Prediction of success in village fish farming in northeast Thailand

Abstract

It is apparent that substantial differences exist between communities with regard to their ability to initiate and sustain community resource management (CRM) projects. The knowledge and identification of these differences is important when formulating policies and planning projects since scarce resources should be allocated to those communities which show the greatest potential for the successful use of these resources and alternative programmes formulated which recognise different community potentials for successful programme implementation.

This study is concerned with inter-community differences as they may affect the success of the "Village Fish Pond" (VFP) project in a sample of 205 Northeast Thailand villages. The specific objective of the thesis is to develop an empirical predictive model to identify community characteristics which underlie outcomes of the VFP in Northeast Thailand. The development of the predictive model is based upon a review of literature on the innovation adoption, collective action, and community development and management. The study is concerned with decision-making at the village level in general, and with the factors associated with the acceptance of development projects in particular.

A backward stepwise logistic regression analysis is used to test the predictive model. Data required for the analysis came from two different sources. First, a questionnaire survey was conducted in a sample of Northeast VFP villages (in which the VFP was established between 1986 to 1989) to evaluate the performance of the VFP in each of the participating villages. Second, for the same 205 villages, predictor variables were selected from a national village-level database which is available from the Institute for the Processing of Information for Education and Development (IPIED) at Thammasat University in Bangkok. The village community is used as the unit of analysis, project performance is used as the dependent variable, and community characteristics are used as the predictor (independent) variables.

For the questionnaire analysis, the sampled 205 VFP villages were classified into three groups according to their degree of project performance: a complete success, a complete failure, and an indeterminate outcome (neither a clear success nor a clear failure). Since the dependent variable consists of three categories, three predictive models were developed. The first model distinguishes villages which are more likely to succeed in implementing the VFP from those that are more likely to fail. Here, the dependent variable is coded as "2" (Success) and "0" (Failure). In the second model, the dependent variable is coded as "1" (Indeterminate) and "0" (Failure) to represent the differences between those villages which are struggling to continue their VFP activities and those which have rejected the project by stopping all project activities. The third model differentiates those villages that continue to practice the VFP but vary in the degree of success. Its dependent variable is coded as "2" (Success) and "1" (Indeterminate).

In order to validate the resulting models, 45 villages (about 20% of the total 205 villages) were randomly selected as a validating sample to be used in testing the models for accuracy and efficiency. The majority of the VFP sampled villages (n=160) was used to test the predictive models. The major results show that only Model 1 achieves a good level of statistical fit with an overall accuracy of 78.16 percent correctly predicted in the model building sample and 70.83 percent correctly predicted in the validating sample. Models 2 and 3 performed less well than expected according to the results of the diagnostic and validation procedures used in the thesis. It is assumed that their low predictive power is due to complication with the indeterminate group, which may result from the short evaluation period used in the research.

In Model 1, nine community characteristics are found to be related to the degree in which community decision-making and performance supported continuing the VFP. The results suggest that property tax (ZPTAX) is the most influential factor in the resulting model. Respectively, in descending ordered of influence, the other variables are: the number household per tubewell and shallow well (WELL), the number of households per television set (HHPERTV), agricultural land use (SOIL_QUALI), percentage of houses in good sanitary condition (GEN_HEALTH2), percentage of households pursuing the most common industry (INDUSTRY1), the number of community sport activities in the past two years (ZACTIVITY2), percentage of households pursuing fish culture (FRE_RAISER), and the number of households per store (SHOP). In general, it appears that communities with a larger population and territory, a higher economic status, lesser migration to work outside sub-district areas, and a more peaceful society are more likely to succeed in independently implementing and sustaining the VFP than other communities.

In light of the above results, the implications of the results in communal fish culture and rural development planning are examined. Finally, suggestions are made for future research based on the results of the study.