in agricultural markets and price outcomes. In doing so, this approach used in this thesis strives for a more precise estimate of the impact of social capital, and its transmission through social networks, on household welfare. While conclusions drawn from these estimates are limited to the specific context of Jambi, Chapter 6 links these findings to the broader Indonesian context and also shows how this approach could be applied in other country contexts and in other markets.     29  Chapter 3: Conceptual framework This thesis is interested in two key constraints faced by smallholder farmers when bargaining over the price paid for their goods: a lack of competition due to a limited availability of buyers, and the necessity of relying on trust in the absence of formal contracts. I propose that the effects of these two constraints can be directly observed through concepts developed in the field of social network analysis.    This chapter traces the conceptual links between the social network concepts of network centralization, degree centrality, and strength of ties, and the economic concepts of competition, outside options, and informal contract enforcement. By demonstrating the association between these concepts, this chapter will show how social capital that is transferred through social networks can be conceptualised as well as quantified such that its impacts on farmers’ bargaining power can be known by adding it to a standard economic model (Chapter 5).  The chapter begins by setting out an approach to viewing the market as a form of network, in a way that captures the measureable economic impact of social capital transferred via social networks. It then traces the links between the social network analysis and economic concepts that will later be applied in an economic model in Chapter 5 that measures the impacts of these factors on price outcomes faced by smallholders in Jambi, Indonesia. Section 3.2 examines network density and market competition, Section 3.3 explores the association between asymmetric information and degree centrality, and Section 3.4 examines the association between bargaining power and strength of ties. In section 3.5, these concepts are developed into a formal bargaining model.  3.1 Conceptualizing the market as a social network Modern economic theory tends to understand the market as an arena of anonymous market transactions (Granovetter, 1985; Jackson, 2008). The ‘boundary’ set by the discipline of economics  30 often implies that psychological, social, or political factors that can conceivably influence the market be designated as ‘non-economic’, or residual, factors (Parsons & Smelser, 1956).   There is an emerging view, however, that such ‘boundaries’ need to be relaxed to allow for the integration of different social science traditions in order to produce a more holistic approach to questions that incorporates both the economic and social factors involved in market exchange. This thesis is a product of such integration, falling under the discipline of ‘integrated studies in land and food systems.’ The concepts derived in the following sections emerge from this view, as they endeavour to integrate concepts from the fields of economics and sociology.  The conceptual approach for understanding the dynamics of market exchange proposed by this thesis is therefore termed a ‘dual’ one. It adopts both an economic and a social view of the market, which is defined as a set of social actors engaging in economic exchange.  This definition proposes a two-way interaction between the social and economic elements of market exchange: social actors are motivated by economic gains, and economic gains are mediated by social factors. The measures that are derived to observe the market must therefore account for this two-way interaction, meaning that the social elements of these measures must sit within a sound economic framework, and vice versa. The following assumptions are made to establish this two-way interaction: 1.  Social capital held by an individual and appropriated through personal relationships (such as family and friendships), or through repeated market interaction, is transferred through social networks. Social capital is a ‘non-physical tangible’ asset which can be acquired, lost, and used as collateral, and which has a quantifiable value (own definition).  2.  Markets are embedded in social networks. Decision-making in markets is influenced by an actors’ position in a network, and market structure is partly determined by the network structure.   Social network analysis can be easily adapted to economic questions since the concept of a network very closely resembles the concept of a market, where individuals (or firms) interact in a particular environment, and patterns of behaviour can be observed. Interactions between farmers and traders are but one form of social interaction among the many possible social interactions that could be studied by social network analysis methods.   31 Burt (1995) makes one of the strongest cases for the networked nature of markets. He argues that there is direct causality between social interactions: the degree of competition within markets and the resulting profit from market transactions. He states: The rate of return is keyed to the social structure of the competitive arena… Each player has a network of contacts in the arena. Something about the structure of the player’s network and the location of the player’s contacts in the social structure of the arena provides a competitive advantage in getting higher rates of return on investment…. Social capital is the final arbiter of competitive success.                        (Burt, 1995, p. 8-9)  The concepts developed in the following sections take Burt’s statement on social capital and competitiveness as a point of departure. Each of these concepts is adapted to understand how social networks, and different positions within them, affect farmers’ bargaining power in the market as evidenced by the price that they receive for their goods. One structural-level concept is developed to observe how the network as a whole affects farmers’ bargaining outcomes, and two individual-level measures are developed to observe how farmers’ positions within the network affect their bargaining outcomes. Table 1 provides a schematic of the economic and social network concepts that are combined to arrive at this two-way understanding of the market as a network. The subsequent sections of this chapter will develop each of these concepts in turn.  Table 1: Economic and social network concepts developed in this thesis   Economic Social network Structural  Market competition  Network centralization Individual  Outside options  Degree centrality Individual Informal enforcement mechanisms Tie strength  32 3.2 Market competition and network centralization Having established that the concept of the social network can be adapted to reflect the functioning of market interactions, I now turn to the first of three social network concepts that I adapt to use in the bargaining model presented at the end of this chapter. This section demonstrates the suitability of the social network measure of network centralization to capture the level of competition in a market.  3.2.1 Market competition and monopsonistic power Market competition is a structural-level feature of the market, and underlies the individual-level bargaining capacities of market actors. When one or a small number of firms are able to restrict the price or supply of goods in a market in a way that maximizes their economic benefit they are said to hold market power, thus reducing market competition. The level of competition in a market sets the context in which bargaining takes place: bargaining opportunities are highest in competitive markets, and lowest in uncompetitive markets.  In the context of agricultural markets which are spatially differentiated, access to villages from the central market varies significantly, as does access to each individual farm. It is therefore important to consider bargaining opportunities alongside location constraints. Access to bargaining opportunities will depend on the willingness of buyers to travel to villages and farms to trade. Though there may be a large number of small traders in a particular agricultural market, which would normally suggest that the market is competitive, each trader is likely to have spatially driven market power, which moves the market towards monopsonistic competition.  Market competition in this context can best be understood by the Salop’s circle model of monopsonistic competition. By adapting this model to the context of agricultural markets, we can consider a large number of farmers uniformly distributed along a circle and N traders evenly spaced along that circle with identical costs (see Figure 1). The traders offer to buy from farmers located to their left and to their right along the circle. Each farmer chooses to sell either to the trader on their left or their right, depending on who offers the highest price.   33 Due to the high transportation costs associated with accessing other traders along the circle available traders are limited, and each trader maintains local market power over the farmers nearest to them on the circle. This local market power enables traders to offer prices below full market value (p). In equilibrium, each trader offers the same price, (𝑝∗) , and transacts with the same number of farmers. The traders’ catchment area is the halfway point between the left and right edge between each pair of traders along the circle.  Figure 1: Monopsonistic competition: Salop's circle Source: Polinomics (2010) The positive margin earned by traders is used to cover their fixed operating costs. In situations where the space between traders means that there are more farmers per trader, there are higher overall gross margins that contribute to the trader’s fixed operating costs. In long run equilibrium, the spacing between traders is such that fixed costs are just covered by overall gross margins, meaning that each trader earns zero economic profits.  3.2.2 Measures of market power Measures of market power are used to “link the organization of firms to the degree of competition in a market and hence to predict the departure of price or rate of return from the competitive level” (White, 1982). These measures can be used to determine whether market power is being exercised by one or a small number of firms, and in doing so, they can capture the level of inequality among firms in a market (OECD, 2002). Commonly applied measures of market power include the  34 Concentration Ratio, Herfindahl-Hirshman Index, Lorenx Curve, Gini Coefficient, Inverse Index, and Entropy measures (ibid). The concentration ratio is one of the most frequently used measures of buyer or seller market power. This metric uses the number of buyers, or top buyers depending on the size of the market, and their combined market share. In markets with fewer buyers, or where a small number have a disproportionately large market share, we can expect those firms to hold monoposonistic or oligopsonistic buying power and the market can thereby deemed uncompetitive.  Typically the Concentration Ratio (CR) is measured at the four firm (𝐶𝑅%, top 4 firms) or eight firm (CR8 , top 8 firms) level. It is obtained by: 𝐶𝑅& = 	   𝑠*&*+,        where 𝑠* is firm market share and n represents the number of firms considered  Equation 1 The Concentration Ratio is considered to be one of the most generic measure of structural market power because it does not contain behavioural predictions that might be distinct to particular markets or vary across them (White, 1982). Therefore, one of the Concentration Ratio’s shortcomings is that if fails to capture the social structure that moderates the degree of market power present in a particular market.  Social mediating factors include the nature of the commodity (export vs. locally consumed, perishable vs. storable), cultural dynamics in the trading area, and economies of scale principles. These factors are certain to vary across different commodities, social settings, and geographical regions. This thesis uses the concept of network centralization (a variation of network density) to account for this variability, by observing the realized relationships among actors in a market rather than what is assumed to exist according to the Concentration Ratio (this differentiation is expanded on below).  3.2.3 Network centralization The more connections present in a network of a given size, the denser that network is. Similar to the concentration indices listed above, the social network measure of network density is used to  35 measure the level of concentration in a network. The main difference between these two concepts is that network density measures the number of actual connections (and compares this to the number of possible connections to determine overall density), whereas the concentration ratio only measures the market share of buyers without taking into account farmers’ abilities to access those buyers (this is similar to the shortcomings of outside options, which are identified and further clarified below).  Network density, a basic social network analysis measure, presents the number of ties in a network as a percentage of all possible ties. It is one of the most widely used measures of network cohesiveness due to its practical applicability across different types of networks types and because it examines the basic ‘connectedness’ among actors in a group (Friedkin, 1981).  For a two-mode network1, network density is calculated by:  ∆= 𝐿𝑛, ∗ 	  𝑛0 − 2 Equation 2  Where L is the number of lines (or ties) in a network, 𝑛, is the number of actors in the reference group (traders) and 𝑛0 is the number of primary actors (farmers).  A network density of 1 (or 100%) describes a network that is completely linked: each node is connected to every other node in the network. This is portrayed in the first panel of Figure 2, where all four actors (or nodes) are connected to each other by a line representing a relationship. Where one or more lines are absent, at least one actor is not fully integrated into the network. This scenario is illustrated in the second panel of Figure 2, where actor x has only one connection to the network while all other actors are connected to each other.                                                  1 The type of network referred to here is defined in more detail in the next section as this relates to the individual level of interactions.  36 Figure 2: Complete vs incomplete network density          Extending the concept of density further, it is also possible to observe the level of inequality among sellers in a market (further aligning the measure to existing market competition measures as described by the OECD above) using a measure of network centralization. Doing so will allow us to understand the impact of actor x’s relative disconnectedness compared to the fully connected actors in the second panel of Figure 2. For a two-mode network, network centralization is calculated by summing the differences between the most central actor in the network C(p*) and all other actors 𝐶(𝑝*) and dividing by the maximum number of connections possible in the network, which is taken over all bipartite graphs of specified node sizes 𝑛,, 𝑛0 (Borgatti & Everett, 1997). 𝐶6 = [𝐶 𝑝∗ − 𝐶(𝑝*)]𝑛, + 𝑛0 𝑛, − 2(𝑛, + 𝑛0 − 1) Equation 3 A centralized network is more likely to have a single dominant actor, or set of actors, with spatially driven market power. This leads to monopsonistic competition (described in section 3.2.1).  In a centralized network, these powerful actors are more central than other actors in the network; in other words, the network is unequal (Wassreman & Faust, 1994). This measure can also be understood as a measure of variability, dispersion, or spread (ibid).    37 3.2.4 The difference between network centralization and the concentration ratio Figure 3 shows how network density and network centralization differ from the classic Concentration Ratio measure. In Market A, the network share is divided equally among four buyers and 12 sellers have equal access to one buyer each, with three sellers per buyer. This market is highly concentrated, but sellers have equal access to the same number of large buyers.  In market B, the concentration ratio remains the same as market A, the same four buyers divide up the market equally between them. Now, however, one of the 12 sellers, F1, has access to all four buyers while the remaining sellers only have access to one buyer each. This increases the number of ties in the network from 12 to 15 (out of a possible 48), therefore increasing network density from 25% to 31.25%. According to the network density measure, Market B is considered more competitive because there are more connections between sellers and buyers. From a pure competition perspective this may be the case, but the increase in competition only benefits one seller; the remaining sellers remain tied to only one buyer each. This example shows how markets with a greater number of connections can still be less competitive.             38 Figure 3: The concentration ratio, network density and network centralization Market A: Concentration ratio = 40% Network density = 25% Network centralization = 0       Market B: Concentration ratio = 40% Network density = 31.25% Network centralization = 11.43%        Network centralization corrects for this by incorporating the unequal access that seller F1 has compared to the other sellers in the market, giving them a greater advantage when bargaining (increased bargaining through additional ties is explored in greater detail in the next section). Due to this actor’s centrality score being higher than others in the market (leading to inequality), market B is more centralized than market A. Market share   =10%  F1 F2 F3 F4 F5 F6 F7 F8 F F10 F11 F12 F1 F2 F3 F4 F5 F6 F7 F8 F F10 F11 F12 Market share   =10%  Market share   =10%  Market share   =10%  Market share   =10%  Market share   =10%  Market share   =10%  Market share   =10%   39 Observing this inequality is important for understanding the degree of competition in a market. Whereas the concentration ratio measures the number of buyers in a market without taking into account farmers’ abilities to access those buyers, as does network density, network centralization measures the number of actual market relationships between farmers and buyers and the extent to which farmers within the network are able to access different buying options. Where access to buyers is unequal among sellers, there is imperfect competition.  This leads to the first hypothesis that will be tested in Chapter 5: Hypothesis 1: Farmers have unequal access to buying options when there is imperfect competition in the market. Controlling for market centralization shows that farmers with a higher degree centrality earn a higher price for their goods on average.  3.3 Outside options and individual degree centrality  Having established a structural level measure of competition in the market using the social network analysis concept of centralization, I now turn to the individual-level access to buying options. This individual level measure will be incorporated into the formal bargaining model developed in section 3.5. This section demonstrates the suitability of individual degree centrality for measuring farmers’ outside options. 3.3.1 Outside options During the course of bargaining, any actor can usually quit the negotiation in order to take up the best option available elsewhere (Cunayat, 1998). When a seller receives an offer from a buyer, they can choose to reject the offer if they expect they can get a higher price from another buyer (after controlling for costs associated with waiting to sell). A farmer’s outside option is the best offer available to them, and the availability of an outside option can be used for strategic negotiation (ibid). If the threat of a failed negotiation is credible – meaning the seller could gain more value from rejecting the offer – the buyer must offer a higher price to match the outside option to avoid a failed bargain.   