Smallholders Coping with Food Sector Transformation

This chapter provides a description of the data and research methodology used in this study. This study employed investigates the smallholders coping with food sector transformation and the role of social enterprises in Thailand. Both qualitative and quantitative data were carefully planned and collected in order to facilitate the achievement of the objectives set for the study. An overview of the stages of data collection, data discussion, scope of the study, sample and case studies selection, and methods of analysis is provided. Qualitative data was used for making a general mapping and characterization of the agri-food value chain in Thailand. This information provided essential input for designing a survey for collecting quantitative data. Furthermore, qualitative information will be very useful for interpreting quantitative results and complementing them. While quantitative data is analyzed using statistical and multivariate data analysis techniques, such as Probit regression, gross margins regression with selectivity bias solution, and Factor Analysis. All these quantitative methods are summarized and discussed in this chapter according to the research objectives and questions.

5.2 Overview of stages of data collection

The different stages of data collection were divided into four steps as follows: First stage, prior to leaving to the field in Thailand, a thorough review of secondary information was undertaken. It consisted of the review of information from international and Thailand sources, including a literature review of previous studies. This stage consisted of interviewing key informants of government institutions, NGOs, private organizations, farmer organizations, universities and donors. The semi-structured interviews explored the information on how agri-food value chains are changing, with special reference to smallholder coping with food sector transformation and the role of social enterprises. The objective of this activity was also determining the role of institutions and organizations in helping small-scale farmers in modern trade supply chains, including the involvement of institutions and organizations in the production and marketing, their current and future plans and limitations. This information was essential for designing pre-structured interviews used in the following stage of the study. Second stage, the main key actors in agri-food supply chain such as supermarkets, suppliers/brokers, wholesalers, Green Net cooperative and the Royal Project were interviewed by using semi-structured questionnaires (please see the questionnaires in Appendix II-1) in order to gather information about their marketing activities, procurement systems, contract farming systems, and marketing development, especially the current and development of participating with modern trade chains. These interviews help to identify the nature and level of participation of smallholders in modern trade chains. It also provided information about market opportunities and threats offered to small-scale farmers by modern trade markets, the constraints faced by farmers trying to enter such supply chains, and key success factors associated with small-scale farmers that are able to gain access. Furthermore, market observation were undertaken by the author and research assistances in top supermarkets and wholesale markets in Bangkok and main cities such as Chiang Mai, Chiang Rai, Khon Kaen, Chachoengsao. As a third stage, in-depth interviews and focus groups were undertaken with farmers participating with social enterprises in producing for modern trade markets and farmers selling in the traditional markets. Interviews encompassed the factors that facilitate of impeded their participation with social enterprise in producing for modern trade markets. Interviewing participant and non-participant farmers enabled the exploration of the motivations, benefits and key success factors of these two groups, and the barriers to entry faced by non-participants. This stage provided information to develop the quantitative questionnaires in stage four. The fourth stage, step in quantitative survey, of data collection was consisted of surveying a sample of participating and comparable non-participating farmers in order to collect quantitative information about the issues addressed in the in-depth interviews after the survey instrument was pre-tested in two different rounds with sets of respondents (see more detail about questionnaire and pre-test in section xxx). This stage provides quantitative data on the benefits, constraints faced by farmers and key success factors of farmers. Both qualitative and quantitative information contribute to achieving the objectives and research questions of the study. 5.3 Data discussion Regarding to stages of data collection, it was divided into two main phases according to type of analyses and objectives of the study. First phase aimed at collecting qualitative (secondary information and semi-structured interviews), and was carried out during the period April 2008 – July 2008. In addition, it aimed at market observation from top supermarkets in main cites, and was carried out during the period April 2008 - September 2008. The phase of the surveys investigated the following issues according to the research objectives and questions: Phase I: Qualitative Analysis: Assessing forces and trends in the restructuring of Agri-food value chain of modern trade and small-scale producers in Thailand. 1: To analyse the changing value chain and transformation in the agri-food industry of small-scale horticulture producers and modern trade in Thailand. RQ1 What is the current agri-food supply chain of small-scale producers and modern trade markets in Thailand? RQ2 What are forces and trends driving the restructuring of agri-food value chain and food sector transformation in Thailand? RQ3 What is the role of social enterprise in linking small-scale farmers into modern trade chains? 