Date of Award
Doctor of Philosophy (PhD)
Economics / Economic Policy
Frank T. Denton
In this study, we have attempted to empirically assess the efficiency of a sample of Bangladeshi farmers in the civilization of rice, the most important crop in the country. That sample is drawn from Khilghati, a village lying about 95 miles north of the capital city of Dhaka. Our empirical analysis is based on survey data collected by Khandker (1982) for the 1981-82 crop year. The importance of farmer efficiency in Bangladeshi architecture cannot be overstated given the predominance of agriculture in the economy, low crop yields and limited land supplies relative to population. While the adoption of "Green Revolution" technologies involving the use of more production seed varieties and fertilizer inputs are undoubtedly important for increasing yields, and while progress has been made in that direction, attention also needs to be paid to improving the efficiency of farmers within the framework of any technology, be it of the traditional or more modern kind. the data for Khilghati provides an opportunity to examine this question. In particular, we construct indices to assess the efficiency of Khilghati farmers in the cultivation of the traditional, wet-season "Aman" and "Aus" rice crops, and the dry season, new technology "Boro" rice crop. Several efficiency indices are estimated for each crop. Thus, we construct multi-factor indices of technical and allocative efficiency, as well as factor-specific efficiency indices which are indicators of the efficiency of individual factor usage. Technical efficiency refers to the efficiency of factor use in the physical sense and is an attribute of the production function, while allocative efficiency is a cost concept and is associated with the question of whether a firm utilizes inputs in the "right" (that is cost-minimizing) proportions. Technical and allocative efficiency can be modeled and estimated in different ways [Schmidt (1986)]. In this study, given the nature of the data available, the efficiency indices are constructed from deterministic and stochastic Cobb-Douglas production frontiers. In the deterministic case, all departures from the frontier are taken to represent inefficiency, while stochastic frontiers frontiers additionally allow for statistical noise. The estimation of the production frontier generally (though not always) involves assumptions about the distribution of the technical inefficiency term in the deterministic case, and additionally statistical noise in the stochastic case. To examine the sensitivity of the estimates, we consider two alternative assumptions about the technical inefficiency term - one, that it follows a half-normal distribution and two, that it follows an exponential distribution. These distributions imply that technical inefficiency places the firm on or below the deterministic/stochastic frontier. In the stochastic case, we assume that the disturbance term reflecting statistical noise is normally distributed with zero mean and constant variance. The deterministic frontier is estimated by two versions of the corrected least squares (COLS) as well as by linear and quadratic programming techniques, while the stochastic frontier is estimated by the COLS and maximum likelihood methods. The distributional assumptions stated earlier are critical aspects of the estimation strategy, particularly in the stochastic case where they are needed in order to separate technical inefficiency from statistical noise. Our results indicate that the efficiency estimates are somewhat sensitive to estimation method and distributional assumptions, though primarily in the deterministic case. More importantly, we find that the relative ranking of farmers along the technical or allocative efficiency sprectrum is largely independent of estimation method and distributional assumptions under the deterministic and stochastic approaches. The major difference between the two approaches is that, under the latter, the average level of technical efficiency is clearly higher, with statistical noise being an important reason for departures from the deterministic kernel. This points to the importance of allowing for statistical noise. The following discussion deals with estimates obtained from the stochastic frontier, unless noted otherwise. The estimates of technical efficiency suggests that Khilghati farmers are highly efficient in the cultivation of all crops, with at least 70 percent of farmers having a technical efficiency index in excess of 80 percent. Average technical efficiency is about 90 percent in Aman, and about 85 percent in Aus and Boro cultivation. In allocative terms, farmers are markedly less efficient, and the inter-crop variation is also greater. Thus, average allocative efficiency is in the 70-75 percent range for Aman and Boro, but only about 50 percent for Aus. Alternative distributional assumptions have only a minor impact on the allocative efficiency estimates and a somewhat larger impact on the technical efficiency estimates. We used our estimates of technical and allocative efficiency to examine a number of issue. Thus, correlation analysis provides limited evidence to indicate that a farm household's technical (allocative) efficiency indices are related across crops, indicating perhaps that efficient cultivation practices related across crops, indicating perhaps that efficient cultivation practices are crop specific, and that farmers' growing experience and/or learning by doing also vary across crops. We also found little evidence to support the view that technically more efficient farmers are also allocatively more efficient. In addition, we examined the widely held view that better educated farmers are relatively more efficient in the technical and/or allocative senses, in a regression context. Only allocative efficiency in Aman cultivation and technical efficiency in Boro cultivation were found to bear a statistically significant positive relationship with farmer education. Our factor-specific estimates of efficiency suggest that farmers are least efficient, in the physical sense, in the use of labour (the relatively abundant factor), and generally most efficient in the use of land (the relatively scarce factor). The inefficiency of labour usage is substantial; however, the greatest gain (in terms of cost saving) would be realized through the elimination of inefficiency in land or other inputs, and not labour. Several implications follow from our findings. Farmers appear to be as efficient in the new-technology Boro crop as in the traditional Aman and Aus crops. A policy of encouraging the adoption of such HYV crops is thus well-founded. However, attention clearly needs to be paid to improving farmer skills within the existing crops. For instance, rural development policies could be geared to improving allocative skills, perhaps through rural education, and more effective management of extension services and rural co-operatives. our estimates point to a substantial cost saving via such an improvement. Those policies would probably have to take account of possible differences in efficient cultivation practices across crops.Policies aimed at improving the efficiency of highly scarce inputs such as land could also go a long way towards improving the overall efficiency of farmers. In fact, since the relative price of land can be expected to increase over time, the cost reductions by improving the efficiency of land use could be substantial. Finally, it may be that institutional constraints on individual behaviour foster, inefficiency. For example, the lack of access of smaller farmers may lead them to make inefficient choices. Ensuring greater access to those farmers could be important in promoting greater efficiency.
Sharif, Najma Rownaq, "Farmer efficiency: The frontier approach applied to rice farming in Bangladesh" (1992). Open Access Dissertations and Theses. Paper 3891.