40 The concept of outside options originates in game theory. Rubestnein (1982) introduces a now widely applied strategic bargaining model that is based on the observation that economists tend to vaguely assert which optimal contract will be chosen by two rational individuals. He argues that this assertion uses a loosely defined concept of ‘bargaining power’. He goes on to develop a model which incorporates actors’ preferences and the rules of the game in an effort to directly observe actors’ actual bargaining power in strategic negotiations (Cunayat, 1998). He models this by introducing fixed bargaining costs and a fixed discounting factor for each actor in the negotiation to determine the perfect equilibrium; strategies chosen at the beginning of the game and all subsequent subgames form an equilibrium. While the concept of outside options has earned a great deal of attention in the theoretical literature, there is a dearth of data available to measure outside options; generic market surveys do not typically capture the level of detail required to examine outside options. As shown below, a simple count of the number of buyers in a market, which many measures of market competition use, is insufficient. A count of market actors may reveal a large number of buyers in a market, suggesting that the market has competitive outside options, but it may be that not all sellers have access to these options. Instead, to observe the number of buyers sellers are actually able to engage, one would need to count of the number of buyers each seller actually has access to. Detailed survey data on trading options is particularly relevant in the context of smallholder farming, where it is typical for there to be a relatively large number of small traders buying from farmers but the reach of those farmers is often limited by physical, economic, social, and other factors. As Stigler (1961) notes, farmers looking to sell their goods will accept a price from among the buyers that they happened to canvas at that time, meaning that a large number of traders in a market may have little impact on a farmer that has no access to them. An important feature of outside options will also be the farmer’s ability to store their product. For that farmers with access to storage, they may choose to hold on to their product when the prices being offered to them fall below the minimum price that they are willing to accept. Storage expands a farmers’ outside options beyond those traders that happen to canvas them at the that time to all future traders that canvas them, up to the length and capacity of their storage, thus increasing their bargaining power. This feature of outside options will not be captured by the social network  41 concept that is described in the next section, however it will be captured in the formal bargaining model in Section 3.5 in the form of costs associated with delay.   3.3.2 Degree centrality I propose the social network measure of degree centrality as a means to observe the effect of outside options on farmers’ bargaining outcomes. This measure is calculated by comparing the number of ties an actor maintains to the total number of possible ties. It is used in social network analysis to measure an actor’s position of prominence in a network. A prominent actor is one that maintains many ties in a network, meaning that they are more “involved” in the network relative to other actors (Wasserman & Faust, 2009). It follows that an actor with more ties, who is more involved in the network, has more options than an actor with fewer ties.  Individual-level degree centrality measures the number of ties an actor (termed node in the social network literature, and hereafter) holds as compared to the total number of actors within the network. In this thesis I am interested only in relationships between farmers and traders (therefore excluding relationships between farmers and between traders). This generates a directed 2-mode network, meaning that there are two distinct groups of interest that relate with one another (farmers and traders) and the analysis will focus on the relationships between those distinct groups but not relationships within them.  The degree of a node in a network is a count that ranges from 0, where the node maintains no other ties, to a maximum of g, where g represents the total reference group size, in this case the number of traders. Actor degree centrality 𝐶;(𝑛*) is calculated by counting the proportion of nodes (d) that are connected to actor 𝑛* as compared to the total reference group size (the total number of possible traders in the network):  𝐶;(𝑛*) = 𝑑(𝑛*)𝑔  Equation 4 Figure 4 shows the focal point of degree centrality. The red highlighted region shows the centrality position of one individual node in the network. This node is connected by five edges and therefore  42 has the highest measure of centrality in this network; all other nodes are connected by four, three, or two edges. Figure 4: Degree centrality   An actor whose position in the network is highly central has greater outside options to access from across the network. In the network depicted in Figure 4, the farmer with only two connections has the fewest outside options, whereas the farmer with five connections has the largest outside options.  This leads to the second hypothesis, which will be tested in chapter 5: Hypothesis 2: Farmers with a relatively higher degree of centrality have greater outside options. Greater outside options lead to increased bargaining power due to the threat of rejecting a low offer. Farmers with higher degree centrality therefore earn a higher price for their goods on average. By incorporating degree centrality into the bargaining model, it is possible to observe the first type of reinforcement described by Kandori in chapter 2:  •   Degree centrality allows us to examine the personal reinforcement mechanism, whereby mutually beneficial outcomes can be sustained in equilibrium if the same set of actors engage frequently through repeated interactions.   43  Similarly, network centralization can be used to observe the second type of reinforcement described by Kandori:   •   Network centralization examines the community level mechanism, whereby dishonest behaviour against any actor in the community leads to sanctions from other members of the community  However, on their own, these centrality measures do not explain the full rules of engagement and preferences that affect actors’ bargaining power within the network. For this, it is also necessary to measure the strength of ties between actors. 3.4 ‘Strength of ties’ and bargaining power The market failures examined so far – resulting from imperfect competition – are key components of the bargaining process for the smallholder. On their own, however, they do not fully capture the social mediating factors that determine the rules of engagement and preferences referred to by game theorists. I now turn to the bargaining power inherent in a farmer’s interaction with a particular trader based on the nature of their relationship. I propose that this bargaining power is largely determined by the level of trust held between the two market actors. I also argue that this trust is particularly important for bargaining in smallholder markets where formal contracts are not used to guarantee transactions.  3.4.1 Strength of ties This thesis employs the concept of tie strength to measure the degree of trust embedded in market interactions. The strength of a tie has been defined by Mark Granovetter as “a combination of the amount of time, the emotional intensity, the intimacy (mutual confiding), and the reciprocal services which characterize the tie”(Granovetter, 1973, p. 1361). Although stronger ties are expected to develop mutual trust and familiarity, it should be noted that they are also thought to reduce economic efficiency because they are more likely to result in networks of redundant contacts.   44 Jackson (2011) posits that strong ties can be leveraged as a form of bargaining power because “repeated interactions with specific partners help mitigate a number of problems related to moral hazard and adverse selection, and thus long-term economic relationships with known partners can dominate shorter-term anonymous transactions”.   Social obligation and reputation are core mechanisms of developing trust through social networks in the absence of formal contracts. These mechanisms motivate market actors to abide by implicit contracts, leading to barriers of entry and exit from the informal contract and maintaining the sustainability of trade relations with longstanding contacts or kinship networks (Greif, 1993).  The relationship between social obligation, reputation, and cooperative economic behaviour has been observed in a number of ways. Grief (1993) constructs a game-theoretical model to show that 11th century traders in the Mediterranean overcame risk in the absence of enforceable contracts through coalitions that rewarded trustworthy traders with access to information and strategic partnerships. Buchan, Croson, and Dawes (2002) employ an ‘experimental investment game’ in four cultural settings to test whether the social identity of actors in the game – be they familiar ‘neighbours’ or distant strangers – affects cooperation in economic exchange. Their study finds that cooperation decreases as social distance increases, with variable effects across cultures. Chandrasekhar, Kinnan, and Larreguy (2015) perform a similar experiment in India where participants who knew each other in different capacities are paired in a high-stakes risk sharing game to measure whether different degrees of social closeness – measured by actors’ centrality – help to determine cooperative outcomes. Their study finds that socially close pairs cooperate in the absence of contract enforcement whereas distant pairs do not. They also find that the underlying structure of the network affects the propensity to cooperate.  3.4.2 Measuring strength of ties in agriculture The importance of close contacts in the context of agriculture has also been measured. A study on the adoption of improved farming methods in Mozambique, empirically shows that information on new technologies diffused through closer social ties (family and friends) is four times more likely to induce a farmer to adopt a new technology than information from weak ties (those in the same religious cohort or those in other religions, Bandiera & Rasul, 2006). Even when controlling for other individual determinants of adoption, such as age, gender, literacy, relative poverty, and  45 whether the farmer has participated in an NGO project in the past, the social network effect on farmers’ propensity to adopt new technologies remains unchanged. The authors conclude that their social network measures are uncorrelated with other determinants of adoption. Songsermsawas et al. (2016) show that a large share of variation in revenues received by farmers in India can be explained by information accessed through peer relationships. Using a spatial econometric technique, the authors show that 60% of crop revenue is explained by relationships held by farmers, particularly farmers’ advisors and relationships held through caste-based networks. The body of empirical work by Marcel Fafchamps and co-authors also contributes evidence on the effect of close social relationships, trust, and cooperation in agricultural markets.  Fafchamps and Gubert (2007) find geographic proximity, which may be correlated with kinship ties, to be a strong correlate of risk sharing networks by facilitating monitoring and enforcement in the Philippines. Using data on trader relationships in Madagascar, Fafchamps and Minten (2002) find social capital to be more important than human capital, such as education and years of experience, for efficiency in agricultural trade. A complimentary study from Madagascar by Minten and Fafchamps (2001) finds that traders rely on social capital to overcome information and search costs in informal marketing relationships.   A measure of tie strength, capturing the familiarity of the relationship between the farmer and the trader, will be used in the bargaining model developed in the following section to observe the effect of strong ties versus weak ties. Four categories of familiarity are employed in the empirical chapter to measure tie strength. These range from family member (most familiar), to neighbour (less familiar than family member), to village contact (less familiar than neighbour), to external market contact (least familiar).  This measure of tie strength will be used to test the following hypothesis: Hypothesis 2: Trust is built through familiar market relationships. Farmers leverage the trust built through longstanding and intimate trade relationships they maintain when bargaining the price for their goods to obtain higher prices on average.  46 3.5 Bargaining model  The concepts described in this chapter will now be added to a formal bargaining model to demonstrate their expected effect on price outcomes. This model will be used to inform the empirical analysis developed in Chapter 5.  3.5.1 The bargaining game The trader and the farmer will bargain over the price W that the trader pays to the farmer at the farm gate. Let P be the market price of rubber in the central market. The price that the trader offers to the farmer will differ from P by the cost of handling and transportation from the farm gate to the central market, denoted by F, and the value of the farmer’s outside options.  The value of the farmer’s outside options (the amount they will earn if the trade fails) at the market-level is captured by the level of competition in the market as described in section 3.2, and is denoted by N. Competition in the market underlies the individual-level bargaining capacity of the famer, determining the set of traders available to them. The farmer’s outside options is measured using the network centralization of the market in which they trade. The trader’s outside option is zero. For the farmer, the profits earned from the trade are equal to the negotiated price W (we ignore production costs because these costs are sunk when the bargaining is in process). The farmer’s contribution to bargaining surplus is therefore equal the negotiated price minus their outside options, W – N.  Profits for the trader are equal to the price at which they sell to the central market, minus the price paid to the famer and the costs of handling and transportation to the central market, P – W – F. Because the outside option for the trader is zero, the trader’s contribution to the bargaining surplus is also equal to P – W – F.  The combined contribution to the bargaining surplus, accounting for the farmers outside options at the market level, can be expressed as S = (W – N) + (P – W – F). This can be rewritten as S = P – N – F.   47 The farmer receives a surplus equal to αS and the trader receives a surplus equal to (1 – α)S. The determinants of α are discussed below. A higher value of α therefore increases the bargaining surplus earned through the negotiation as the the famer leverages the trust maintained in the trade relationship to earn reciprocal benefits from the trade. Substituting P – Z – F for S the farmer’s bargaining surplus is equal to α(P - N – F). The price received by the farmer is their outside option, N, plus the negotiated surplus, α(P - N – F).  With Nash bargaining, the combined surplus is split according to the relative bargaining power of the farmer, α. The relative bargaining power of the farmer, α, is a function of the farmer’s level outside options (measured by network centralization) and the level of trust between the farmer and the trader (measured by strength of ties). More on this below.  As noted, the price received by the farmer in equilibrium is:  𝑊∗ 	  = 𝑁 + 𝛼(𝑃 − 𝑁 − 𝐹) Equation 5  Notice that equation 5 can also be rewritten as:  𝑊∗ 	  = 1 − 𝛼 𝑁 + 𝛼(𝑃 − 𝐹) Equation 6 Equation 6 shows that the price received by the farmer is a weighted average of the farmer’s outside option and the market’s competitive price (i.e., the price that results in zero surplus for the trader). The number of bargaining rounds the farmer may undertake is determined by the size of their individual network, as measured by degree centrality, denoted by NZ. The bargaining game is played as follows. First, Nature draws a value for α from the cumulative probability function G(α) and assigns the value to the first trader who approaches the farmer.  Let the selected value of α be denoted α1. During the first round of bargaining the farmer’s outside option is denoted by N1. To keep the analysis simple, I assume the value of N1 is equal to the expected value of W if the farmer bargains with the next trader who is randomly selected from G(α) minus a cost of bargaining delay and transaction equal to C. The cost of delay is the result of  48 direct costs, such as storage (as described in Section 3.3.1) as well indirect costs, such as costs to the household incurred by a delay in income (as in the case of poorer households that face an income shock when they are not able to sell their goods at regular intervals). As noted in section 3.3, the farmer’s threat of a failed bargain must be credible – meaning the farmer could gain more value from rejecting the offer – to be valid. In other words, the farmer must hold additional market contacts (degree centrality greater than one) for the threat to be credible. If the farmer only holds one market contact (degree centrality equal to 1), their outside option is zero. The expected value of W, referred to as E(W2), requires an assumption about the farmer’s outside option if bargaining with the second trader fails (i.e., N2). Similar to bargaining with the first trader, N2 is equal to the expected value of W if the farmer bargains with a third trader who is randomly selected from G(α) minus a cost of delay equal to C. This process of defining W and N continues up to NZ. Formally, let W1 and N1 denote the equilibrium price and the farmer’s outside option when the farmer bargains with the first trader. Similarly, let E(W2) and E(N2) denote the expected equilibrium price and the farmer’s expected outside option if bargaining with the first trader fails, let E(W3) and E(N3) denote the expected equilibrium price and the farmer’s expected outside option if bargaining with the second trader fails, etc. Using equation 6, the system of recursive equations that defines the equilibrium is therefore: 𝑊,∗ = [1 − 𝐸 𝛼 ] 𝐸 𝑊0∗ − 𝐶 + 𝐸𝛼(𝑃 − 𝐹) 𝑊0∗ = [1 − 𝐸 𝛼 ] 𝐸 𝑊D∗ − 𝐶 + 𝐸𝛼(𝑃 − 𝐹) … 𝑊E∗ = [1 − 𝐸 𝛼 ] 𝐸 𝑊E∗ − 𝐶 + 𝐸𝛼(𝑃 − 𝐹) Equation 7 Notice that E(α) has substituted for α in the first expression in equation 5. This implies that W1* in equation 7 is a measure of the price the farmer should expect to receive prior to the arrival of the first trader.  49 Because E(α) is the same for each new trader it follows that the E(W2*) = E(W3*), E(W3*) = E(W4*), etc.2 Consequently, the second expression in equation 7 can be solved for this common value of expected price, E(W2*) = E(W3*) = … = E(W*): 𝐸(𝑊∗) = 𝑃 − 𝐹 − 1 − 𝐸(𝛼)𝐸(𝛼) 𝐶 Equation 8            The expression for E(W*) in equation 8 can be substituted into the first expression in equation 7 to obtain  𝑊,∗ = 𝑃 − 𝐹 − 1 − 𝐸(𝛼)𝐸(𝛼) 𝐶 Equation 9 The first two terms on the right side of equation 9, P – F, is a measure of the competitive price (i.e., the price that results in zero surplus for the trader). The last term is the adjusted cost of delay that the farmer would incur if bargaining with the first trader broke down and bargaining with a second trader began. Similar to Rubinstein’s (1982) model of sequential non-cooperative bargaining, a larger cost of delay for the farmer implies a lower outside option, which in turn implies a lower equilibrium price for the farmer. This feature of the model is also supported by social capital theory, as discussed in Section 2.2.3, through the recognition that social capital induces costs given that the maintenance of trusted relationships requires investment of time and costly favours.  Equation 9 shows the reduced form expression for the equilibrium price that will emerge in the bargaining game between the first trader and the farmer. At this point the model is only capturing market level competition, and the farmer is not aware of their position relative to other farmers because it is not possible to observe the entire market; in other words, the farmer does not know the network centralization of the market they are trading in (this is an important feature that will be revisited in Chapter 6.   Therefore, although the seller’s outside option allows them to credibly threaten to walk away (if they maintain degree centrality greater than 1), they never actually do so. The bargaining outcome is always successful under these conditions.                                                  2 This restriction only holds if the number of potential buyers in the pool is large enough to ensure that the effects of sampling without replacement can be ignored.  