2: Outline the terms under which small-scale producers interact with modern trade. RQ 4 Do they have proper contracts, or are the transactions more informal and ad-hoc, etc? RQ 5 Do contracts change over time?, and why? RQ 6 How suppliers/producers adjust to/bargain with modern trade? This phase (regarding to objective 1 and 2) refers to the patterns of changes in the different modern food supply chains in Thailand and the impact of policy on these changes, including the role of social enterprises in linking small-scale farmers to the modern trade chains. The study reviewed grey literature and interviewed with key informants from many key informants and institutions industry ‘middlemen’- wholesalers, buyers, social enterprises and cooperative managers, the World Vegetable Center- AVRDC, government institutions, NGOs, and Universities (see Appendix II-3). The general of this phase is to make a general characterization of agri-food value chain in Thailand, make a map of the system and collect inputs for the questionnaire design in the second phase. Second phase aimed at collecting quantitative data (survey), and carry out during June 2008 – November 2008 for the Royal Project case study, and September - December 2009 for Green Net case study. Phase II: Survey and Quantitative Analysis: Assessing motives for and impacts of small-scale producers’ participating with social enterprises in producing for modern trade markets for selected products. This phase based on two case studies (Green Net, and the Royal Project) producer surveys which include both participants and non-participants with modern trade contract farming. The surveys investigated the following issues according to the research objectives and questions: 3: Analyses the motivation and challenges of participation of small-scale producers in producing for modern trade chains in Thailand. RQ7 What are the determinants of small-scale farmers’ participation with social enterprise in producing for modern trade chains? RQ8 What are advantage and challenges faced by small-scale farmers participating with social enterprise in producing for modern trade market markets in Thailand? 4: Estimating the impacts on small-scale farmers participating with social enterprises in producing for modern trade. RQ 9 Do producers participating with social enterprise in producing for modern trade obtain better outcomes (profits/income) compared to non-participant? RQ 10 Do producers participating with social enterprise gain other non-financial benefits from participation? This phase (regarding to objective 3 and 4), the study involved field survey and collect the data from questionnaires (please see scope of the study in the next section and Appendix II-2). The objective of this phase is to collect quantitative information about variables that determine the participation of small-scale farmers in modern trade chains. 5.4 Scope of the study and survey 5.4.1Qualitative Information The qualitative phase was carried out during the period April - July 2008, consisting of in-depth interviews with different actors involved in the agri-food sector, including government, private sectors (eg., modern trade and supermarkets, suppliers/brokers, buyers and/or wholesalers), NGOs, farmer organizations, universities and farmers. (see Appendix II- list of key informants) A total of 20 in-depth interviews conducted with organizations and institutions, 20 with small-scale farmers that participate in the modern food chains and 20 with small-scale farmers that selling to traditional markets. A “Snowball”? Sampling used to contact research informants (see example Blandon, 2006 and Robson, 1993). Semi-structured interviews were differently designed for organizations and institutions, buyer/wholesalers, and farmers; participants and non-participants (Please see Appendix II-1). Each interview lasted between one hour and one hour and a half, and in the case of farmers most of the interviews were conducted on their farms. For each interview hand-writing notes and/or tape-record were taken for analyzing and reviews. For analysing the data, notes and tapes were reviewed. This analysis allowed the making of a map of the agri-food system and identifying categories related to the research objectives. 5.4.2Quantitative data collection The objective of the second phase was to collect quantitative information about variables that determine the participation of small-scale farmers with social enterprise in producing for modern trade markets. For this purpose, a survey including participant and non-participant farmers was carried out during June 2008 – November 2008 for the Royal Project case study and September - December 2009 for Green Net case study. The quantitative portion of the study based on small-scale producer analyses (vegetable and rice case studies) which supplying modern trade markets as well as international markets. The survey questionnaires administered from face-to-face interviews 240 (120 for each case study) small-scale farmers excluding pilot tests. This stage that some attempt made to follow a random sampling procedure, but field reality may result in convenience based sampling. Since the qualitative data collection precedes and feeds into the quantitative data collection, I first outline the schedule for qualitative data collection and then show how this transition into the quantitative part (see Table 5.1). Table 5.1 - Steps in quantitative survey data collection

Task

Description of task

1 Information gathering and preparation of materials needed for the in-depth interviews 2 Pre-test: Semi-structured interviews, focus group 3 In-depth interviews (semi-structured interviews) conducted on a group of key informants 4 In-depth interviews (semi-structured interviews) conducted on a group of suppliers/buyers/wholesalers 5 In-depth interviews (semi-structured interviews) conducted on a group of small-scale farmers 6 Initial Analysis of results from in-depth interviews and comparisons. This will provide inputs to development of the quantitative part, commencing in step 7 below. 7 Development of initial questionnaire for pre-testing 8 Pre-test of questionnaire on a sample of participant farmers 9 Pre-test of questionnaire on a sample of non-participant farmers 10 Evaluation of questionnaire for possible corrections and re-designing of questionnaire 11 Final administration of questionnaire on sample of A participant farmers 12 Review of answers given to ensure clarity and uniformity, at this stage, new information revealed will be added to the questionnaire. 13 Final administration of questionnaire on sample of non-participants 14 Review of answers given to ensure clarity and uniformity, at this stage, new information revealed will be added to the questionnaire 15 Revisit[1] of farmers by the same interviewer to clarify ambiguous issues and ensure consistency [1] Repeat visit will be done if there are any data problems.

