Efficiency Performance of Rice Farms in Northern Bangladesh: An Application of the Stochastic Frontier and Data Envelopment Analysis (DEA) Mode

Abstract

This study examines farm level efficiency of the rice producers in Northern Bangladesh by estimating technical efficiency (TE) for the period of aman season 2009-10 and boro season 2010 using farm level cross sectional field survey primary data. Two methods for measuring efficiency are applied, Stochastic Frontier Analysis (SFA) and Data Envelopment Analysis (DEA). Technical efficiency is computed by estimating the Cobb-Douglas stochastic frontier. We also investigate the factors associated with technical inefficiency. Technical inefficiency effects are modeled as a function of age and experience of farmers, years of schooling of the rice producers, size of plots, credit facilities, extension services and land degradation factors in a single stage estimation technique using maximum likelihood method. Technical and Scale efficiencies are derived from solving input-oriented and output-oriented constant return to scale (CRS), variable returns to scale (VRS) DEA models. Tobit inefficiency effects model is used to quantify factors associated with technical and scale inefficiency for both input-oriented and output-oriented CRS and VRS DEA frontiers. We compare the results obtained from both the Cobb-Douglas stochastic frontier and DEA frontiers. We conduct a survey of 251 rice producers from ten villages of three upazilas of different three districts. We select Tanore upazilla from Rajshahi district, Manda upazilla from Naogaon district and Nachole upazilla from Chapai Nawabganj district of northern Bangladesh. These upazillas and 10 villages are selected by applying purposive and stratified sampling. Finally, a simple random sampling technique is applied to each stratum. The cross sectional primary data are collected from rice producers by using a questionnaire. We collect data mainly on rice output and related inputs, socio-economic characteristics, environmental degradation and other information. The Cobb-Douglas stochastic frontier results exhibit that the rice producers of our study area in aman season (S1), boro season (S2) and both aman and boro seasons (S1+S2) together average technical efficiency (TE) are 85.17 per cent, 80.42 per cent and 86.85 per cent respectively. The minimum TE scores are 55.73 per cent, 48.73 per cent and 57.51 per cent in aman season (S1), boro season (S2) and both aman and boro seasons (S1+S2) together and maximum TE scores are 98.22 per cent, 99.93 per cent and 98.50 per cent respectively with standard deviations of 9.23, 13.01 and 8.49 for aman season (S1), boro season (S2) and both seasons (S1+S2) together respectively. Both input-oriented and output-oriented CRS and VRS DEA methods give almost similar results in aman season (S1). The input-oriented DEA frontier results show that the average technical efficiency estimates in aman season (S1) are 86 per cent and 88 per cent under constant returns to scale (CRS) and variable returns to scale (VRS) assumptions and the average scale efficiency (SE) is 97 per cent. The average values for technical efficiency measures and scale efficiency from the output-oriented CRS and VRS frontiers are 86 per cent, 89 per cent and 96 per cent respectively. The corresponding values for boro season (S2) from input-oriented DEA methods are 83 per cent, 85 per cent and 97 per cent and from output-oriented DEA methods these value are 83 per cent, 89 per cent and 93 per cent per cent under CRS and VRS respectively. The average values for technical efficiency measures and scale efficiency from the input-oriented and output-oriented CRS and VRS frontier for both aman and boro seasons (S1+S2) together are 86 per cent, 89 per cent and 97 per cent and 86 per cent, 89 per cent and 96 per cent respectively. The efficiency estimates from stochastic frontier analysis (SFA) model for aman season (S1) are slightly higher than boro season (S2) as expected. According to stochastic frontier, results technical efficiency (TE) in aman season (S1) 14.83 per cent, boro season (S2) 19.58 per cent and both aman and boro seasons (S1+S2) together 13.15 per cent, which could be improved if the rice producers could operate at full efficiency levels. The efficiency estimates of data envelopment analysis (DEA) model from input-oriented CRS and VRS results for aman season (S1) are slightly higher than boro season (S2) and similar result exists from scale efficiency (SE) of both aman and boro seasons (.S1+S2) together. According to DEA, results of technical efficiency (TE) from input-oriented CRS and VRS in aman season (S1) shows that 14 per cent and 12 per cent could be increased and 17 per cent and 15 per cent could be increased in boro season (S2). The corresponding values for both aman and boro seasons (S1+S2) together are 14 per cent and 11 per cent, which could be enhanced if the farmers could operate at full efficiency levels. Inefficiency effects model shows that age and experience of farmers, years of schooling of farmers, land size, credit facilities, quality extension services are inversely related to inefficiency of farms in both aman and boro seasons. Environmental factor, such as land degradation is the most statistically significant factor effecting technical inefficiency. This implies that land degradation is not only reducing output from given inputs but also causing sub-optimal cost minimizing input decisions. Policies should be taken to reduce land fragmentation, to increase rural credit facilities and the quality extension services, and also to reduce factors which cause land degradation. As a result, these will increase rice production and rice producer’s income and hence bring the welfare of farmers.