50 3.5.2 Determinants of E(α) Most importantly, equation 9 also shows that W1* is an increasing function of E(α). In fact, equation 9 shows that the only link between W1* and the structural parameters of the model is through the E(α) variable. It is now time to discuss specifically E(α) links to these structural parameters. Because E(α) depends on the cumulative distribution function, G(α), from which α1 is randomly selected, it is sufficient to focus on the relationship between G(α) and the structural parameters. The two most important structural parameters are measures of trust and number of buyers. Let T be the level of trust between the farmer and the trader, measured by their strength of ties (as discussed in section 3.4). Strength of ties accounts for the bargaining power inherent in the farmer’s interaction with the trader based on the nature of their relationship. A higher value of T indicates a strong tie between the farmer and the trader while a lower value of T indicates a weak tie between the farmer and the trader. The trust variable T is a continuous variable ranging from a low of 0 and having no upper limit. As defined earlier, Z is an integer measure of the number of buyers that would potentially purchase from the seller. It follows that Z ranges from a low of 1 to an upper limit of K, with K representing the total number of buyers available in the network (taken from the denominator of the network centralization measure). A fully integrated farmer would have a degree centrality equal to K, meaning that they have access to all buyers in the market. More specifically, the case of perfect competition between buyers corresponds to Z = K and the case of perfect trust between buyers corresponds to T →∞ . Consequently if either Z approaches K or T →∞  then we should expect to see G(α) shift to the right and collapse so that ( ) 1E α → . To keep the analysis as simple as possible, assume G(α) is a uniform distribution with lower support αL and upper support αH where 0 ≤ αL < αH < 1. Consequently, E(α) = 0.5(αL + αH) Equation 10 The two parameters of the distribution, αL and αH, depend on T and Z. However, it is easier to assume αL = 0 and αH = 2E(α), and then model the relationship between E(α), T and Z rather the relationship between the two parameters, αL and αH, with T and Z. Specifically, assume  51 ( )11( ) 1 1T ZKE eλρα − −−= − −  Equation 11 It is straightforward to show using Equation 11 that as T gets large, E(α) goes to 1 for all values of Z. Conversely, when T is equal to (or close to) 0 in value, then E(α) goes to 0 with low trust and goes to 1 with high trust. This indicates that as trust increases between the farmer and the trader, the importance of the number of ties the trader maintains on the price that they receive diminishes. This feature of the model is important as it shows that trust can be used to compensate for limited buying options available to the trader.  Equations 9 and 11 combined, links the expected price received by the farmers to two key features of the market: Z, which is the number of traders the farmer has access to (the farmer’s degree centrality), and T, which is the level of trust that characterizes the relationship between this set of Z traders and the seller in question. This dependency of W1* on Z and T is the combination of a direct bargaining power pathway (i.e. how Z and T affect E(α) directly) and an indirect outside option pathway (i.e. how Z and T affect the seller’s outside option, which in turn affects W1*). A simulation model can be used to confirm this relationship between trust and degree centrality as described in the model. Figure 5 shows a simulation of the average value of 𝛼* and Zi for a hypothetical famer.3 We see that the average value of E(α)	  increases with increasing values of T  as well as with increasing values of Z.                                                  3 Figure 5 was generated by assigning values to the two parameters (ρ = 0.5 and λ = 1.5) and then using equation 11 within Excel to generate values for E(α) for alternative values of Z.   52 Figure 5: Simulation model of relative bargaining power E(α)  In this scenario, at the point Z=1 and T=0, E (α) = 0, the farmer has no outside options and no trust in the trader, therefore the trader captures all surplus value. When the farmer moves from one trader without trust (T=0) to two traders without trust (moving from the first line to the second line, Z=2), E(α) increases from 0 to 0.16, therefore increasing the farmer’s surplus through their outside options irrespective of trust in the trader. Moving to the third line (Z=3) increases E(α)=0.31, and so on.  As T increases from 0 to 1 (moving along the first line, Z=1), E(α) increases from 0 to 0.39, meaning that the farmer captures surplus through increased trust. Holding Z constant and increasing trust to T=2 leads to further increase in E(α) (0.63), and so on. Specifically, as T gets large, E(α) goes to 1 for all values of Z.  This simulation therefore shows that both trust and outside options can be used to increase the price received by farmers, with the implication being that trust can be used to compensate for a lack of outside options. This finding is crucial to the analysis as it shows that in the absence of outside options, farmers can engage their trusted market contacts to negotiate a higher price for their goods. This principle has important policy implications which will be explored in chapter 5.  00.10.20.30.40.50.60.70.80.910 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8E(α) TrustZ=1Z=2Z=3Z=4Z=5Z=6Z=7Z=8Z=9Z=10 53 3.6 Conclusion This chapter has built the conceptual foundation for this thesis. Having established that social network analysis concepts can be used to bridge the gap between neoclassical economic and sociological approaches to market analysis, this chapter then proceeded with developing these concepts into an economic framework.  A wholly integrated approach has been used to develop these concepts. Although the final model is anchored in an economic framework, a balanced combination of economic and social network concepts have been integrated throughout the model, making this approach accessible to both disciplines. I propose that this process on its own is a unique contribution to the literature in light of the historical divide between economics and sociology as explained in Chapter 2.  The model developed in this chapter will be applied in Chapter 5 using an empirical approach to determine whether these concepts hold in a real world setting. The social network measures described in this chapter will be estimated and used to predict the bargained prices received by farmers in three villages in Jambi, Indonesia.    54  Chapter 4: Characteristics of farm trade relationships in Jambi  This chapter describes the agricultural markets of a sample of villages in Jambi province, Indonesia. As discussed in earlier chapters, this contextual analysis is an essential precursor to the empirical study of the impact of social factors on market outcomes.  The chapter begins with an overview of the study site’s geographic, social, and economic characteristics, including a presentation of the agricultural market context. Section 4.2 outlines the methods used to collect data for the study and the sampling frame used to construct the social network sample. Section 4.3 provides an overview of the specific areas sampled through the survey, and Section 4.4 describes how the survey questions were constructed. Section 4.5 draws conclusions from this analysis by linking the case study’s context to the research question posed, and highlights some of the key contextual factors that will be used to interpret the econometric results in Chapter 5.  4.1 Study site background  The province of Jambi is located on the island of Sumatra, Indonesia’s largest and second most populous island (see Figure 6). Jambi has historically been one of the less densely populated provinces of the country; since 1970, however, its population has tripled and it was home to 3 million people as of 2010 (BPS Indonesia, 2013a). The province is geographically diverse, with hills in the West reaching 3000m and the low-lying coastal regions to the East largely composed of swamp area (Stolle, Chomitz, Lambin, & Tomich, 2003). The majority of land in Jambi consists of low-lying forest and plains.   55 Figure 6: Jambi Province, Indonesia  Source: Creative Commons  The Batanghari river connects the capital city of Kota Jambi to the coast, though no major port connects the province to export markets. Kota Jambi is the commercial centre of the province; the majority of trade from the province is channelled through Kota Jambi before being transported to other regional centres. Road infrastructure in the province is poor, so water transport is still relied upon for those farmers that have settled near rivers. Jambi is ethnically populated by Malay, Javanese, Minangkabau, and Batak people, as well an indigenous population known as Kubu (Daulay, 2011). Islam is the dominant religion. In the last 40 years, there has been a significant shift in Jambi’s demography through in-migration and the province’s poluation density has increased rapidly from 48 persons per km2 in 2000 to 62 in 2010 (BPS Indonesia, 2012). This can partly be explained by high population density of neighboring Java – one of the most densely populated islands in the world – where land scarcity has driven many agriculturally active people to migrate to other parts of the country, including Jambi.  Another factor behind Jambi’s population growth has been the National Program of Transmigration. Begninning in 1967, the program aimed to resettle people from densely populated areas of Java and Bali, as well as refugees from other areas, and migrants from within the area. Given Jambi’s relatively low population density, the province was among the top destinations for out-migrants from other areas. In Jambi, the transmigration programme resettled 340,203 people from outside the province and 86,330 people from within the province (Daulay, 2011).  56 Transmigrants tended to take up farming once settled in Jambi and most were given plots of land by the government and often encouraged into palm oil production (Daulay, 2011).  As Figure 7 shows, agriculture is the largest sector in the province, making up 29.83% of regional GDP; followed by mining (17.38%), commerce (15.77%), and industry (10.91%) (BPS Provinsi Jambi, 2013a). Agriculture also employs the greatest share of the population, with 49% of those employed in 2014 reported to be working in agriculture, followed by trade (19%), and services (18%) (BPS Provinsi Jambi, 2015). Tree crops (defined locally as plantation crops), largely consisting of rubber and palm oil, are the largest sectors in Jambi; followed by food crops, including rice and horticulture. Tree and food crops are also increasing in output at a higher rate than all other sectors (see Figure 8). Rubber production has been the dominant livelihood for most farmers as it is best suited to the region’s poor acid soils, while also suiting the province’s physical and human characteristics in that it has greater land requirements than labour (Levang, Yoza, & Tasman, 1999).  Figure 7: Structure of Jambi's economy  Source: (Budan Pusat Statistik Provinsi Jambi, 2013a)   AgricultureElectricityTransportMiningConstructionFinancial ServicesIndustryCommerceServices 57 Figure 8: Agricultural sectors in Jambi Province  Source: (Budan Pusat Statistik Provinsi Jambi, 2013b) 4.2 Survey sampling design A range of sampling design techniques for social network analysis research exist; in most cases, the choice of sampling design is subject to the form and size of the network to be observed and to the relations of interest (Rothenberg, 1995). This thesis has adopted a conventional, ego-centric design as described by Handcock et al. (2008).  The sampling frame for this study involved three stages. First, four of the province’s ten regencies were sampled, each exhibiting the different spatial, economic, infrastructure, and agro-ecological zones of the province. In the second stage, villages within these regencies were selected as representative of the primary crops and transport characteristics of the regency, with an emphasis on crops that displayed different farmer terms of trade (rubber, palm oil, coffee and pineapple). Table 2 provides a schematic of regency selection. Villages were purposively sampled because limited resources meant that the sample of farmers interviewed would be small and it was necessary to have a reasonable concentration of farmers of a certain crop in the same village to conduct the analysis. Regency and village selection was carried out through consultation with the Jambi Ministry of Research and Development and researchers from the University of Jambi. There 0500000100000015000002000000250000030000002008 2009 2010 2011 2012Regional Gross Domestic Product (millions of Rupiahs)Plantation crops Food crops Livestock Foresty Fisheries 58 is the potential of sampling bias towards villages that are more accessible by road, as staff from the ministry may have limited knowledge about more remote villages. This should be considered when interpreting the results as the sample may be more representative of villages with better transportation options than it is of more remote and less accessible villages.  Table 2: Regency selection by area characteristics  Regency Muara Jambi Tanjung Jabung Barat Sarolangun Merangin Villages Pondok Meja Sungai Bertam Tang Baru Teluk Sialang Bram Itam Kampung Nelaya Bukit Murau Muara Siau Dusun Tuo  Spatial characteristics Close to provincial capital       Good land transport Access to sea    Poor land transport Distant from regional markets  Poor land transport Distant from regional markets    Poor land transport Agro ecological characteristics  Lowland  Coastal, Swamp  Lowland  Highlands Primary crops Rubber, palm oil, local fruits Fruit, rubber, palm oil, rice, fisheries Rubber Coffee, rubber, rice, horticulture   The third sampling stage involved a random sample of farmers in each of the nine villages selected. Population data was not available for each village, meaning that it was not possible to calculate a precise sample size needed to represent the full population. An interview with a village administrative officer for Pondok Meja, the largest village in the sample, revealed an estimated 1000 households in the village. A minimum sampling threshold of 50 farmers per village, or 5% of households, was set based on this estimate. This threshold is above 30, the  commonly used in social research as a general rule of thumb (David & Sutton, 2004), and within the time and resource constraints of the study.   59 Survey enumerators were dispersed across the spatial area of each village to ensure a completely random sample. Maps were used to assign enumerators to different parts of the village and a range of sites were randomly chosen, with efforts made to ensure that remote parts of the village were included in addition to those closer to the main road. Enumerators travelled through the village on foot and were therefore able to survey those farmers with poorer road access. Variability in road access within villages was confirmed in the data as farmers were asked to rank the quality of the road used to access their farm, and a spread of quality is observed within villages (see Sections 4.3.2 and 4.3.3). Surveying took place at different times of the day to account for the fact that farmers of some crops would be away from home during the afternoon hours. Men and women were included in the sample, though culturally it would be expected that the male head of household would be responsible for household economic decision making.  4.3 Overview of sampled areas in the study site  Muara Jambi The Muara Jambi regency was selected for its proximity to Jambi City, a local economic hub with processing firms and economic links to regional markets. The primary commodities for this region are rubber and local horticultural crops, such as duku (a tree fruit that is not transportable over long distances), chilies (with some regional and export trade), and durians. An emerging market for palm oil has also been developing recently to replace ageing rubber plantations (Balit Bangda, personal interview, 3 February 2012). The villages of Pondok Meja and Sungai Bertam were selected for their proximity to the provincial capital and for having the strongest transport access to the regional markets. The village of Muara Siau was selected as an area which has seen a considerable amount inward migration through the government transmigration program.  Tanjung Jabung Barat  The Tanjung Jabung Barat regency was selected as a coastal region that is relatively connected to sea transport (primarily to the national capital of Jakarta and neighbouring country Singapore), though is less easily accessed by land from the provincial capital. This means that trade between Tanjung Jabung Barat and the provincial capital is restricted, as is trade from the rest of the  60 province via sea transport. Tanjung Jabung Barat’s primary commodities include rubber and palm oil. Food crops, including rice, horticulture, and fisheries, are also widely produced. Villages with high concentrations of these commodities were selected within the regency (Government of Tanjung Jabung Barat, 2012). The village of Teluk Sialang was selected to represent a fishing community, as this sector was identified as being particularly constrained by powerful traders, with small producers earning very low margins. The village of Tang Baru was selected as an emerging pineapple producing area, also considered relatively uncompetitive (Balit Bangda , personal interview, 2012).  Merangin The Merangin regency represents the highland area of Jambi. Although poor roads and long distances mean it is relatively disconnected to the provincial capital, it benefits from its central location in the province and has possible transport connections to West Sumatra. Rubber is the primary crop in this regency (124,568 ha), though coffee (10,577 ha), rice (19,782 ha), and horticulture (38,454 ha) are also dominant commodities (Government of Merangin Regency, 2008). The village of Dusun Tuo was selected as representative of the Regency’s primary crop, rubber, and the village of Muara Siau was selected to represent a coffee producing area of the Regency.  Sarolangun The Sarolangun Regency represents an ecologically diverse area where rubber remains the dominant crop. In 2008, 11,995 ha of rubber was planted and the government invested in rubber plantation regeneration through local and provincial programs. The regency borders the Merangin regency and shares similar transport characteristics. The village of Bukit Muraru was selected as representative of a rubber producing area that is further from regional markets than the other rubber producing villages in the sample.  4.4 Survey design and variables used in the empirical analysis  The full set of key variables employed in the empirical analysis and their construction in the survey is summarized in Table 3.   61 Table 3: Variables used in the empirical model                        Variables  Instrument Type  Survey question Average price received  Continuous  The recalled average price received by the farmer in Indonesian Rupiahs per kilogram (Rp/kg) over the last year. Quality (frequency of harvest per year) Continuous  The number of days the farmer harvests rubber in an average week, month or year; reported in days per year. Road quality Likert  Farmers were asked to rank the quality of the road to reach their farm ranging from 1 (poor quality) to 5 (good quality). Access to credit Binary  Farmers were asked whether they had access to formal credit (bank) and/or non-formal credit (trader, friend, family). Responses are coded as yes or no. Age of farmer Continuous  In years  Size of farmland Continuous  In hectares (ha) Distance of village from capital Continuous  The distance required to drive from Jambi City to the village as measured by Google maps. Network centralization Continuous  Calculated using equation 3 comparing the most central farmers’ reported degree centrality (see below) to the number of possible ties between all farmers and traders in the village.  Degree centrality  Continuous  Farmers were asked to estimate the number of traders that they have access to in order to trade rubber. This was an open-ended question, with responses coded as the value reported by farmers.    Tie strength (nature) Categorical  Farmers were asked to list the nature of the relationship between themselves and the traders with whom they traded. Options provided were family member or friend, neighbor, from the same village, market contact only. See discussion below for values assigned to relationship types.        62 The three social network variables used in the empirical analysis were developed using survey questions that set out to observe the wider social structure, respondents’ positions within the structure as well as values and statuses within the structure, three aspects of social networks highlighted by Coleman (1958) as potential areas of investigation through surveys.  