5.4.3 Administration of questionnaire and survey:

The questionnaire was developed following five steps suggested by Aaker et al. (1998) and Masakure (2005). Step one involved planning what to measure (revisiting research questions, focusing on research issues and getting additional data from secondary and exploratory research). Step two entailed formatting the questionnaire (determining the content of questions, the framing for each question). Step three involved consideration of question wording (evaluating each question according to how respondents would comprehend and their ability to answer). Step four involved sequencing and layout decisions (ordering of questions to create a single questionnaire). Finally, step five involved pre-testing the questionnaire and correcting problems. The entire design was guided by the in in-depth interviews and as shown in steps in quantitative survey data collection (see Table 3). It is important to know that the research supervisory visited in Thailand also involved in these steps for evaluation of questionnaire for possible corrections and re-designing of questionnaire. The questionnaire divided into five sections. Section one of the questionnaire contained questions relating to Basic household information. Section two collected data on farm characteristics including costs and returns. Section three focused on income and assets of household. Section four assessed the marketing details and the factor of choosing markets. Section five focused on the history and experience of growing. A copy is provided in Appendix II-2. In questionnaire different scales of measurement were used, such as nominal, ordinal, interval and ratio scales. A number of multi-item scales were included in the questionnaire following Masakure (2005) and Oppenheim (1992). It is important to note that multi-item scales are widely used in marketing research to measure phenomena that cannot be captured directory with one attitude-based question (Masakure, 2005). They are particularly useful when it is not possible to rely on behaviour as an indication of phenomena (Oppenheim, 1992). For example, in this study a multi-item scales was used to measure the perceptions of small-scale farmers on participating with social enterprise in producing for modern trade markets as opposed to using a single item. The advantages of multi-item scales is that specificity of items can be averaged when all items are combined and the researcher is able to make a clear distinction between individuals and factors through combining these items (Masakure, 2005). The final questionnaire was administered by personal interviews, after the completion of each pilot questionnaire, from................. through ....... . Two trained research assistances from Mae Fah Luang University and Chiang Mai University were recruited and helped the author in this task. The author carried out half of the interviews. Respondents were selected using a two-stage stratified sampling procedure based on the number of centres operated and the geographical areas of operation. To aid the interview process, research assistants first met farmers at an informal gathering. Farmers were free to ask questions related to the research. Participant farmers were told that the information generated would be used sorely for academic purposes. Each interview began with a brief explanation of the research objectives and its purpose. Questionnaires were filled in by the interviewer. The process could be adjourned several times to enable farmers to undertake their normal duties. On completion the questionnaire, the respondent was thanked for their participation.

5.4.4 Case of the study

Secondary information from different sources (triangulation) and the information from in-depth interviews with key informants was a key to identify the case studies. The research design, data gathering, analysis and interpretation were based on two case studies approach. Case studies are better at investigating contemporary phenomenon within its real life context, especially when the boundaries between the phenomenon and context are not clear (Yin, 1994). The case study approaches are also powerful in combining qualitative and quantitative data and provide a description and test theory or even generate theory (Masakure, 2005). This study mostly concerned with the role of social enterprises assist smallholders with production and marketing for modern trade markets. Therefore, case studies are more useful in addressing the implications of the restructuring agri-food value chain on small-scale farmers (the relationship-returns, network and social aspects). In consequence, this study used two case studies as Green Net and the Royal Project. Below is a brief overview of the current case studies. 1. Green Net (rice case study) Having explained briefly about the Green Net, it is established in 1993 by the group of people wishing to support the environmental and social responsible business. In present, Green Net is one of the largest producers and wholesaler of organic food in Thailand. It also plays as important social enterprise in supporting sustainable development for a better livelihood of small-scale producers and consumers as well as a clean environment for Thailand. At present, there are over 20 product assortments (e.g. organic rice (majority product), vegetables, fruits, teas, cotton etc.) sold through approximately 40 retail outlets in Bangkok and around the country. Beside domestic market, Green Net Cooperative operates fair-trade exports to Europe and the nearby countries in Asia. Please see more detail about Green Net in Chapter 6, section 6.xx. Green Net is currently purchased from farmer groups in the North-eastern, Northern and Central regions of Thailand. Regarding organic rice, majority product, main organic rice price producers for Green Net are cooperatives in Yasothorn Province (North-eastern) , one of the largest organic rice provinces in Thailand, such as Naso Organic Rice Cooperative and Bakruea Organic Rice Cooperative. Therefore, this study is focused on these cooperatives’ members who are small-scale farmers growing organic Jasmine rice as participant farmers. 2. The Royal Project (vegetable case study) The Royal Project (RP) has been playing as an important social enterprise in developing and promoting quality of life for the highland small-scale farmers in various aspects. The RP also becomes one of important agri-food suppliers for both domestic and international markets especially for low-chemical and organic products. The RP has developed the household subsistence farming into the commercial-based production under Good Agricultural Practice (GAP) emphasizing on quality and safety standards in all links of the supply chain. Therefore, the RP, as it now becomes, has represented the linkages between the small-scale farmers and modern trade markets to help themselves in growing useful crops which enable them to have a better benefits. Please see more detail about the Royal Project in chapter 6, section xxx. The RP have 4 research centres and 37 Agricultural Development Centres within 5 provinces in the North of Thailand as Chiang Mai, Chiang Rai, Lamphoon, Phayao and Mae Hong Sorn. The Agricultural Development Centre plays as a main collaboration centre between farmers and the RP to support the RP production and marketing plans. There are 27 Agricultural Development Centre in Chiang Mai (72.79%), 6 centres in Chiang Rai (16.22%), 2 Mae Hong Sorn (5.1%), 1 centre in Payao (2.70%) and 1 centre in Lampoon (2.70). In addition, there are 26,174 household’s members, 257 villages from 5 provinces, under 37 The RPF’s development centres. Most farmers are in Chiang Mai (69.25%), Chiang Rai (16.27%), Lampoon (7.38%), Mea Hong Sorn (4.90%), and Phayao (1.51%) in ordered. Vegetable and fruit production are main income resources of the RP. In 2008, The RP had total income about 427.47 million Baht. The main income resource came from vegetables 56.29%, fruits and coffee 16.97% (coffee is about 4.61%), and flowers and trees 5.88%. Chinese cabbage is one of importance crops for the RP which is selling all the year to modern trade markets. Therefore, regarding to in-depth interviews and survey, this second case study is focused on the RP members who are Chinese cabbage producers in Chiang Mai province (Chiang Mai is one of the largest vegetable (cold weather vegetable) production areas in Thailand) as participant farmers. The participant famers come from 3 Agricultural Development Centres in Chiang Mai as Nong Hoi, Mae Hea, and ........... The following table (Table 4) shows the summary of scope of the study. Table 4: Summary of scope of the study Phase of study Analysis Data sources Phase 1 :Qualitative Trends Changes in retail points policy. Documents/interviews from key informants, organizations and other administrations. Change in producer and supplier points, and contract policy. Documents/interviews from key-informants from supermarkets, suppliers and farmers. Phase 2: Quantitative Producers/Suppliers’ access Small-scale producer analyses: - Royal Project: vegetable producer - Green Net: rice producer (Total is 240 farmers) Surveys of small-scale farmers who participate and non-participate with supermarkets, 60: 60 PER CASE STUDY

5. 5 Validity, Reliability and Practicability

Regarding to data collection and research methods, one important issue to take into account in this study is that measurement tools used meet the criteria of validity, reliability and practicability. A research instrument meets validity criterion if it is able to measure what is supposed to measure; reliability if it provides accurate and stable measurements; and practicability if it is appropriate according to economic, convenience, and interpretability (Blandon, 2006; Cooper and Emory, 1995 and Robson, 1994). The quantitative research is usually associated with the terms of validity and reliability. On the one hand, qualitative research usually looks for alternative ways to deal with credibility issues (Blandon, 2005). The above measurement criteria are very difficult to evaluate. However, according to Janesick, 2000 and Robson, 1994, an alternative and important tool is triangulation which refers to the use of several data sources and/or the use of multiple methods in the research. Therefore, in this research, qualitative data come from different sources (for example, supermarket, buyer, social enterprise, cooperative, NGOs, donor, farmer, public, private and academic research participants). In addition, several statistical and multivariate research techniques that facilitate methodological triangulation are used in the quantitative analyses. Furthermore, qualitative and quantitative results can mutually validate. Several authors have raised concern about the issue of selectivity bias, or self-selection bias (see example Ali and Abdulai, 2009, Mendola, 2007, Shankar and Thirtle, 2005), such as in this study farmers are not random assigned by the researcher into participation and non-participation groups, and instead self-select themselves into groups. There are many solutions to the selectivity problem, including Heckman correction models, panel data methods, etc. In this study, therefore, the propensity score matching method was used to be addressing this issue. Finally, the practicability is also taken into account, especially for the quantitative phase, which was conducted in two case studies. The case studies selected for the survey are important social enterprises, which are most important producers/suppliers for modern trade markets in Thailand. In addition, kind of products and the geographical selections were very practical in terms of budget and crop season limitations as well as in term of socio-economic conditions.