Two important considerations in social network survey design are: whether respondents’ social contacts are identified with a list or roster of actors or by asking the respondent to use free recall; and whether a fixed maximum number of network contacts are elicited or an open-ended number of network contacts are requested (Wasserman and Faust, 1995). Because the full set of potential network actors was unknown when the survey was constructed, it was not possible to use a social network roster instrument in the survey. Free recall of network contacts among respondents was therefore used, which can be affected by respondents’ abilities to recall all market contacts.  With regards to the number of elicited contacts, both an open-ended question on the number of contacts available to the farmer was used as well as a recall of a maximum of five network contacts. Preference was given to the open ended question, recognising the inherent selection bias that can be introduced by limiting the number contacts listed cited by Holland and Leinhardt (1973) as well as the potential for recall fatigue introduced that could bias the number of contacts listed identified by Bernard, Killworth & Sailer (1982). This open ended question is also intended to compensate for the measurement error that could be introduced by respondents’ difficulties in recalling all market contacts by name. By asking for an estimated number of contacts, there is less onus on the respondent to list all market contacts by name.  This open ended question asked to list how many traders they knew of. Farmers were also asked to name up to 5 traders that they have recently traded with among these trading options. As the interest of the study is to ascertain farmers’ potential buying options (their full set of outside options) the open-ended question referring to farmer’s access to traders is used in the empirical analysis to determine degree centrality. This response is also used to calculate village-level network centralization using equation 3.   Measures of tie strength are highly dependent on the type of network being observed and the research question posed by a study of social networks. Marsden and Campbell (1984) distinguish  63 between two aspects of tie strength that one might measure: tie strength as it relates to the duration of the relationship between two actors, and tie strength as it relates to the nature of the relationship (such as the intimacy and emotional intensity). This study examines the nature of the relationship between the farmer and the trader.  Farmers were asked to select from four options to describe the nature of the relationship they have with their trader – family member or friend, neighbour, village contact or simply market contact (someone that they know purely through market exchange). These four categories were selected to distinguish between levels of intimacy and emotional intensity as described by Granovetter (1973) (see Section 1.7). Based on the assumption that levels of intimacy and emotional intensity are not linear between these four categories, they have then been assigned values using an exponential function in the empirical analysis: 80 (family/friend); 40 (neighbour); 20 (village contact); 10 (market contact only).   4.5 Conclusion This chapter has set out the social and economic context of the study site and described the sample design and survey construction. As explained in Chapter 2, this background information is important for interpreting the social network measures that are estimated in the model presented in Chapter 5. For example, knowing that the economy of the province of Jambi is reliant on agriculture demonstrates the signficance of agricultural markets in social and economic relations in the region. The history of social and economic relations in the province also serves as a background upon which to select important factors that could affect bargaining in agricultural markets in the region, as well as to assist in interpreting the effects of these factors in the subsequent analysis. The demographic characterstics of the region are also important background for the empirical analysis, particularly given the importance that individual characterists are likely to have on relationships between market actors.     64  Chapter 5: Empirical analysis 5.1 Introduction The purpose of this empirical study is twofold: (1) to test the hypotheses presented in Chapter 3, demonstrating the efficacy of the conceptual framework developed in this thesis; (2) to observe the functioning of a real world agricultural market using a social network approach, showing how it can be used to map market relationships and their effects on farmers’ bargaining outcomes.  Social network measures are applied to a standard economic model to demonstrate how these can be used to capture the social factors that affect farmers’ bargaining power. Chapter 6 will pick-up on the possible applications of this model in policy and program design and explore the replicability of this model.  This chapter is organized as follows: Section 5.2 describes the methods of data collection and sample refinement based on the sample collected through the survey. Section 5.3 describes the summary statistics of the markets sampled and highlights any association between variables that could impact the econometric analysis presented later in the chapter. Specifically, it explores any relationship between each predictor variable and the prices received by farmers as well as the relationship between the predictors and key social network measures. Sections 5.4 and 5.5 describe the production characteristics of sampled farms and demographic characteristics of sampled farmers, respectively. They also explore any possible relationships these may have with the outcome variables and key social network characteristics. Section 5.6 summarizes the key social network measures that will be the focus of the subsequent econometric analysis. The full results of the econometric analysis are summarized and discussed in Section 5.7. Section 5.8 sets out the limitations of these findings and proposes areas for further inquiry, and section 5.9 draws conclusions from the empirical analysis with these limitations taken into account.  65 5.2 Data  The data used in this study is taken from an original survey carried with support and oversight from the Jambi Ministry of Research and Development and the University of Jambi. The survey was piloted in the Muara Jambi regency in the first week of February, 2011 and the full survey administered between Febraruary 16th and March 4th, 2011. The survey contains three core modules: a farmer background module, a farm production module and a social network module. See annex 1 for the full survey. The survey was designed in consultation with empirical social network analysis experts and local experts at the University of Jambi. It was piloted in one village to test the survey instruments and to improve the language of the questionnaire to make it locally understood. Villages were selected to represent varying production and transportation characteristics. A conventional ego-centric4 sample design was used, consisting of a simple random sampling of a subset of the actors in each village, involving a complete observation of the relationships originating from randomly sampled farmer (a full description of the sampling design can be found in Section 4.2.1) . The survey sample includes 287 rubber producers, 52 coffee producers, 52 pineapple producers, and 42 palm oil producers. Additional products were identified in the sample but represent fewer than 30 producers and are not sufficiently representative. These products include areca nut, bananas, chilies, rice, and turmeric.  Table 4 summarizes the number of producers of each product by village. Most villages in the sample are represented by one dominant crop: rubber in Bukit Murau, Muara Siau, Pondok Meja, and Sungai Bertam; coffee in Dusun Tuo; and pineapple in Tang Baru. Two villages, Bram Hitam and Teluk Sialang, do not include a sufficient number of producers (30) of any particular crop and are therefore excluded from the social network analysis that follows in this chapter.                                                   4 An ego-centric network is used when it is not possible to sample the full network. It involves sampling a set of actors and identifying the actors that they are connected to them to establish a sample of network connections in the population of interest.   66 Table 4: Number of farmers by crop  Areca nut Banana Chili Coconut Coffee Palm oil Pineapple Rice Rubber Turmeric  Bram Hitam  3  4  0  7  1  25  1  0  0  0 Bukit Murau 0 0 0 0 0 4 0 0 62 0 Dusun Tuo 0 0 1 0 51 0 0 1 1 0 Kampung 0 0 0 0 0 1 0 0 0 0 Muara Siau 0 0 0 0 0 1 0 0 63 0 Pondok Meja 0 0 2 1 0 1 0 0 76 6 Sungai Bertam 0 1 2 0 0 5 0 0 85 0 Tang Baru 0 0 0 0 0 0 51 0 0 0 Teluk Sialang 10 0 0 14 0 5 0 28 0 0            Total 13 5 5 22 52 42 52 29 287 6  Networks are mapped by village and crop. Only villages with a sufficient number of producers (30) of one crop are included in the analysis. Given that the empirical analysis involves the comparison of social factors between villages (particularly for the measure of network centralization), it was necessary to limit the sample to those crops that had a representative sample (30 or more observations) in two or more villages. Therefore, it was necessary to drop the villages primarily composed of coffee, palm oil, and pineapple producers, as there was only one representative village for each of these crops. This limited the analysis to rubber farmers. Once missing data for key variables used in the analysis was taken into account, two additional villages were dropped from the analysis. The final sample includes 179 rubber farmers in Muara Siau, Pondok Meja, and Sungai Bertam.   The following four sections analyse market-level characteristics across the sampled villages. Section 5.3 examines agricultural products, Section 5.4 examines farm production characteristics, Section 5.5 examines farmer background characteristics, and Section 5.6 examines the key social network measures that were developed in Chapter 3. The set of predictor variables that are examined have been shown to affect farmers’ bargaining power according to the economic  67 literature. Studies that use a similar set of predictors for farmer bargaining include Courtois and Subervie (2014), and Labonne and Chase (2009).5 Justification for the inclusion of these variables is presented alongside descriptive statistics and an analysis of the association between a selected set of variables that are relevant to the model presented that is presented later in the chapter. This analysis will help to contextualize rubber bargaining in the three sampled villages and will also be used to support the econometric analysis. Specifically, this analysis will help to anticipate any correlation between key variables and price outcomes in order to test any multicollinearity between independent variables that could lead to the misinterpretation of the effect of these predictors.  5.3 Market characteristics A full summary of descriptive statistics for each of the key variables relevant to the econometric analysis is provided in Table 5.                                                    5 Education was considered as a control variable but the body of evidence on its effect on farmer bargaining power is very limited. It has therefore been dropped from the analysis in the interest of keeping the model efficient. Farmers’ access to credit was also considered as a factor that could affect their wiliness to take on risk, but very few respondents reported access to formal or informal credit. Farmers were also asked whether they maintained a formal contract with their traders, but like access to credit, formal contracts were nearly absent in the sample.   68 Table 5: Descriptive statistics for variables included in the model Note 1: Standard deviations in parentheses 5.3.1 Price received for rubber The variability of crop prices across villages can show the level of competition and the range of bargaining options in the market. Price variation, however, may also be explained by factors beyond competition and bargaining options, including the scale of production, the product quality, transaction costs, and access to financial resources that allows farmers to avoid selling when prices are low (Bijman & Meijernik, 2007). Each of these factors will be examined in the following sections.  The average price the farmer received (in Indonesian Rupiahs) per kilogram (Rp/kg) over the last year is used to measure bargaining outcomes between the farmer and trader. Farmers were asked                                                 6 For ease of interpretation, I present degree centrality instead of normalized degree centrality for all estimates in this chapter. This is equal to the number of traders the farmer has access to.                                  Villages   Instrument Pondok Meja Sungai Bertam Bukit Muarau Average price   14,504 Rp/kg (1,513) 12,867 Rp/kg (1,267) 12,552 Rp/kg (1234) Median harvest  frequency  52 days/year (141) 104 days/year (139) 52 days/year (63) Average road  quality 3.17 (0.87) 3.27 (0.82) 2.81 (1.09) Average age  of farmer 47 (13) 45 (13) 47 (12)  Average number of hectares 2.32 (2.00) 3.15 (2.30) 2.26 (2.18) Average degree  centrality6  3.22 (3.04) 3.90 (3.44) 2.73 (3.39) Median tie strength 21.42 (20.70)  26.91 (23.91) 36.60 (21.04)  69 to average out the prices they received for rubber over the last year (2011), in order to measure the typical price they received as opposed to a specific price which could be subject to temporal factors.  To confirm that this measure of price is robust, farmers were also asked to list the maximum and minimum prices they received in the last year. This was then compared to the mean between these extremes and the observed average prices. The average price reported through farmers’ recollection over the past year was 13,195Rp/kg with a standard deviation of 1,870.13Rp/kg; the mean average price received (calculated by averaging the maximum and minimum prices) was 10,126.82 Rp/Kg with a standard deviation of 5,607.50Rp/kg.  The deviation between the recollected average price received over the last year and the difference between the maximum and minimum prices received over the last year could be explained in two ways. One possible explanation is that farmer’s recollected average prices are higher than the actual average price. The second possible explanation is that the frequency of maximum and minimum prices received is not uniform, with minimum recalled prices being less frequent than maximum recalled prices over the past year.  Given the high price variability of natural rubber on international markets it is likely that the price received by farmers was not uniform throughout the year. This supports the second explanation for this deviation. Furthermore, a significant fall in international rubber prices over the course of 2011 from approximately 450 JPY7  in January 2011 to approximately 275 JPY in December  2011 (Trading Econoimcs & Tokyo Commodity Exchange, 2016) could indicate that the reported minimum price received was limited to September onwards.  In any case, the recollected average price falls well within one standard deviation of the difference between the maximum and minimum prices received. This measure is therefore considered to be a robust indicator of prices received.  It is also important to note that the average prices received by farmers varies between villages. Figure 9 shows that the highest average prices are received by farmers in Pondok Meja, followed by Sungai Bertam and Bukit Murau. It is important to note that this difference in prices corresponds                                                 7 Natural rubber prices are shown in JPY as rubber futures are available for trading primarily on the Tokyo Commodity Exchange.   70 with the distance of these villages from Kota Jambi, the regional center.  This variation in average prices, and the importance of distance, will be explored in detail in the following sections.  Figure 9: Variation in average price between villages   5.3.2 Product quality The price negotiated by the farmer will be highly dependent on the quality of the goods they produce. Rubber quality is measured by Dry Rubber Content, with the highest quality of rubber containing 100% Dry Rubber Content. Most rubber is downgraded due to water content, contaminants from the production process, and according to the chemicals used to coagulate the rubber (Kopp & Brummer, 2015).  It was not feasible to measure product quality directly (due to time constraints and insufficient information on the desired characteristics of the product). Notwithstanding these limitations, product quality is typically determined through visual inspection at the farmgate (Peramune & Budiman, 2007). In other words, quality is negotiated alongside price and traders are assumed to judge quality on their own terms as a part of the bargaining process.  5,00010,00015,00020,000Average pricePondok Meja Sungai Bertam Bukit MurauDistance from Kota Jambi  71 This implies that the final negotiated price is a function of both the intrinsic quality of the product (according to its Dry Rubber Content) as well as the farmer’s ability to negotiate with the trader’s perception of the quality of the good. To approximate the intrinsic quality of the good, the frequency of harvest is used to indicate the level of engagement the farmer has with their farm enterprise. Farmers who harvest more frequently are expected to be more engaged in their farming enterprise and therefore produce higher quality goods (ibid).  It is recognized that this approximation of quality is a weak predictor and that there is a risk of omitted variable bias by failing to account of product quality. However, it is expected that quality will at least be partly captured through the weak, but direct, approximation of frequency of harvest in addition to the farmer’s bargaining power in negotiating the perceived quality of the product through social network factors (indirect). Figure 10 shows that frequency of harvest varies by village, but does not have a strong relationship with the price received by farmers (shown by the line of fit in each graph). All three villages have a significant proportion of traders harvesting weekly (52 days per year), though Sungai Bertam has a larger share of farmers that harvest more than once per week, with a significant proportion of farmers harvesting daily. As a result, the median harvest frequency in Sungai Bertam is 104 days per year, compared to 52 days per year in Pondok Meja and Bukit Murau. This systematic difference in farmers’ frequency of market interaction will be considered as a confounding factor in the empirical model, though given that there is little relationship between frequency of harvest and the price received by farmers (the explanatory variable in the analysis) this is not expected to affect the results significantly.   72 Figure 10: Frequency of harvest and average price received by village  5.3.3 Transportation costs Traders in this context absorb transportation costs, given that transactions take place at the farmgate and these costs are reflected in the price offered to the farmer as deductions from the true market price (as discussed in the bargaining model presented in Section 3.5.1). Travel to villages from regional trading centers varies by distance and road quality and makes up the largest share of transportation costs. This is controlled for in the model by including a village level estimate for the amount of time required to reach the village from Kota Jambi, the province’s commercial center. These distances are summarized in Table 6.  5000100001500020000Average price (Rp)5000100001500020000Average price (Rp)5000100001500020000Average price (Rp)0 100 200 300 400Number of days0 100 200 300 400Number of days0 100 200 300 400Number of daysPondok Meja Sungai BertamBukit Murau 73 Table 6: Village distances  Distance to Kota Jambi (hours) Distance to Kota Jambi (km) Pondok Meja 0.38 12 Sungai Bertam 4.32 210 Bukit Muaru 5.72 293     Prices paid to farmers in villages further from the regional trading center are expected to be lower than those paid to farmers closer to the regional trading center. This is confirmed by running a Spearman’s correlation to assess the relationship between distance to reach the village and average prices received by farmers (both village-level estimates). There is a strong negative correlation between distance and price which is statistically significant, rs = - 0.5354, p = 0.000.  