5.6 Methods of Analyses

The study has two parts of analysis methods, according to the objectives of the study, which organized into two empirical studies as the following; Value Chain Analysis; value chain mapping – in order to achieve the objective 1 and 2. Econometric and Statistic Processing of Survey data – in order to achieve the objective 3 and 4.

5.6.1 Value Chain Mapping

The value chain mapping helped to understand the patterns of agri-food supply chains, forces and trends driving the future food value chain and market development, including the role of social enterprises in helping smallholders with food retail transformation in Thailand. There are many different dimensions of Value Chain Analysis techniques (see example Kaplinsky and Morris, 2002 and Roduner, 2004). However, this research followed the value chain work plan as the following. Value chain work plan This study adopted an approach follow a simple practical methodology borrowed from Miles (2002) and Holtzman (2002) found in the World Bank’s online “Guide to Developing Agricultural Markets and Agro-enterprises”?.The development of a 9 steps work plan is the combination of these approaches that describing the sequence of efforts needed to construct a viable and representative value chain map for the selected case studies. A flow chart illustrating the overall procedure is depicted in Figure 4. Figure 4. Employed work methodology: chain construction and sector analysis. Part I Establish initial understanding of commodity subsetor Product selection Step 1 Review of existing literature & data Step 2 Preliminary interviews/fieldwork Step 3 Identification key issues & questionnaire design Step 4 Drawing of preliminary (Value Chain) map Step 5 Part II Refine map and subsector understanding Extensive fieldwork: interview of chain actors Step 6 Visiting of physical facilities & institutions Step 7 Quantification and refinement of map Step 8 Re-assessment of results by actors and map finalization Step 9 The value chain mapping helped to understand the patterns of agri-food supply chains, forces and trends driving the future food value chain and market development, including the role of social enterprises in helping smallholders with food retail transformation in Thailand. This methodology helped to understand the patterns of agri-food supply chains in Thailand. In addition, this part gained understanding of the patterns of agri-food supply chains in Thailand and business potential such as markets, inter-firm relationships, and critical constrains that limit small-scale farmers growth and industry competitiveness.

บอ�ว่าส่วนนี้จะได้อะไร�ล้วข้อมูลที่ได้จะเชื่อมโยง�ละมีประโยชน์ต่อ�ารวิเคราะห์เชิงปริมาณในส่วนที่สอง

5.6.2 Econometric and Statistic Processing of Survey data

Two different sets of econometric models were applied on collected survey data; Probit models of the determinants of smallholder participation with social enterprise in producing for modern trade markets. With data on participants and comparable non-participants in place, these models estimated the following kinds of relationships: Probability of Participation = f(demographic, socio-economic, attitudinal variables) Here, demographic information included variables such as age, family size, education of household head, etc; socio-economic variables may include income, experience in farming, farm size, etc; attitudinal variables included a small number of scale variables that attempt to proxy the smallholder’s welfare priorities, attitudes to risk, etc. Regression models of the determinants of farm economic outcomes (profits per rai). Profits per rai = f(farm variables, socio-economic and farmer variables, participation in modern trade chains). These helped determine the effect of participation with social enterprise in producing for modern trade chains on key farm outcomes, while controlling for other variables that may affect outcomes. A selectivity/endogeneity problem is recognized in such estimation equations, and ways to overcome such problems was explored during analysis stage. In addition, the statistic tool; Factor analysis, was applied on collected survey data; Factor Analysis of the motivation and potential benefits of participant farmers including problems faced by participant farmers. This analysis helped us to look at factor of the motivation and problems faced by participant farmers. The factor interpreted by identifying the variables that have a large loading on the same factor. These methods are described below.

5.6.2.1 Probit regression analysis.

In analysis of dependence when the dependent variable is discrete, choice or probability models are used. A particular dependent variable used in this research is the participation with social enterprise in producing for modern trade markets. Explanatory variables are used for determining the probability of the participation with social enterprise in producing for modern trade markets. Probit regression is associated with the estimation of the probability of participation (see example Lattin et al., 2003; Greene, 2000, and Blandon, 2007). To test the determinants of participation with social enterprise in production for modern trade, a probit model is estimated in which the dependent variable equals 1 if the farmer is participate with social enterprise in producing for modern trade markets, and zero otherwise: Yi* = β' Xi + ui , (1) where Y = 1 if Yi > 0, otherwise Y = 0, and Probability (Yi = 1) = Probability (ui > β' Xi) = 1 – F(– β' Xi), where F is the cumulative distribution function for u [1] . The β' are maximum likelihood estimates. For a description and discussion of the probit model, see, for example, Maddala, 1998, 22-27. The theoretical concept of the probit model application with a list of factors that were indentified from previous studies (see for example Braun, Hotchkiss and Immink, 1989 and Blandon, 2006) and the information provided by in-depth interviews and survey is the following: It is hypothesized that the choice to become a participant farmer was determined by the expected income increase, which can be assumed to be determined by the resource endowments of the farm (farm size, soil quality, land elevation, distance of farm from main road and distance of farm to market). In addition, income potentially earned non-farm determines the of opportunity cost of working on-farm. In the long-run, farmers are facing a choice of earning non-farm income versus on-farm work growing the labour intensive for modern trade crops, especially for organic farming. This choice is determined by the non-farm versus on-farm opportunity costs of family labour. Endowment of human capital and established non-farm employment opportunities determine these relationships for a specific household. It is further hypothesised that household labour force size and composition (women’s share) may be a factor for adoption. A higher share of women’s labour may enhance participation of the modern trade crop. Since the decision is mainly that of the male head of household, his age, education level, and experience in farm are other factors of hypothetical impact for the participation. Based on these hypotheses, the participation model is specified as follows: Participation = f (Hsize, Adult, Female, HHsex, Hhage, Hhedu, Expf, Froad, Fcoop, Fsize, Land, Qsoil, Hincome, Nfincome, NfinL, WealthIndex), where Hsize = house size (all members in the family), Labour = total labour available in the household (that is, persons of working age), Female = total female in the household Hhsex = sex of head of household (1 male, 2 female), Hhage = age of head of household (years), Hhedu = head of household education (years) Expf = head of household expericenc in farming (years), Froad = distance of farm from main road (km) Fcoop = distance of farm from cooperative (km) Fsize = farm size (rai). Land = land elevation (1very low land – 4 high) Qsoil = soil quality (1 very bad – 5 excellent) Hincome = yearly household income (Baht) Nfincome = non-farm income (%) NfinL = non-farm income from providing labour share (%) Wealth Index = calculated from house area and household’s assets e.g. car, bicycle and TV The empirical analysis was conducted using the STATA statistical package. The probit estimates can be used to derive linear probability of participating modern trade crop product, which can be approximated by dF/dx (marginal effect). Table 5.xx is shown a list of socio-economic, farm characteristic variables and attitudinal variables? that hypothetically determine small-scale farmers participation with social enterprise in the modern trade chain which has been used in probit regression. This original set of variables has been chosen considering the literature review and the information provided by in-depth interviews and survey. (see for example Braun, Hotchkiss and Immink, 1989 and Blandon, 2006) Further discussion about these variables presented in the results (please see chapter 7: Results 1 - Green Net, and Chapter 8: Results II -The Royal Project). The expected relationships of the explanatory variables and the probability of participating with social enterprise in modern trade chains are also presented in Table 5.xx.

Table 5.xx – Explanatory variables and descriptive statistics determining participation with

social enterprise in producing for modern trade.

Variables Description Expected Sign Type* Dependent Variable: Participation with Green Net (social enterprise) in production for modern trade 1=yes, 0 = no N/A

-

Independent variables: House size (members in the family) person

+

SE Total labour available in household person

?

SE Total females in house hold person

-

SE Sex of head of household Male/Female

?

SE Age of head of household Years

?

SE Education of head of household Years

+

SE Experience in farm of head of household Years

+

SE Distance of farm from main road km

-

FC Distance of farm from social enterprise km

-

FC Area of farm size rai

+

FC Land elevation 1 very low land – 4 high

+

FC Soil quality 1 very bad – 5 excellent

+

FC Yearly household income Baht

+

FC Proportion of non-farm income

%

-

FC Proportion of providing labour share

%

-

FC Wealth Index scale

+

AV Note: * SE stands for socio-economic; FC for farm characteristics, AC for attitudinal variables 5.6.2.2 Regression analysis (Gross Margins Regression) This study used the simple form multiple regression analysis finding the best predicted by a linear combination of the possible explanatory variables to explain how the variation in farm economic outcomes (or dependent) variable, Y, depends on the variation in a predictor (or independent or explanatory) variable, X. The general regression model is given by: Yi = β0 + βiXi + ui where the values of β0, βi are called the regression coefficients and are estimated from the study data called least squares, explained by Lomax [2] (1992). The regression models was employed to explore the determinants of farm economics outcomes which helped determine the effect of participation on key farm outcomes (profits per rai), while controlling for other variables that may affect outcomes. It is hypothesized that the farm outcomes (profits per rai) was determined by the farm variables, socio-economic, farmer variables, and participation with social enterprise in production for modern trade. Based on the hypothesized, the profits per rai regression model is specified as follow: Profits per rai = f(farm variables, socio-economic and farmer variables, and participation with GN) The empirical analysis was conducted using the STATA statistical package. In addition, a selectivity/endogeneity problem is recognized in such estimation equations, and ways to overcome such problems will be explored in the next section (section 5.6.2.3).