Within villages some farmers are harder to reach due to the remoteness of their farms or due to poor quality roads leading to their farms. Travel costs within the village is measured through farmers’ observations of the quality of roads to reach their farm on a scale ranging from one to five.  There does not appear to be a strong relationship between road quality within villages and average prices received by farmers, with a Pearson significance of p = 0.1947. However, it is important to note that there does appear to be a relationship between road quality within villages and degree centrality. This will be explored in Subsection 4.6.2 below.   5.4 Farm production characteristics Alongside social network characteristics, farm production characteristics are expected to impact the manner in which social interactions form and operate in agricultural markets in Jambi. Farm production characteristics are considered in this section to determine any impact these might have on market relationship formation (i.e. whether production size has any relationship with the number of buyers a farmer trades with).   74 5.4.1 Size of farm production The number of hectares under production is indicative of the scale of a farmer’s enterprise. Farmers with greater economies of scale are predicted to have greater bargaining power (World Bank, 2009). The impact of production scale on the price offered can be explained by the discounts traders receive from larger shipments, the convenience of transacting with larger enterprises, and the knowledge of farm management and price negotiation that is accrued by farmers with larger production volumes through negotiations over larger transactions (ibid).  However, there does not appear to be a statistically significant relationship between the number of hectares of rubber production and the average price received by the farmer (see Figure 11); a Pearson’s correlation shows p = 0.4136. Nor does there appear to be a correlation between the number of hectares under production and either of the social network variables.   Average farm size varies both from village to village as well as within villages. Table 7 summarises the average number of hectares under production among farmers in each village. Sungai Bertam has the highest average production size at approximately 3 ha, but also has the highest variance in farm sizes, with farms ranging from less than 1 ha to 20 ha.  Table 7: Average farm size by village  Mean Standard deviation Minimum Maximum Pondok Meja 2.28 2.0737 0.1 15 Sungai Bertam 3.35 3.0898 0.3 20 Bukit Murau 2.25 2.0822 0.5 14 Full sample 2.64 2.2243 0.1 20     75 Figure 11: Size of production and average price received by village  5.4.2 Formal contracts and access to credit  In order to measure access to credit, the survey asked farmers whether they had access to credit from a bank, trader, friend, family member, or any other lender. Only 14% of respondents identified as having access to credit and the majority of those identified government input subsidies as falling under access to credit. Access to credit therefore appears to be insignificant in the sample and is not included in the econometric analysis.  One further factor to consider with regards to farm characteristics, and their potential impact on bargaining power, is whether the farm maintains a contract with a trader, cooperative, or directly with a processing facility. Farmers were asked whether they maintained any form of contract with any of the buyers they had listed and, if so, to indicate the reliability of that contract (i.e. had it ever been broken). Only 5% of respondents listed any form of contract, the majority of which appear to be informal. Contracts are therefore considered insignificant to the bargaining analysis in this context.  5000100001500020000Average price (Rp)5000100001500020000Average price (Rp)0 5 10 15Number of hectares0 5 10 15Number of hectares0 5 10 15Number of hectaresPondok Meja Sungai BertamBukit Muarau 76 5.5 Demographic characteristics of farmers 5.5.1 Age of farmer Older farmers may have more longstanding trade relationships and have better knowledge of how the market has operated historically (Attoh, Martey, Kwadzo, Etwire, & Wiredu, 2014). Alternatively, younger farmers may have better access to price information than older farmers through information technology. The farmer’s age is included in the model to control for the possible effects of older farmers potentially having greater historical knowledge and younger farmers potentially having greater access to price information.  There appears to be a very weak positive relationship between average price received and age of the farmer. A Pearson’s correlation run to assess this relationship shows it to be insignificant, p = 0.3015. However, it is important to note that there is a positive statistically significant relationship between age of the farmer and tie strength; this is explored in Subsection 4.6.3 below. 5.5.2 Gender Traditionally, women are not expected to earn an income in Indonesian society (FAO, n.d.). Although this social norm has been changing over time, women continue to face difficulties in engaging in formal economic activities (de Vries & Sutarti, 2006). This is expected to negatively affect women’s bargaining power. The farmer’s gender is included in the model to control for any effects that gender may have on bargaining outcomes.  Only 24% of respondents in the sample are women. This distribution is spread fairly even across the three villages: women account for 29% of respondent in Pondok Meja, and 25% of respondents in both Bukit Muaru and Sungai Bertam. There does not appear to be statistically significant relationship between gender and average price received, with a Spearman’s correlation showing p = 0.3204. Nor does there appear to be a statistically significant relationship between gender and either of the social network measures.   77 5.5.3 Farmer’s origin Another potential factor in bargaining relations is whether a farmer originates from the area in which they are currently trading or if they have moved there from elsewhere. This could determine the amount of time they have had in which to develop market relationships. It could also influence whether they are considered as ‘outsiders’ for having moved to the area from elsewhere. This is particularly relevant for areas which have recently experienced an influx of farmers from elsewhere, both through the Indonesian government’s transmigration program and the scarcity of land in neighbouring islands that has pushed people to this less densely populated area of the country.  There does not appear to be any relationship between the number of buyers a farmer knows and their origins. The Pearson correlation is -0.01 and insignificant, suggesting that the number of market relationships formed is independent of whether or not a person is from the local area or whether they moved from elsewhere. Similarly, the correlation between transmigration and the number of buyers known is -0.0256, suggesting having been placed in the area through government intervention has very little impact on the number of market relationships formed.  Given that this factor is context specific (i.e. not part of a conventional bargaining analysis, but rather considered to reflect the socio-economic context of Jambi) and has not shown to have any significant relationship with the primary factors of interest. Therefore, farmer’s origins are not included in the econometric analysis.  5.6 Social network characteristics In addition to the standard economic predictors listed above, this model includes the social network measures developed in Chapter 3 to measure the effect of social network factors on farmers’ bargaining power and to test the hypotheses developed in Chapter 3. This section summarizes the social network characteristics of the sample and notes any significant relationships to other predictor variables described above.     78 5.6.1 Network centralization Network centralization, which measures the structural inequality in access to buyers among sellers in the market, is used to observe the level of overall competition in each network. Network centralization is used to test Hypothesis 1: Farmers have unequal access to buying options when there is imperfect competition in the market. Controlling for market centralization shows that farmers with higher degree centrality earn a higher price for their goods on average. As explained in Subsection 3.2.3, network centralization is calculated by summing the differences between the most central actor in the network (p*) and all other actors (𝑝*) and dividing by the maximum possible number of connections, which is taken over all bipartite graphs of specified node sizes 𝑛,, 𝑛0 	  (Equation 3).   According to this measure, Pondok Meja is the least centralized network at 5%, followed by Bukit Murau at 8%, and Sungai Bertam at 10%. This indicates that inequalities in access to buyers are lowest in Pondok Meja and highest in Sungai Bertam.  There is a strong positive relationship between network centralization and distance from the commercial centre, Kota Jambi. Village distances from Kota Jambi are listed in Table 8 alongside network centralization measures. A Spearman’s correlation was run to test the relationship between distance to the commercial centre and network centralization. This found a strong positive relationship between these two factors; r2 = 0.5753, p = 0.0000. This finding shows that distance from the central market affects the level of competition in the network; there are higher inequalities in access to buyers in more distant markets than in more centrally located markets.   Table 8: Village distance and network centralization  Distance to Kota Jambi (hours) Network Centralization Pondok Meja 0.38 5% Sungai Bertam 4.32 10% Bukit Muaru 5.72 8%        79 This finding is intuitive: greater distance from the commercial center means higher transportation costs, and higher costs leads, which may lead to fewer buyers.8 This restricted access to buyers increases the likelihood of inequalities among farmers in accessing available buying options. This finding also supports the location model of competition described in Subsection 3.2.1. It is important to acknowledge this when interpreting the econometric results of network centralization later in this chapter.   5.6.2 Degree centrality  The number of ties an actor holds, as compared to the total number of actors within the network, is used to observe outside options available to the farmer. Degree centrality is used to test Hypothesis 2: Farmers with relatively higher degree centrality have greater outside options. Greater outside options lead to increased bargaining power due to the threat of rejecting a low offer. Farmers with higher degree centrality therefore earn a higher price for their goods on average. As explained in Subsection 3.3.2, farmers’ normalized degree centrality is calculated by counting the proportion of traders that the famer (𝑛*) reports having access to compared to the total number of traders in the village, g (Equation 4).  Normalized degree centrality within villages is concentrated below 0.1, with a few farmers in each village maintaining a higher number of traders than their peers (see Figure 12). This supports the findings in Subsection 4.6.1 that access to traders is unequal within villages. The highest average degree centrality is found in Bukit Murau (0.059), followed by Pondok Meja (0.054), and Sungai Bertam (0.041).                                                  8 It is important to note that the number of traders in each location may also depend on the size of total rubber output per location. Villages with larger pooled outputs may receive more traders than those with smaller pooled outputs. It was not possible to observe total village output through this survey.   80 Figure 12: Degree centrality and average price received by village  There does not appear to be a relationship between distance to reach the village and degree centrality. Bukit Muara is the furthest village from the regional center, yet farmers maintain an average degree centrality score slightly higher than farmers in Pondok Meja, the closest village to the regional center. This finding suggests that, although distance to reach villages affects the level of inequalities among villages (the level of competition at the market level), it does not appear to affect farmers’ outside options within those markets.  As village distance is likely to be a strong determinant of the prices offered to farmers, due to the transportation costs that traders pass on to the farmer, it is important to distinguish this effect from social network factors. The correlation between distance and network centralization, identified in the previous section, is somewhat problematic for the measurement of market-level competition because distinguishing between the effects of these two factors on the prices received by farmers is challenging. However, the lack of correlation between distance and individual degree centrality shows that these effects can be distinguished at the individual-level, with distance showing no relationship to farmers’ available outside options in the sample. This indicates that, although the 5000100001500020000Average price (Rp)5000100001500020000Average price (Rp)0 .1 .2 .3 .4 .5Degree centrality0 .1 .2 .3 .4 .5Degree centrality0 .1 .2 .3 .4 .5Degree centralityPondok Meja Sungai BertamBukit Muaru 81 level of market competition at the village-level may be partly determined by the village’s distance to the market center, distance does not appear to affect the level of competition within villages.  Degree centrality does not appear to have a statistically significant relationship with farm production characteristics or with farmers’ demographic characteristics, with the exception of the quality of roads between the farm and the village. A Spearman’s correlation was run to test the relationship between road quality and degree centrality, finding a statistically significant negative relationship between degree centrality and road quality; r2 = -0.2249, p = 0.0024.9 This finding is surprising; it would be expected that farmers with higher road quality would maintain more market relationships due to the comparable ease of accessing their farms over those farmers with poor road access. It may instead suggest that farmers with lower road quality compensate for their limited access by maintaining more outside options to increase their bargaining power.  5.6.3 Strength of ties The nature of the relationship between market actors beyond market interactions (e.g. family members, neighbors, village contacts, or external market contacts) is used to measure the strength of ties between the farmer and the trader. Values are assigned to the different types of relationships farmers hold with their traders, and the strength of a tie is assumed to increase from external market contact to village contact, from village contact to neighbor, and from neighbor to family member. Strength of ties is used to test Hypothesis 3: Trust is built through familiar market relationships. Farmers leverage the trust built through longstanding and intimate trade relationships when bargaining the price paid for their goods to obtain a higher price on average.  The distribution of tie strengths between farmers and traders varies by village (see Figure 13).                                                  9 To test this relationship further, an ordered logit regression was performed to test the effect of road quality on degree centrality while controlling for village. This confirmed the significance of the relationship at the 1% level, with a coefficient of -0.41.   82 Figure 13: Distribution of tie strength by village   In Pondok Meja and Sungai Bertam, the majority of traders maintain relationships with market contacts only (56% and 51% respectively). In Bukit Murau, on the other hand, the largest share of market relationships is among traders from the same village (41%), followed by neighbors (39%), and market contacts (15%). Family and friends make up the smallest share of market contacts in Sungai Bertam (14%) and Bukit Murau (12%), and the second smallest in Pondok Meja (10%). This variability in tie strength between villages suggests there is a positive relationship between tie strength and distance to the commercial center. A Spearman’s correlation was run to test the relationship between distance to the commercial centre and tie strength, finding a strong positive relationship between these two factors; r2 = 0.2726, p = 0.0002. This relationship implies that farmers in more distant markets maintain trading relationships with stronger ties to compensate for lower competition (as determined by the relationship between network centralization and village distance).  Correlations run on tie strength and other production and demographic characteristics did not reveal any significant relationships, with the exception of the age of the farmer. A statistically Pondok Meja Sungai BertamBukit MurauMarket only Same villageNeighbour Family/friend 83 significant positive relationship was observed between tie strength and age of the farmers (see Figure 14). A Pearson correlation between these two factors shows r = 0.1239, p = 0.0984. This suggests that older farmers maintain closer ties than younger farmers.  Figure 14: Distribution of relationship type by age    5.7 Econometric results The base model includes only standard economic predictors of bargained price outcomes – product quality (approximated by frequency of harvest), farm size, road quality, age and gender. To control for differences between villages, most importantly transportation costs incurred by the trader, a village identifier which measures the distance (measured in hours) to reach the village from the region market center, Kota Jambi is included in the model.  The base model used to predict average rubber prices received by farmers over the previous year, using standard economic measures, takes the following form: 𝑝H = 	  𝛽, + 	  𝛽0𝑄H +	  𝛽D𝑆H − 𝛽%TH + 𝛽M𝐺H + 𝛽O𝑉HQ + 𝑒H  20406080ageMarket only Village contact Neighbour Family 84 Where:  𝑃H          = Price received by farmer f for rubber QH     = Quality of rubber sold by farmer f  𝑆H	  	    = Size of rubber production produced by farmer f TH     = Road quality to reach farmer f 𝐺H     = Gender of farmer f     𝑉HQ   = Village distance for farmer f trading in village g Equation 12 The second model introduces network centralization,  𝐶HQ, to the base model model to observe the effect of structural market power in each village, g, on each individual farmer, f. The third model introduces farmers’ degree centrality, 𝐷H, and the fourth model introduces farmers’ strength of ties with their traders,  𝑆𝑇H. Model 5 introduces an interaction term between degree centrality and tie strength, 𝐷𝑥𝑆𝑇H  to capture the combined effect of these two social network factors and to test whether tie strength has a different effect at different levels of degree centrality (expanded on in Subsection 5.4.5).  The complete model is therefore: 𝑝H = 	  𝛽, + 	  𝛽0𝑄H +	  𝛽D𝑆H − 𝛽%TH + 𝛽M𝐺H + 𝛽O𝑉HQ + 𝛽W𝐶HQ + 𝛽X𝐷H +	  𝛽Y𝑆𝑇H +	  𝛽,Z𝐷𝑥𝑆𝑇H + 𝑒H  Equation 13 5.7.1 Base model  Results from the base model (summarized in table 9) show road quality, gender and village distance to be significant predictors of bargained prices.    85 Table 9: Base model and model 2 (network centralization)              Gender shows a significant negative effect on average prices received, with women receiving an average price 368.70 Rp/kg lower than men. After distance to reach the village, gender shows the largest effect on prices received by the farmer. Although the correlation analysis in Subsection 4.5.2 found no significant relationship between gender and price received on its own, this finding suggests that when controlling for other factors, gender does have a significant impact on the price paid to farmers.  Road quality within villages shows a positive effect on average prices received with each additional level of perceived road quality resulting in an additional 153.32 Rp/kg. This means that farmers ranking their road quality at the maximum level report an additional 613.28 Rp/kg when compared to farmers that rank their road quality at the lowest level.     Base Model Model 2  Harvest frequency   0.22 (0.30)  0.58 (0.78) Farm size  46.16 51.21   (1.09) (1.21) Road quality  153.32* 181.95*   (1.57) (1.86) Gender  -368.70* -367.62*   (-1.79) (-1.80) Distance  -382.84*** -278.35***   (-9.13) (-4.15) Network centralization   -134.73**                            (-1.99) Constant  14,083.41*** 14,618.78***   (36.35) (31.13)  R-squared  0.3289 0.3426 Adjusted R-Squared  0.3113 0.3218 No. of observations  196 196 Note 1: t-statistic in ( )      Note 2: *p<0.1 ** p<0.05; *** p<0.01  86 Village distance shows the largest effect on price outcomes, with each additional hour of travel distance reducing the average price paid to farmers by 382.84 Rp/kg. As noted above, this effect is presumed to be largely explained by transport costs to carry their goods to the regional trading center, Jambi City.  The residuals of the model have been examined to confirm that they do not violate the assumptions of linear regression and no outliers were observed.  5.7.2 Model 2 – network centralization The second model incorporates the measure of market level competition – network centralization – to the base model. The results are summarized in Table 9.  The same three factors remain statistically significant, with the one notable difference being that the effect of village distance decreases from -382.84Rp/kg to - 278.35Rp/kg in the second model. Given that these two factors are highly correlated – as shown in Subsection 4.6.1, this in not surprising. This change in effect size indicates that network centralization is now capturing some of the village level effect that was being measured by the distance variable in the base model. However, network centralization does show to be statistically significant, with a 1% increase in network centralization leading to a 134.73Rp/kg decrease in average price paid to the farmer.  It is not possible to verify the independent of effect of network centralization separate from village distance, and it is likely that these two factors are directly related as more distant markets are likely to receive fewer buyers. The sample of villages is also too small to draw any significant conclusions from comparisons between village distances. However, once the individual level social network measures are added to the model in the subsequent Sub-sections, it should become clearer that network centralization does play a role in determining average price paid to farmers alongside village distance.   By adding network centralization to the model, the adjusted R2 increases from 0.3113 to 0.3218. Although a crude measure of model fit, this slight increase in the adjusted R2 supports network centralization’s inclusion in the model.  87 The residuals of the model have been examined to confirm that they do not violate the assumptions of linear regression and no outliers were observed. This was done by plotting the residuals against the model’s fitted values to observe any correlation between the residuals and fitted values or any noticeable outliers. This procedure is followed for each subsequent regression in this chapter.  5.7.3 Model 3 – degree centrality and outside options The third model summarized in Table 10 includes the estimated degree centrality for each farmer.  Table 10: Model 3 (degree centrality)   Base Model Model 3  Harvest frequency   0.22 (0.30)  0.58 (0.78) Farm size  46.16 49.50   (1.09) (1.15) Road quality  153.32* 185.62*   (1.57) (1.87) Gender  -368.70* -373.02*   (-1.79) (-1.81) Distance  -382.84*** -276.81***   (-9.13) (-4.15) Network centralization   -136.44**                           (-1.99) Degree Centrality   6.69    (0.24) Constant  14,083.41*** 14,598.52***   (36.35) (30.52)  R-squared  0.3289 0.3428 Adjusted R-Squared  0.3113 0.3184 No. of observations  196 196 Note 1: t-statistic in ( )      Note 2: *p<0.1 ** p<0.05; *** p<0.01   88 The number of trading relationships a farmer maintains does not show a significant effect on the average price farmers receive on its own. As will be shown below, it is not until all social network measures are added to the model that degree centrality shows to be statistically significant. As with model 2, road quality, gender, village distance and network centralization are statistically significant. There is a slight increase in the effect size of each of these factors with the exception of village distance, which see a slight decrease in effect size.   The residuals of the model have been examined to confirm that they do not violate the assumptions of linear regression and no outliers were observed.   5.7.4 Model 4 –tie strength  Model 4 introduces the measure of tie strength to the model and is summarized in Table 11.  Table 11: Model 4 (strength of ties)   Base Model Model 4 Harvest frequency  0.22 (0.30) 0.56 (0.76) Farm size  46.16 48.58   (1.09) (1.13) Road quality  153.32* 181.79*   (1.57) (1.83) Gender  -368.70* -360.60*   (-1.79) (-1.74) Distance  -382.84*** -288.35***   (-9.13) (-4.17) Centralization   -130.66*    (-1.90) Degree centrality   7.87    (0.28) Tie strength   3.11    (0.80) _cons  14,083.41*** 14,517.20***   (36.35) (29.66)  R-squared  0.3289 0.3451 Adjusted R-Squared  0.3113 0.3170 No. of observations  196 196 Note 1: t-statistic in ( )       Note 2: *p<0.1 ** p<0.05; *** p<0.01  89 As with degree centrality in model 3, tie strength on its own does not show to be statistically significant. The same set of factors remain statistically significant with slight increases in effect size for road quality and distance and slight decreases in effect size for road quality and network centralization.  The residuals of the model have been examined to confirm that they do not violate the assumptions of linear regression and no outliers were observed.   5.7.5 Model 5 – combined effect of degree centrality and tie strength I add an interaction term to the model to test whether different levels of degree centrality and trust in traders have different effects on the price received by farmers. Looking back at the simulation results from chapter 3 we see that the combined effect of increasing N and T leads to an increase in E(α). Adding an interaction term to the model captures this combined effect.  When the interaction term is added to the model (Model 5, Table 12), both degree centrality and relationship type become significant. The estimate for degree centrality shows that for each additional trader, the average price received by a farmer increases by 83.97Rp/kg. The estimate for tie strength shows that each additional level of tie strength leads to an increase of 11.29 Rp/kg in the average price received by the farmer. This indicates that a farmer bargaining with a trader from their village receives 11.29Rp/kg more than a farmer bargaining with a market contact from outside the village; a farmer trading with a neighbor receives 22.58Rp/kg more than a farmer trading with a market contact from outside the village; and a farmer trading with a family member or friend receives 33.87Rp/kg more than a farmer bargaining with a market contact from outside the village.       90 Table 12: Model 5 (interaction between degree centrality and tie strength)   Base Model Model 5 Harvest frequency  0.22 (0.30) 0.61 (0.83) Farm size  46.16 42.64   (1.09) (1.00) Road quality  153.32* 174.95*   (1.57) (1.83) Gender  -368.70* -314.01   (-1.79) (-1.52) Distance  -382.84*** -290.15***   (-9.13) (-4.23) Centralization   -146.36**    (-2.13) Degree centrality   83.97*    (1.79) Tie strength   11.29**    (2.02) Centrality / Tie strength interaction   -2.95**                         (-2.02) _cons  14,083.41*** 14,438.58***   (36.35) (29.65)  R-squared  0.3289 0.3592 Adjusted R-Squared  0.3113 0.3282 No. of observations  196 196         Note 2: *p<0.1 ** p<0.05; *** p<0.01  The effect size of these two factors is small compared to the effects of other factors in the model. However, their statistical significance in the model is indicative a relationship between these social factors and the prices received by farmers.  It is possible that more advanced social network measures could capture more of the social effects of bargaining that are not being reflected by these measures.  The fact that these measures were only statistically significant once the interaction term was added to the model is an important finding. This suggests that different relationship types have different effects on prices received depending on how many relationships a farmer maintains relative to other actors in the network.  This may be an indication that search and maintenance costs for these relationships affect the prices received by farmers and that some combinations of the number of Note 1: t-statistic in ( )  91 traders and their tie strength may be better than others, however it is beyond the scope of this thesis to explore this effect further. See below on limitations and suggestions for further inquiry. Network centralization remains significant in this model, and its effect size is the largest among the three models that include it with farmers receiving 146.36Rp less for each additional 1% of network centralization. The statistical significant of this estimate also increases from the 90% level in the previous models to 95% in this model. It appears that the inclusion of each of the social network measures developed in this thesis increases their specificity in the model. This finding supports the proposition made above that with more advanced social network analysis measures, it may be possible to improve the explanatory power of these estimates. It is also worthy of note that the statistical significance of gender has fallen below the 90% level in model 5. With each model iteration, the effect of gender gradually decreased, and once all three social network measures are added to the model and the interaction between degree centrality and tie strength is incorporated, the significance of gender has fallen below the threshold. It was noted in Sub-section 4.5.2 that gender did not appear to have a relationship with any of the variables of interest, however in the base model, the effect of gender was relatively large and significant at the 90% level.  Given that the statistical significance of gender only dropped slightly (it was previously just above the 90% threshold) caution should be taken in drawing any definitive conclusions, but it is interesting to note that as social network factors were added to the model, the significance of gender gradually decreased. This would suggest that degree centrality or tie strength may be used by women to compensate for gender bias in negotiations that would otherwise lead them to receive lower prices.  Finally, it is worthy of note that the adjusted R2 of model 5, at 0.3283, is the highest of the five models. As previously noted, this is only a crude measure of model fit, but the slight increase in this estimate (which control for the degrees of freedom lost by the addition of more variables) suggests that the inclusion of these social network measure lead to an improved model specification compared to the other four models.  The residuals of the model have been examined to confirm that they do not violate the assumptions of linear regression and no outliers were observed.    92 5.8 Limitations and suggestions for further inquiry 5.8.1 Instruments The specific social network measures applied in this model were designed with agricultural markets in Jambi in mind, therefore interpretations of the results are limited to that context. In particular, the nature of social and economic relations was taken into consideration when setting the values for the strength of ties measure. In other settings these particular categories might be entirely different as might the values applied to them. While the model can be transferred to other contexts, the measures themselves would need to be adapted.  The effect size of the individual social network measures was shown to be small. This may be due to the basic nature of these measures, in that the entirety of social network factors is not being captured by these measures. Testing of additional social network measures could improve the specificity of these measures and uncover further social effects that remain uncaptured my these estimates. The model does not directly measure search costs, as described in the conceptual model in chapter 3. Instead search costs are implied through the number of traders that farmers maintained and measured by degree centrality. Adding a direct measure of search costs to the model would likely improve the specificity of the model by capturing the trade-off between searching for more traders and the costs of searching for them.  Likewise, the model does not include a direct measure of the cost of maintaining strong ties. As noted in chapter 2, Coleman (1990) and Jackson (2010) argue that the maintenance of trusted relationships requires the investment of time and may include costly favours. The model might be further improved by adding a measure of maintenance costs to better distinguish the optimal relationship type and duration of trade relationship.  5.8.2 Endogeneity There are two possible concerns of endogeneity in this analysis: reverse causality and omitted variable bias  93 Perhaps the most important limitation to be aware of with this analysis is the possibility that the social network measures in the model are endogenous to the dependent variable, average price received by farmers. In other words, it is possible that the price received by farmers increases their opportunities for search and investment in trusted relationship maintenance– given that they are likely to have more resources to invest in these social factors. Given that search and maintenance costs were not measured in the survey it is not possible to control for these factors to determine whether there is a problem with endogeneity.  I would argue however that search costs in this context are likely to be quite low given that the trading network in each village is relatively diffuse (meaning there are a number of traders available) and the population density in the area would suggest that the physical costs of search (such as transport) would be relatively low.  With regards to maintenance costs, this could explain the lack of distinction between 3 of the 4 relationship-type categories observed. However, family relationship was the one significant category among the 4 and I would argue that the marginal costs of investing in this type of relationship are likely to be lower given that family relations can be maintained through a variety of structures, not just the market. For example, trust could be formed through marriages, reputations within families spanning generations, or through the increased threat of repercussion from malfeasance. Any added resource gained through higher prices received by family members is unlikely to have a significant impact of the ability of the farmer to invest in the maintenance of family trading relationships. 5.9 Conclusions  This empirical study has provided support for the two hypotheses set out in this thesis. In the context of rubber farming in Jambi, this study finds a positive relationship between the size of a farmer’s network (degree centrality) and the prices that they receive. It also finds as a positive relationship between the strength of trading relationships that a farmer maintains (strength of ties) and the prices they receive – specifically in the case of trade with friends or family members.   94 Having drawn the conceptual links between degree centrality and price information asymmetry in chapter 3, it is possible to conclude that in the province of Jambi, farmers with smaller networks are less able to access information about the prevailing market price when compared to farmers who maintain a larger number of trading relationships. This supports the hypothesis that information asymmetry offers a better opportunity for the trader to realize gains from arbitrage. Following the conceptual links drawn between strength of ties, social capital and risk mitigation, it is also possible to conclude that farmers in Jambi who trade with close contacts are able to leverage their social capital to increase the prices they receive. This supports the hypothesis that trading with familiar market contacts minimizes the risk of malfeasance as well as the likelihood that the trader will take advantage of gains from arbitrage. Although these conclusions are drawn from a limited sample in Jambi and cannot be generalized beyond that context, this analysis has shown that it is possible to measure the effect of social relations in a market setting while applying a classic economic approach. Furthermore, the models presented in this analysis demonstrate that failing to capture social network factors in market analysis risks omitting key factors in the determination of prices at the farmgate. The implication is that policies that do not account for these social network factors may overlook important determinants of prices paid to farmers.    95 Chapter 6: Conclusion  This thesis set out to improve our understanding of the social factors which determine smallholder farmers’ bargaining constraints. The goal of this research is to help inform policies aimed at mitigating market failures faced by famers which fail to incorporate the effect of social factors in their analyses and therefore overlook an important dimension of the bargaining process and how it affects farmers’ welfare.  This chapter revisits this research aim and demonstrates how the evidence presented in this thesis has responded to the research questions set out at the beginning of this thesis. Section 6.2. situates the research findings and conclusions in light of current knowledge in the field, while Section 6.3 comments on the contribution of this thesis to the fields of economics and sociology. Section 6.4 discusses the strengths and limitation of the research. Section 6.5 Discusses the potential application of the research findings, which possible applications in agricultural development research and policy more broadly, but also in the context of Jambi. Finally, Section 6.7 discusses possible future research directions based on the findings and limitations of this thesis.  6.1 Research aims Three overarching questions have guided the research undertaken by this thesis.  (1)  Can concepts developed in the field of social network analysis be used to improve our understanding of bargaining in agricultural markets?  A review of the literature presented in Chapter 2 outlined the emergence of the field of social network analysis and demonstrated the linkages between the aims of this field of inquiry to those of neoclassical economics. It was shown that social network analysis can be used to respond to some of the challenges of conceptualising and measuring social factors in economics, with a focus on the transmission of social capital through economic exchange. The findings in this chapter have shown that the fields of economics and sociology are complimentary, and that an interdisciplinary approach incorporating principles developed both disciplines can be used to advance our methods of understanding bargaining in agricultural markets.  96 This interdisciplinary approach was developed further in Chapter 3, where well-known concepts related to bargaining from the field of economics were linked to well-known concepts from the field of social network analysis. The relationship between the concepts of market competition and network centralization, individual degree centrality and outside options and trust and informal contract enforcement was shown in order to demonstrate that these concepts share common foundations, and through their combination, can be used to understand both the economic and social factors that underlie bargaining in agricultural markets. A formal bargaining model was presented, providing a strong theoretical framework from which to interpret the relationship between these economic and social factors and to demonstrate how we might predict the effect of these factors on farmers’ bargaining outcomes.  (2)  How does the structure of a market and a farmer's position within it affect the level of competition in the market and, by extension, the prices they receive for their goods? The model presented in Chapter 3 demonstrated that the density and structural composition of the market can lead to differentiated outcomes for actors in the market based on their ability to access buying options. Specifically, it was shown that centralization in a network – leading to unequal access to buying options among sellers – can lead to monopsonistic competition. The result is that buyers are able to exude some degree of control over the prices paid to sellers in the market, thereby capturing additional surplus.  