5.6.2.3 Selectivity bias

The above gross margin regression analysis explores the determinants of farm economics outcomes which help determine the effect of participation on key farm outcomes. There is however, one very important econometric issue with the gross margin regression which needs to be addressed. This is the issue of selectivity bias, or self-selection bias because farmers are not randomly assigned by the researcher into participation and non-participant groups, and instead self-select themselves into groups. In consequence, this profitability difference does not necessarily indicate that participant farmers have a positive impact on profits because it could be caused by selection bias. It is possible that more talented or more enterprising farmers tend to become members of Green Net. Since more talented farmers make higher profits than less talented farmers, it may appear that Green Net membership is increasing profits. In reality, it may be the higher underlying talent levels of participating farmers that creates extra profits. If this is true, then the regression coefficient of gross margins on participation would not really reflect the effect of participation, but rather the mix of the effects of participation and the underlying talent levels. There are many solutions to the selectivity problem, including Heckman correction methods, panel data method, etc. The method for this study is “propensity score matching”?method. The basic idea behind propensity score matching method is as follows. The probit model of participation produces a probability of participation for every observation in the sample, including participants and non-participants. This predicted probability (called the propensity score) is based on the observed values for the independent variables and the coefficient estimates from the probit model. In one version of propensity score matching, every participant will be compared to a non-participant based on similarity of propensity scores. Their outcomes will be compared, i.e, the difference between their gross margins will be computed. Once this is done for all participants, the differences will be averaged and reported as the average difference. This version is called ‘nearest neighbour’ matching (NNM). The intuition is that, controlling for the probability of participation, ie., comparisons of participants and non-participants with similar propensity scores, is similar to random assignment to control and treatment groups. There are other versions of propensity score matching. Another method is called ‘kernel-based’ matching (KBM). Here, the outcome of each participant is compared to the weighted average outcomes of all non-participants, where the weights depend on the probability of participation. The output will show a row called ‘ATT’, the average treatment effect on the treated. The value in this row shown as ‘difference’ is the average difference between gross margins of participants and non-participants after matching. It also gives a t-statistic that used for doing a t-test. This study employs statistical matching to address the problem of selection bias. This involves pairing participants and non-participants that are similar in terms of their observable characteristics (Dehejia and Wahba, 2002). When outcomes are independent of assignment to treatment, conditional on pretreatment covariates, matching methods can yield an unbiased estimate of the treatment impact (Ali and Abdulai, 2010). It follows that the expected treatment effect for the treated population is of primary significance. This effect may be given as Ï„ I I=1 = E (Ï„ I I = 1) = E (R1 I I = 1) - E (R0 I I = 1) whereA Ï„A is the average treatment effect for the treated (ATT),A R1A denotes the value of the outcome for participant farmers and R0A is the value of the same variable for non-participant farmers. As noted above, a major problem is that we do not observe E (R0 I I = 1).A  Although the difference [ Ï„e = E (R1 I I = 1) - E (R0 I I = 1) ] can be estimated, it is a potentially biased estimator. In the absence of experimental data, the propensity score-matching model (PSM) can be employed to account for this sample selection bias (see for example Ali and Abdulai, 2010; Dehejia and Wahba, 2002). The PSM is defined as the conditional probability that a farmer participation, given pre-participation characteristics (see for example Ali and Abdulai, 2010; Rosenbaum and Rubin, 1983). To create the conditions of a randomised experiment, the PSM employs the unconfoundedness assumption also known as conditional independence assumption (CIA), which implies that onceA ZA is controlled for, participation is random and uncorrelated with the outcome variables as pointed out by Imbens and Wooldridge (2009).A The PSM can be expressed as, p(Z) = Pr {I = 1IZ} = E{I I Z} Where I = {0, 1}A is the indicator for participation andA ZA is the vector of pre-participation characteristics. The conditional distribution ofA Z, given p(Z) is similar in both groups of participants and non-participants. The empirical analysis was conducted using the STATA statistical package.