This proposition was supported in the empirical analysis by testing the following hypothesis: Hypothesis 1: Farmers have unequal access to buying options when there is imperfect competition in the market. Controlling for market centralization shows that farmers with a higher degree centrality earn a higher price for their goods on average. The empirical results confirmed that network centralization has a negative effect on the prices received by farmers overall, with the implication that less central farmers are disadvantaged in the bargaining process given their limited access to available buying options. This finding was confirmed in the next stage of the analysis which examined individual level outcomes.  Recognising that market competition underlies the individual-level bargaining capacities of market actors, the thesis went on to show how farmer’s outside options within this competitive structure affect their ability to bargain with available buyers. Using the social network concept of degree  97 centrality, it was shown that an actor whose position in the network is highly central has greater outside options and that these can be leveraged to bargain a higher price.  This proposition was supported by testing the following hypothesis in Chapter 5:  Hypothesis 2: Farmers with a relatively higher degree of centrality have greater outside options. Greater outside options lead to increased bargaining power due to the threat of rejecting a low offer. Farmers with higher degree centrality therefore earn a higher price for their goods on average. The empirical analysis showed that on its own, degree centrality did not demonstrate a significant effect on the price received by rubber farmers in Jambi, but in combination with strength of ties, higher degree centrality led to higher prices received by farmers on average.  (3)   Does the strength of the relationships a farmer maintains with their traders affect their bargaining power and, by extension, the prices they receive for their goods? The model developed in Chapter 3 crucially showed that in addition to outside options, trust between the buyer and the seller can be leveraged to negotiate a higher price paid to the seller. Specifically, the share of the bargaining surplus going to the seller is increased when there is trust between them and the buyer. This finding is supported by the literature on informal contract enforcement discussed in Chapter 2, which shows that community and individual level informal enforcement increases the likelihood of mutually beneficial bargaining outcomes. This proposition was also supported in the empirical analysis by testing the following hypothesis: Hypothesis 3: Trust is built through familiar market relationships. Farmers leverage the trust built through longstanding and intimate trade relationships when bargaining the price paid for their goods to obtain a higher price on average.  The empirical analysis showed that when controlling for market competition (network centralization) and individual outside options (degree centrality), the familiarity of the relationship between the seller and the buyer is a significant determinant of the price paid to the farmer. Specifically, farmers trading with more familiar market contacts such as family members, friends or neighbours negotiate a higher price on average.   98 6.2 Situating the findings in the field Social effects on economic outcomes have been recognised as a gap in research, and the increasing attention paid to social capital theory in the last few decades is a testament to that. This thesis is situated in the discussion of how to integrate social effects into neoclassical economic approaches. While the role that certain manifestations of social capital play in market exchange has widely been acknowledged, there is not yet a widely accepted measure of social capital that is suited to sit alongside other market measures in economic models. The application of social network analysis is presented in this thesis as one approach that can be used alongside economic measures to understand the bargaining process that market actors undertake. The application of these measures to the context of bargaining in agricultural markets can also be situated in the field of agricultural marketing, with a specific focus on access to markets among smallholder famers in developing country contexts. The importance of social factors in market exchange in such contexts has been long understood, and a number of studies cited in this thesis have set out to measure these effects empirically. The intention of this thesis has been to anchor the evidence of social factors in agricultural market exchange in a strong theoretical framework that is easily translated across disciplines. New measures of social capital are proposed which are rooted in both social network anlaysis and economic principles, and new evidence is presented on role that social factors play in agricultural markets. Conclusions drawn from that evidence are limited to the context under investigation, but lessons can be drawn from the approach to measurement itself.  6.3 Research contributions This thesis has contributed to firstly to the field of interdisciplinary research by showing how one common approach to a line of inquiry can be devised by developing a strong theoretical framework rooted in the basic principles of two disciplines. Although the fields of economics and sociology are well understood to be complimentary, and traditions exists that cut across both of these disciplines, still the language, concepts and measurements used in these disciplines leads them to diverge in many ways.   99 A key challenge that this thesis set out to overcome was to find a unifying way of approaching a research problem that spoke to both of these traditions. The concepts presented in Chapter 3 required a great deal of testing from both disciplines to ensure that these held up to the conventions established in each field, and this exercise itself is a demonstration of how to approach a research problem in a wholly integrated way.  This exercise in integration has also led to contributions to each discipline in its own right. In economics, while social factors have long been seen as playing a role in economic interactions, it has been a challenge to measure these factors directly using existing methods. The social network analysis measures presented in this thesis have been shown to be complimentary to existing bargaining measures in the field of economics. By establishing this bridge between economics and social network analysis, it may be possible to extend this merger further into other areas of economic inquiry where there is a desire to incorporate social effects.  Similarly, this exercise has shown that sociological concepts such as social capital can be integrated into a formal economic model and empirically tested using econometric methods. While this thesis is not the first to attempt this, this in an under-investigated area in sociology. Similar to the challenges faced by economists, sociologists often struggle to present their findings in way that is suited to economic modelling while maintain the level of complexity that can be observed in social relations. The field of social network analysis has been shown to be a way of translating findings between these two disciplines. Finally, the empirical analysis put forward in this thesis is a unique contribution in its own right. The data set collected for this thesis has provided an evidence base to support the lines of inquiry discussed in this section. Furthermore, the survey which was designed to respond to the research questions posed by this thesis shows how one might approach the empirical measurement of social factors in economic exchange. This approach could be adapted and refined to study these factors in other contexts.   100 6.4 Strengths and limitations The interdisciplinary approach of this thesis, and the equal weight it has given to the fundamental principles in both economics and sociology is one of the most significant strength of this thesis. By doing so, the approach can be adapted by practitioners in either discipline. The simplicity of concepts used from both disciplines also facilitates the accessibility of this approach and was designed with practical applications in policy and program design in mind.  One limitation, which stems from the simplicity sought in the design of this integrated approach, is that the conceptual model and empirical instruments developed in this thesis do not offer significant extensions of existing approaches in their respective disciplines. For example, the econometric procedures used in this thesis are limited, with emphasis placed on integrating new variables to an otherwise basic econometric design. Likewise, the social network measures applied in this thesis are basic ones, with the emphasis placed on adapting their interpretation into an economic framework rather than developing the measures further in their own right. A further limitation is that the specific social network measures applied in the empirical model were designed with agricultural markets in Jambi in mind, therefore interpretations of these results are limited to that context. As noted throughout the thesis, the nature of social and economic relations is highly context dependent, and this must be taken into account when designing instruments to measure social factors in economic exchange in different settings and in different types of markets.  6.5 Research applications It has been shown that market imperfections within the trade of agricultural goods tend to restrict market competition and leave farmers reliant on informal mechanisms to ensure a beneficial outcome from price negotiations. Supporting evidence has been provided which confirms that these factors often have a negative impact on smallholder farmers’ bargaining power, with the lower prices they receive having a negative impact on their welfare.  101 Local, national and international level policies have been designed to help smallholder farmers overcome these market imperfections and improve rural welfare. To determine their approaches, various organizations and governments have introduced market analyses to their rural development programing. However, these analyses tend to be limited to price, trade flows and transactions costs (the latter only when measureable, and then only imperfectly) so that even the most comprehensive market analyses for agricultural development are still limited in their scope (Barrett, 1996). Furthermore, these analyses tend to examine only market level prices and aggregate transaction costs, meaning that farm-level characteristics are not accounted for.  With a limited analytical scope, there is room left for misrepresentations of market data and a failure to account for farmers’ abilities to effectively access identified markets, to negotiate within them, and to confront the risks they present. In some cases, this misinformation may worsen the situation for smallholder farmers where the outcome of such analyses divert them from existing marketing opportunities that were in fact better suited to the position of their firm in the broader market structure.  Taking the case of Jambi as an example, recent policies implemented by both the Indonesian and Jambi Governments have encouraged smallholder farmers to convert from rubber to palm oil production. In 2000, the governor of Jambi announced plans to develop 1 million hectares of palm oil plantations within five years. This push toward palm oil production is based on the fact that the international price for natural rubber has fallen, while the global demand and international price for palm oil has been increasing (Feintrenie, Chong, & Levang, 2010).  The analysis presented in this thesis encourages these policy-makers to examine the existing market structure of palm oil and farmers’ access to available buying options within it before encouraging all smallholders into a new market that is for the most part unknown to them. Given that the market structure and farmers’ positions within it has been shown to affect the prices farmers’ received for rubber in this area, it would be important to understand how this new market structure might affect them.  Equally, given that the nature of market relationships has been shown to have an effect on the prices that farmers receive, it would important to understand whether forgoing their  102 longstanding relationships in the rubber market would be compensated through higher prices in the palm oil market. It may be that the lost surplus through bargaining with untrusted palm oil traders would be greater than the increased surplus earned by converting from low priced rubber to high priced palm oil.  Agricultural value chains are expanding and smallholder farmers are being encouraged, attracted or absorbed into new trading relationships. The nature of this expansion means the introduction of new trade actors, new contractual instruments and new forms of market interaction for many smallholders. Where access to different marketing channels exist, it is important that smallholder farmers are aware of the trade-offs between each so as to minimize transaction costs and risk while maximizing the prices that they receive. An improved understanding of the implications of these new trade relationships that incorporates social factors which determine the bargaining power of farmers will help to design policies and programmes better suited to the opportunities and constraints that these farmers face.    103 Bibliography Afrizal, J. (2013). Cartels control staple foods in Jambi, BI says. Jakarta Post. Accessed November 27, 2013, from http://www.thejakartapost.com/news/2013/02/07/cartels-control-staple-foods-jambi-bi-says.html Aker, J.& Fafchamps, M. (2013). Mobile Phone Coverage and Producer Markets: Evidence from West Afrcia. CSAE Working Paper WPS/2013 09. Oxford: Centre for the Study of African Economies. Aker, J. & Ksoll, C. (2012). Information Technology and Farm Households in Niger. WP 2012-005: February 2012. United Nations Development Programme Regional Bureau for Africa. Andriani, L., & Karyampas, D. (2009). A New Proxy of Social Capital and the Economic Performance across the Italian Regions (No. BWPEF 0903). Attoh, C., Martey, E., Kwadzo, G. T. M., Etwire, P. M., & Wiredu, A. N. (2014). Can Farmers Receive Their Expected Seasonal Tomato Price in Ghana? A Probit Regression Analysis. Sustainable Agriculture Research, 3(2), 16–23.  Arias, P., Hallam, D., Krivonos, E., & Morrison, J. (2013). Smallholder integration in changing food markets. Rome: FAO. Arifin, B. (2005). Supply-chain of Natural Rubber in Indonesia. Jurnal Manajemen & Agribisnis, 2(1).  Arrow, K. (1999). Observations on Social Capital. In P. Dasgupta & I. Serageldin (Eds.), Social Capital: A Multifacteted Perspective. Washington: The World Bank. Badan Informasi Geospasial (2016). Atlas Provinsi Sumbar, Jambi dan Bengkulu.Accessed August 13, 2016 from http://www.bakosurtanal.go.id/atlas-provinsi-sumbar-jambi-dan-bengkulu Bandiera, O., & Rasul, I. (2006). Social Networks and Technology Adoption in Northern Mozambique. The Economic Journal, 116, 869–902. Banerjee, A., Chandrasekhar, A., Duflo, E., & Jackson, M. O. (2013). The Diffusion of Microfinance. Science, 341(6144). Barrett, C. B. (1996). Market Analysis Methods: Are our Enriched Toolkits Well Suited to Enlivenevd Markets? American Journal of Agricultural Economics, 78(3), 825–829. Barrett, C. & Mutambatsere, E. (2008). "Agricultural Markets in Developing Countries" in Blume, L., Durlauf, S. (eds.) The New Palgrave Dictionary of Economics, 2nd Edition. London: Palgrave Macmillan.  104 Becker, G. & Murphy, K. (2000). Social Economics: market behavoir in a social environment. The President and Fellows of Harvard College. Beckert, J. (2007). The Great Transformation of Embeddedness: Karl Polanyi and the New Economic Sociology (No. 07/1). Benhabib, J., Bisin, A., & Jackson, M.O. (2010). Social economics: brief introduction to the handbook, in: Handbook of Social Economics, Vol. 1A. North Holland: Amsterdam, xvii-xxi. Berg, J. & Dickhaut, J. & McCabe, K. (1994). Trust, Reciprocity, and Social History. Gaems and Economic Behaviour, 10, 122-142). Bijman, J., Ton, G., & Meijernik, G. (2007). Empowering Small holder Farmers in Markets. (ESFIM Working Paper 1). Wageningen. Binswager, H. P., & Quizon, J. B. (1986). What Can Agriculture Do for the Poorest Rural Groups? (No. 57). Washington, D.C. Bissonette, J. & De Koninck, R. (2015). Large Plantations versus Smallholdings in Southeast Asia: Hisotircal and Contemporary Trends. BRICS Initiative for Critical Agrarian Studies.  Borgatti, S. * Everett, M. (1997). Network analysis of 2-mode data. Social Networks, 19, 243-269. Bourdieu, P. (1986). The Forms of Capital. In Handbook of Theory and Research for the Sociology of Education (pp. 241–258). Westport, CT: Greenwood. Bourguignon, F., Silva, L. P., & Stern, N. (2002). Evaluating the Poverty Impact of Economic Policies: Some Analytical Challenges.Washington, D.C. Badan Pusat Statistik Indonesia. (2012). Statistical Yearbook of Indonesia 2012. Accessed November 30, 2013, from http://www.bps.go.id/eng/aboutus.php?pub=1&pubs=47 Badan Pusat Statistik Indonesia. (2013a). Population of Indonesia by Province 1971, 1980, 1990, 1995 , 2000 and 2010. Population Statistics. Accessed November 29, 2013, from http://www.bps.go.id/eng/tab_sub/view.php?kat=1&tabel=1&daftar=1&id_subyek=12&notab=1 Badan Pusat Statistik Indonesia. (2013b). Number and Percentage of Poor People, Poverty Line, Poverty Gap Index, Poverty Severity Index by Province, September 2012. Accessed November 28, 2013, from http://www.bps.go.id/eng/tab_sub/view.php?kat=1&tabel=1&daftar=1&id_subyek=23&notab=1  105 Badan Pusat Statistik Provinsi Jambi. (2013a). Produk Domestik REgional Bruto. Accessed November 30, 2013 from http://jambi.bps.go.id/pub/fb/2013/pdrblapus2012/index.html  Badan Pusat Statistik Provinsi Jambi. (2013b). Jambi in Figures 2013. Accessed November 30, 2013, from http://jambi.bps.go.id/pub/fb/2013/jda2012/index.html  Badan Pusat Statistik Provinsi Jambi. (2015). Jambi in Figures 2015. Accessed August 13, 2016 from http://jambi.bps.go.id/website/pdf_publikasi/Jambi-Dalam-Angka-2015.pdf Bernard, H., Killworth, P. & L. Sailer (1982). Informant Accuracy in Social-Network Data V: An Experimental Attempt to Predcit Actual Communication from Recall Data. Social Science Research, 11: 30-66.  Brandes, U., & Fleischer, D. (2005). Centrality measures based on current flow. Stacs 2005, 533–544.  Buchan, N. R., Croson, R. T. a., & Dawes, R. M. (2002). Swift Neighbors and Persistent Strangers: A Cross-Cultural Investigation of Trust and Reciprocity in Social Exchange. American Journal of Sociology, 108(1), 168–206.  Burt, R. S. (1995). Structural Holes: the social structure of competition. Harvard University Press. Burt, R. S. (2000). The network structure of social capital. Research in Organizational Behavior, 22, 345–423.  Buskens, V. (2002) Social Network Analysis and Game Theory: Basic Concepts and Assumptions. In Buskens, V. (ed.), Social Networks and Trust. Dordrecht: Kluwer Academic Publishers,. 31-52.  Cervantes-Godoy, D., Kimura, S. & Anton, J. (2013). Smallholder Risk Management in Developing Countries. OECD Food, Agriculture and Fisheries Papers No. 61. Paris: OECD. Chamberlin, J. & Jayne, T. (2012). Unpacking the meaning of 'Market Access': Evidence from Rual Kenya. World development, 41, 245-264. Chambers, Robert, and Gordon R. Conway. “Sustainable Rural Livelihoods: Practical Concepts for the 21st Century.” IDS Discussion Paper. Brighton, 1991. Chandrasekhar, A. G., Kinnan, C., & Larreguy, H. (2015). Social networks as contract enforcement: Evidence from a lab experiment in the field. Boston. Coleman, J. (1958). Relational Analyis: The Study of Social Organizations with Survey Methods. Human Organization, 17(4), 28-36.  Coleman, J. (1990). Foundation of Social Theory. Harvard University Press.  106 Conley, T. G., & Udry, C. (2001). Social Learning through Networks: The Adoption of New Agricultural Technologies in Ghana. American Journal of Agricultural Economics, 83(3), 668–673. Courtois, P., & Subervie, J. (2014). Farmer Bargaining Power And Market Information Services. American Journal of Agricultural Economics, 1–25.  Cox, D., & Fafchamps, M. (2008). Extended Family and Kinship Networks: Economic Insights and Evolutionary Directions. In P. Schultz & J. Strauss (Eds.), Handbook of Development Economics (4th ed., Vol. 44, pp. 0–109). Elsvier. Cribb, R., & Brown, C. (1995). Modern Indonesia: A History since 1945. London: Longman Press. Crisp, J. (2001). Mind the Gap! UNHCR, Humanitarian Assistance and the Development Process. International Migration Review, 35(1), 168–191. Cross, J. G. (1965). A Theory of the Bargaining Process. American Economic Review, 55, 67–94. Cunyat, A. (1998). Strategic Negotiations and Outside Options. Daulay, A. R. (2011). Regional Autonomy and Sustainable Development in Indonesia: The case study of oil palm development in Jambi province. The University of Queensland. Accessed July 14, 2016 from http://asnelly69.wordpress.com/2011/04/21/regional-autonomy-and-sustainable-development-in-indonesia-the-case-study-of-oil-palm-development-in-jambi-province/. David, M. & Sutton, C. (2004). Social Research: The Basics. Sage Publications Ltd: London. Dercon, S., & Krishnan, P. (2000). In Sickness and in Health: Risk Sharing within Households in Rural Ethiopia. The Journal of Political Economy, 108(4), 688–727. De Vries, D. & Sutarti, N. (2006). Gender Equity: revealing the reality of Jambi's women. Governance Brief 29b: 1-7. CIFOR, Bogor, Indonesia.  Drakeley, S. (2005). The History of Indoensia. Greenwood Press, Westport, CT.  Durlauf, S. N. (2002). On The Empirics Of Social Capital. The Economic Journal, 112(483), F459–F479. Ely, J. & Valimaki, J. (2003). Bad Reputation. The Quarterly Journal of Economics. 118(3), 785-814. Fafchamps, M. (1999). Rural poverty, risk and development (No. 144).  107 Fafchamps, M. (2006). Social Capital and Development. Journal of Development Studies, 42(7), 1180–98. Fafchamps, M. (2010). Vulnerability, Risk Management, and Agricultural Development. African Journal of Agricultural Economics, 5(1), 1–29. Fafchamps, M., & Gabre-madhin, E. (2006). Agricultural Markets in Benin and Malawi. African Journal of Agricultural and Resource Economics, 1(1). Fafchamps, M., Gabre-madhin, E., & Minten, B. (2005). Increasing Returns and Market Effciency in Agricultural Trade. Journal of Development Economics, 78(2), 717–734. Fafchamps, M., & Gubert, F. (2007). Risk Sharing and Network Formation. American Economic Review Papers and Proceedings, 97(2), 75–79. Fafchamps, M., & Hill, R. V. (2008). Price Transmission and Trader Entry in Domestic Commodity Markets. Economic Development and Cultural Change, 56(4), 729–766. Fafchamps, M., & Lund, S. (2003). Risk-sharing networks in rural Philippines. Journal of Development Economics, 71(2), 261–287.  Fafchamps, M., & Minten, B. (2002). Returns to Social Network Capital Among Traders. Oxford Economic Papers, 54, 173–206. Fafchamps, M., & Minten, B. (2012). Impact of SMS-based agricultural information on Indian farmers. World Bank Economic Review, 26(3), 383–414.  Fathoni, Z. (2009). Evaluation of Market System and Market Integration for Rubber Cultivation in Jambi Province - Indonesia. Wageningen University and Research. Feintrenie, L., Chong, W. K., & Levang, P. (2010). Why do farmers prefer oil palm? Lessons learnt from Bungo district, Indonesia. Small-Scale Forestry, 9(3), 379–396. Freeman, L. C. (1977). A Set of Measures of Centrality based on Betweeness. Sociometry, 40, 35–41. Freeman, L. C. (1979). Centrality in Social Networks: Conceptual clarification. Social Networks, 1, 215–239. Frenzen, J., & Nakamoto, K. (1993). Structure, Cooperation, and the Flow of Market Information. Journal of Consumer Research, 20(3), 360–375. Fukuyama, F. (2001). Social capital, civil society and development. Third World Quarterly, 22(1), 7 – 20.   108 Gabre-madhin, E. Z. (2001). Market Institutions, transaction costs, and social capital in the Ethiopian grain market. Washington, D.C.: International Food Policy Research Institute. Ghezzi, S., & Mingione, E. (2007). Embeddedness, Path Dependency and Social Institutions: An Economic Sociology Approach. Current Sociology, 55(1), 11–23.  Goyal, A. (2010). Information, Direct Access to Farmers, and Rural Market Performance in Central India. American Economic Journal: Applied Economics. 2 (July 2010), 22-45.  Granovetter, M. (1973). The Strength of Weak Ties. American Journal of Sociology, 78(6), 1360–1380. Granovetter, M. (1985). Economic Action and Social Structure: The Problem of Embeddedness. American Journal of Sociology, 91(3), 481–510. Greif, A. (1993). 1993 Greif AER 1993.pdf. The American Economic Review, 83(3), 525–548. Grootaert, C. (1999). Social Capital, Household Welfare and Poverty in Indonesia. World Bank Policy Research Working Paper No. 2148. Grootaert, C., Oh, G.-T., & Swamy, A. (2002). Social Capital , Household Welfare and Poverty in Burkina Faso. Journal of African Economies, 11(1), 4–38. Gulati, A., Minot, N., Delgado, C., & Bora, S. (2005). Growth in high-value agriculture in Asia and the emergence of vertical links with farmers. Paper presented and the Symposium Toward High-Value Agriculture and Vertical Coordination: Implizations for Agribusiness and Smallholders. National Agricultural Science Centre, Pusa, New Delhi, India, 7 March 2005.  Gunter, B. G., Taylor, L., & Yeldan, E. (2005). Analysing Macro-Poverty Linkages of External Liberalisation: Gaps , Achievements and Alternatives, 23(3), 285–298. Hanneman, R., & Riddle, M. (2005). Introduction to Social Network Methods. Riverside CA.: University of California, Riverside. Hadi, P. U., & Budhi, G. S. (1997). Analysis of the Economic Efficiency and Comparative Advantage of the Sumatran Smallholder Rubber Using “PAM” Method.  Hamilton, Kirk, and Gang Liu. “Human Capital, Tangible Wealth, and the Intangible Capital Residual.” Policy Research Working Paper, 2013. Hammouda, H. & Osakwe, P. (2008). Global Trade Models and Economic Policy Analysis: Relevance, Risks and Repurcussions for Africa. Development Policy Review, 26 (2), 151-170.   109 Handcock, M. S., & Gile, K. J. (2008). Modeling Social Networks from Sampled Data. Annals of Applied Statistics. Hardaker, J. ., Hurine, R. B. M., Anderson, J. R., & Lien, G. (2004). Coping with Risk in Agriculture. Wallingford: CABI Publishing. Hinrichs, C Clare. Embeddedness and Local Food Systems: Notes on Two Types of Direct Agricultural Markets. Journal of Rural Studies, 16 (2000): 295–303. Holland, P. & S. Leinhardt (1973). The structural implications of measurement error in sociometry. The Journal of Mathematical Sociology, 3 (1): 85-111.  Ifeoma, O. & Mthitwa, H. (2015). An Analysis of the Impact of the Use of Mobile Communication Technologies by Farmers in Zimbabwe. A Case Study of Esoko and EcoFarmer Platforms. Proceedings of SIG GlobDev 2015 Pre-ECIS Workshop: Munster, Germany, May 26, 2015. International Fund for Agriculture and Development. (2010). Rural Poverty Report 2011, New realities, new challenges: new opportunities for tomorrow’s generation. Rome: International Fund for Agriculture and Development Jackson, M. O. (2008). Social and Economic Networks. Princetion: Princeton University Press. Jackson, M. O. (2009). Networks and Economic Behavior. Annual Review of Economics, 1(1), 489–511.  Jackson, M. O. (2010). An Overview of Social Networks and Economic. In J. Benhabib, A. Bisin, & M. O. Jackson (Eds.), Handbook of Social Economics (pp. 511–585). North Holland Press. Jamison, D. T., & Lau, L. J. (1982). Farmer Education and Farm Efficiency. Washington, D.C.: The World Bank. Kähkönen, S. & Meagher, P. (1997). Opportunism Knocks? Legal institutions, contracting, and economic performance in Africa. USAID/EAGER Project. Working Paper No. 204.  Kandpal, E., & Baylis, K. (2013). Expanding horizons: Can women’s support groups diversify peer networks in rural India? American Journal of Agricultural Economics, 95(2), 360–367.  Kandpal, E. & K. Baylis, K. (2016). The Social Lives of Married Women: Peer Effects in Female Autonomy and Children’s Food Consumption.”  Working paper (2015).  Kandori, M. (1992). Social Norms and Community Enforcement. Review of Economic Studies, 59, 63-80.  110 Knack, S., & Keefer, P. (1997). Does social capital have an economic payoff? a cross-country investigation. Quarterly Journal of Economics, 112(4), 1251–1288. Kopp, T.; Alamsyah, Z., Fatricia, R. & Brummer, B. (2014). Have Indonesian Rubber Processors Formed a Cartel? Analysis of Intertemporal Marketing Margin Manipulation. Paper prepared for presentation at the EAAE 2014 Congress. Kopp, T. & Brummer, B. (2015). Traders and Credit Constrained Farmers: Market Power along Indonesian Rubber Value Chains. Paper prepared for the International Conference of Agriculrual Economists, Milan: 2015.  Krackhardt, D. (2003). The Strength of Strong Ties. Chapter 3 in Cross, R., Parker, A. & Sasson, L. (eds.). Networks in the Knowledge Economy. Oxford University Press, Oxford.  Kraev, E., & Akolgo, B. (2005). Assessing Modelling Approaches to the Distributional Effects of Macroeconomic Policy. Development Policy Review, 23(3), 299–312.  Labonne, J., & Chase, R. S. (2009). The Power of Information The Impact of Mobile Phones on Farmers’ Welfare in the Philippines, (July), 26. Accessed March 14 2014 from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1435202 Lerman, K., & Ghosh, R. (2010). Information Contagion: an Empirical Study of the Spread of News on Digg and Twitter Social Networks. In Proceedings of the Fourth International AAAI Conference on Weblogs and Social Med (pp. 90–97).  Levang, P., Yoza, B. K., & Tasman, A. (1999). In the shadow of rubber. Jakarta: Institut de recherce pour le development. Lin, N. (1999). Building a Network Theory of Social Capital ’. Connections, 22(1), 28–51. Lüdeke, M., Petschel-Held, G. & Schellnhuber, H. (2004) Syndromes of Global Change: The First Panoramic View. GAIA 13(1), 41-49.  Maertens, a., & Barrett, C. B. (2012). Measuring Social Networks’ Effects on Agricultural Technology Adoption. American Journal of Agricultural Economics, 95(2), 353–359.  Maluccio, J., Haddad, L., & May, J. (2000). Maluccio Haddad May South Africa.pdf. Journal of Development Studies, 36(6), 54–81. Manski, C. F. (2000). Economic Analysis of Social Interactions (NBER Working Paper Series). Cambridge, MA. Mavridis, D. (2014). Ethnic Diversity and Social Capital in Indonesia. World Development, 67, 376–395.  111 Mcintosh, C., Gertler, P. & Falcao, L. (2015). Barriers to market access for smallholder farmers. International Growth Centre: http://www.theigc.org/project/barriers-to-market-access-for-smallholder-farmers/ Mendoza, R. & Thelen, N. (2008). Innovation to Make Markets More Inclusive for the Poor. Development Policy Review, 24 (4), 427-458. Miguel, E. A., Gertler, P., & Levine, D. I. (2003). Did Industrialization Destroy Social Capital in Indonesia (No. C03-131). Minten, B., & Fafchamps, M. (2001). Social Capital and the Firm: Evidence from Agricultural Trade. American Journal of Agricultural Economics, 83(3), 65–98. Mochalova, A., & Nanopoulos, A. (2013). On the role of centrality in information diffusion in social networks. European Conference on Information Systems, 1–12. Montgomery, M. & Casterline, J. (1996). Social Learning, Social Influence, and New Models of Fertility. Population and Development Review, 22, 151-175.  Narayan, D., & Pritchett, L. (1999). Cents and Sociability  : Household Income and Social Capital in Rural Tanzania. Economic Development and Cultural Change, 47(4), 871–897. Nerlove, M. (1956). Estimate of the Elasticities of Supply of Selected Agricultural Commodities. American Journal of Agricultural Economics, 38 (2), 496-509. Nugroho, A. S. (2013). Evaluation of Transmigration (transmigrasi) in Indonesia: Changes in socioeconomic status, community health and environmental qualities of two specific migrant populations. Kagoshima University. OECD (2002). Glossary of Industrial and Organisation Econoimcs and Competition Law. Compiled by R. S. Khemani, & Shapiro, D. M, commissioned by the Directorate for Financial, Fiscal and Enterprise Affairs, OECD, 1993. Peramune, M. & A. Budiman (2007). A Value Chain Assessment of the Rubber Industry in Indonesia. U.S. Agency for International Development. Petroczi, A., Nepsz, T & F. Bazso (2007). Measuring tie-strength in virtual social networks. Connections, 27(2), 39-52. Polinomics (2010). Salop's Circular Theory. Accessed August 29, 2016, from http://www.policonomics.com/salops-circular-city/ Provinsi Jambi. (2012). Nilai Tukar Petani dan Inflasi Perdesaan. Accessed November 23, 2013, from http://jambi.bps.go.id/pub/fb/2013/ntp 2012/  112 Putnam, R. D. (2000). Bowling Alone: The Collapse and Revival of American Community. New York: Simon & Schuster. Rapsomanikis, George, David Hallam, and Piero Conforti. “Market Integration and Price Transmission in Selected Food and Cash Crop Market of Developing Countries: Review and Applications.” In Agricultural Commodity Market and Trade: New Approaches to Analyzing Market Structure and Instability, edited by Alexander Sarris and David Hallam, 187–219. Edward Elgar Publishing, 2006. Ravallion, M. (1986). Testing Market Integration. American Journal of Agricultural Economics, 68(102-109). Reardon, T., Barrett, C., Berdegue, J. & Swinned, J. (2009.). Agrifood Industry transformation and Small Farmers in Developing Countries. World Development, 37(11).  Robinson, L. J., Schmid, A. A., & Siles, M. E. (2002). Is Social Capital Really Capital? Review of Social Economy, 60(1), 1–21. Rola, A. C., Jamias, S. B., & Quizon, J. B. (2002). Do Farmer Field School Graduates Retain and Share What They Learn? An Investigation in Iloilo, Philippines. Journal Lof International Agricultural and Extension Education, 9(1), 65–76. Rothenberg, R. B. (1995). Commentary: Sampling in Social Networks. Connections, 18(1), 104–110. Rubinstein, A. (1982). Perfect Equilibrium in a Bargaining Model. Econometrica, 50(1), 97–109. Salami, A., Kamara, A. & Brixiova, Z. (2010). Smallholder Agriculture in East Africa: Trends, Constraints and Opportunities. African Development Bank Group Working Paper Series. No. 105, April 2010.  Schultz, T. (1979). The Economics of Being Poor. In Lecture to the memory of Alfred Nobel. Accessed July 2, 2013, from http://www.nobelprize.org/nobel_prizes/economics/laureates/1979/schultz-lecture.html Scott, J. (1988). Social Network Analysis. Sociology, 22(1), 109–127.  Shepherd, A. W. (1997). Market information services: theory and practice. FAO Agricultural Services Bulletin, 125. Silvey, R., & Elmhirst, R. (2003). Engendering Social Capital: Women Workers and Rural-Urban Network in Indoensia’s Crisis. World Development, 31(5), 865–879. Songsermsawas, T., Baylis, K., Chhatre, A. & Michelson, H.  (2016) Can Peers Improve Agricultural Revenue? World Development, 83(July), 163-178.  113 Sorensen, B. C. (2000). Social Capital and Rural Development: A Discussion of Issues (No. 10). Washington, D.C. Stifel, D. C., & Thorbecke, E. (2003). A dual-dual CGE model of an archetype African economy: trade reform, migration and poverty. Journal of Policy Modeling, 25(3), 207–235. Stigler, G. (1961). The Economics of Information. The Journal of Political Economy, 69(3), 213-225.  Stolle, F., Chomitz, K. M., Lambin, E. F., & Tomich, T. P. (2003). Land use and vegetation fires in Jambi Province, Sumatra, Indonesia. Forest Ecology and Management, 179(1-3), 277–292.  The World Bank. (2008). Agriculture and Poverty Reduction. Agriculture for Development Policy Brief. Townsend, R. M. (1995). An Evaluation of Risk-Bearing Systems in low-income Economies. The Journal of Economic Perspectives, 9(3), 83–102. Trading Econoimcs & Tokyo Commodity Exchange (2016). Rubber. Accessed August 2, 2016, from http://www.tradingeconomics.com/commodity/rubber Turcoy, T. & von Stengel, B. (2001). Game Theory. CDAM Research Report LDE-CDAM-2001-09.  Uzzi, B. (1997). Social Structure and Competition in Interfirm Networks: The Paradox of Embeddedness. Administrative Science Quarterly, 42(1), 35–67. Volij, O., & Winter, E. (2002). On risk aversion and bargaining outcomes.pdf. Games and Economic Bahviour, 41, 120–140. Von Maltitz, G., & Stafford, W. (2011). Assessing opportunities and constraints for biofuel development in sub-Saharan Africa (No. 58). Bogor, Indonesia. Wasserman, S., & Faust, K. (1994). Social Network Analysis. Cambridge: Cambridge University Press. Welch, F. (1970). Education in Production. Journal of Political Economy, 78(1), 35.  Wetterberg, A. (2007). Crisis, Connections, and Class: How Social Ties Affect Household Welfare. World Development, 35(4), 585–606.  White, Alice P. “A Note on Market Structure Measures and the Characteristics of Markets That They ‘Measure.’” Southern Economic Journal 49, no. 2 (1982): 542–49.  114 Woolcock, M. (1998). Social capital and economic development: Toward a theoretical synthesis and policy framework. Theory and Society, 27(2), 151–208. Woolcock, M., & Narayan, D. (2000). Social Capital: Implications for Development Theory, Research , and Policy. World Bank Research Observer, 15(2). Zhang, Y., Wang, L., Duan, Y. (2016) Agricultural information dissemination using ICTs: A review and analysis of information dissemination models in China. Information Processing in Agriculture, 3(1), 17-29.    115  Appendix 1: Survey The following survey has been translated from the original Indonesian. Name  Gender  Place of origin  Reason for moving here?  Religion  Education level  Age   Production Information 1. How many hectares did you harvest in total last year?  2. Of this land, how much do you own? 3. Of this land, how much do you rent? 4. Please list all of the crops that you produce for sale ______________________________________________________________________________________________________________________________________________________ 5. What are your main crops for sale?  Crops  No. of Hectares What was the lowest and highest price you received last year (Rp) How often do you harvest (Days) How much do you harvest per ha (Kg) 5.1   Rp.                   Rp.   5.2  Rp.                   Rp.   5.3  Rp.                   Rp.   5.4  Rp.                   Rp.   5.5  Rp.                   Rp.   6. Do you have any other business, work or source of income?  Business/work Wage (daily/weekly/monthly) (Rp)    116   7. Farm expenditures Paid labour Number of people  Rp/month                                      How many hours do you and your family work (unpaid)             hrs/day   days/week Seeds 1. 2. 3. 4. kg/year                                        Rp/kg Fertilizer 1. 2. 3. 4.                 kg/year Rp/kg Equipment    Rp/year Rent   Rp/year Transportation    Rp/month Feed   Rp/month   8. Does your family consume any of these crops  9. If yes, how much  Crop Consumption per month 9.1                 Kg 9.2  Kg  9.3  Kg   Yes No  117 10.1 Did you encounter any obstacles last season (such as pest infestation, flooding or drought, shortage of labour)?   10.2 If yes, what proportion of your harvest was lost?  11.1 Do you have a loan or credit related to your agricultural production? 11.2 If yes, what type?  Bank  Friend/Family  Trader Cooperative  Other: __________________________   Social Network 12. How many traders do you know?                                                     13. Please name the traders you sell to most often?  Name and product  How do you know them (family/friend/same village, from the market only)? Name: Product:  Name: Product:  Name: Product:  Name: Product:  Name: Product:  Yes No  118     14. How often do these traders buy your crops? Name Frequency  Day they come by How long have you been selling to them (years)? 1.     2.     3.     4.      5.      15. What is the average price you received from the trader last year? Name Crop Average price 1.    2.    3.    4.    5.   16. Do you have a contract or formal agreement with the trader? Name Contract? Contract upheld? Reason for breaking contract 1.    Yes    No     Yes    No     119 2.    Yes    No     Yes    No    3.    Yes    No     Yes    No    4.    Yes    No     Yes    No    5.    Yes    No     Yes    No    17. Do you believe that the trader purchases your products at a fair market price? Please rate your level of confidence from 1 (don’t trust at all) to 5 (fully trust). Name Don’t    Somewhat   Adequate   High       Fully  Trust          Trust         Trust      Trust       Trust       1              2              3                4               5         1.     2.   3.   4.   5.     18. Do you expect that this trader will contribute to buy from you? Please rate your level of confidence in whether they will continue to buy from you from 1 (don’t trust at all) to 5 (fully trust).  Name Don’t    Somewhat   Adequate   High       Fully  Trust          Trust         Trust      Trust       Trust       1              2              3               4               5  1.        2   3   4   5    120  19. What is the condition of the road to reach your farm?  Road condition Very      Poor     Adequate    Good      Very  poor                                                        good       1              2              3              4             5            Information 20. Please list up five people that you talk to about prices or production (farmers or buyers)   Name Farmer or trader? Frequency you discuss 1.                           2.    3.    4.    5.      

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:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0340488/manifest

Comment

Related Items