5.6.2.4 Factor analysis

This analysis explores the motivation and main problems faced by participants farmers. The respondents were presented with a list of factors indentified from the literature, focus group and in-depth interviews suggesting potential problems faced by small-scale farmers. They were asked to indicate the important each issue on a Likert scale ranging. First, the mean important score method was used to indicate the important motivation and main problems faced by participant farmers. Then, to enable the factor of participation and problems faced by participant farmers to be better understood and classified into subsets, the importance score were subjected to Factor Analysis. Factor analysis is a multivariate method of exploring the structure of data with the object of data reduction and interpretation, particularly in marketing research which may consist of a number of variables and must be reduced to a manageable level. Therefore, factor analysis allows us to look at a group of variables that tend to be correlated to each other and allows us to indentify underlying dimensions that explain these correlations (Malhotra, 2007). In terms of the results, the variables included in this analysis were categorised into groups of variables. The next step is testing the appropriateness of the factor model. The useful statistic is the Kaiser-Meyer-Olkin (KMO), which normally states that a value of KMO greater than 0.5 indicates that the correlation between the pair of variables is desirable. Once the factor analysis demonstrated that it is a proper technique for analysing the data, Principle Component Analysis (PCA), which is one of the most common approaches of factor analysis, is implemented due to the recommendation for the data use in subsequent multivariate analysis. The small numbers of variables were extracted from PCA, which there are several procedures for determining the number of factors (or so-called, principle components). The common approaches are based on Eigenvalue, scree plot (a plot of the Eigenvalue against the number of factors) and the percentage of variance, etc. The first determination is done by only using factors with an Eigenvalue equal to 1 or greater in the analysis. Finally, the factor can be interpreted by indentifying the variable that have a large loading on the same factor. In addition, the factor rotation following the “Varimax”? method by means of orthogonalization of the factor can help the interpretation to become simpler and more accurate (Malhotra, 2007). The empirical analysis was conducted using the SPSS statistical package Finally, the study gathered together the value chain analysis (supply chain map) and discussions from the first stage, and the quantitative insights from the second stage to comment on what has been added to the existing stock of knowledge on smallholders coping with food transformation and the role of social enterprises, and make broad policy recommendations for the sector.

Table 5.xx Decision tree of analysis techniques:

1. Quantitative Analysis

Econometric methods

Research Questions Number/kind of dependent variable Number/kind of independent variables Analytic strategy Goal of analysis RQ7 Multiple (discrete) Multiple (continuous) Probit Analysis InverseA cumulative distribution functionA  associated with the standardA normal distribution RQ9 Multiple (discrete) Multiple (continuous) Multiple Regression: Gross Margin Regression Relationship between several independent or predictor variables and a dependent or criterion variable Strategy: Propensity score matching method Solve the problem of selectivity bias, or self-selection bias

Statistic methods

Research Questions Variables Analytic strategy Goal of analysis RQ8 Multiple (continuous) Factor Analysis Correlation linear combination of dependents variables with independent variables RQ8 Significant difference of mean scores? Strategy: Wilcoxon Signed rank (0.05%) Significant difference of paired mean scores

2. Qualitative Analysis

Research Questions Index Analytic strategy Goal of analysis RQ1, RQ2, RQ3 RQ4, RQ5, RQ6 FIGURES xxxxxxxxxxxxxxxxxxxxxxxxxxxxx Xxxxxxxxxxxxxxxxxxxxxxxxxxxxx TABLES xxxxxxxxxxxxxxxxxxxxxxxxxxxxx Xxxxxxxxxxxxxxxxxxxxxxxxxxxxx Value Chain Analysis Strategy: supply chain mapping technique Understanding the patterns of agri-food value chains and business potential such as markets, relationships, and critical constrains that limit small-scale farmers growth and industry competitiveness.

5.7 Conclusions

In this chapter the data used and research methodology in this study were presented and discussed. The data collection was divided into four stages (2 phases) in order to facilitate the achievement of research objectives. Both qualitative and quantitative information was collected. The first one consisted of secondary information and in-depth interviews (using semi-structured questions) with different players of the agri-food value chains and social enterprises in Thailand. This information was very useful for making a general characterization of the agri-food value chain restructuring and the role of social enterprises in helping small-scale farmers participate with the modern trade markets as well as for designing the questionnaire used for collecting quantitative data through a survey of participant and non-participants farmers in the modern trade chains. In the same way, qualitative information complements and helps to interpret quantitative results. A summary and discussion of quantitative methods were provided in this chapter. Probit regression analysis can assess the determinant factors of small-scale farmers participating with social enterprise in producing for modern trade markets. Gross margin analysis and selectivity bias methods can compare the profitability of participant and non-participant farmers. Moreover, the mean importance scores and Factor Analysis can assess the impact on farmers of a number of problems (variables) faced by small-scale farmers. The following chapters (Chapter 7 and Chapter 8) place the study within the findings regarding the characterization of the agri-food value chain and the role of social enterprises in Thailand.

Cite this page