Measuring efficiency using frontier profit function Production inefficiency is usually analysed by its two components – technical and allocative efficiency. In a production context, technical efficiency relates to the degree to which a farmer produces the maximum feasible output from a given bundle of inputs (an output oriented measure), or uses the minimum feasible level of inputs to produce a given level of output (an input oriented measure). Allocative efficiency, on the other hand, relates to the degree to which a farmer utilizes inputs in optimal proportions, given the observed input prices (for details, see Coelli et al., 2002). Recent developments combine both measures into one system, which enables more efficient estimates to be obtained by simultaneous estimation of the system (e.g., Ali and Flinn, 1989; and Wang, et al., 1996). The popular approach to measure efficiency, the technical efficiency component, is the use of frontier production function. However, Yotopolous and others argue that a production function approach to measure efficiency may not be appropriate when farmers face different prices and have different factor endowments (Ali and Flinn, 1989). This led to the application of stochastic profit function models to estimate farm specific efficiency directly2 (e.g., Ali and Flinn, 1989; Kumbhakar and Bhattacharya, 1992; Ali et al., 1994; and Wang et al., 1996). The profit function approach combines the concepts of technical and allocative efficiency in the profit relationship and any errors in the production decision are assumed to be translated into lower profits or revenue for the producer (Ali et al., 1994). Profit efficiency, therefore, is defined as the ability of a farm to achieve highest possible profit given the prices and levels of fixed factors of that farm and profit inefficiency in this context is defined as loss of profit from not operating on the frontier (Ali and Flinn, 1989). Also, in a number of studies on efficiency measurement (e.g., Sharif and Dar, 1996; Wang et al., 1996), the predicted efficiency indices were regressed against a number of household characteristics, in an attempt to explain the observed differences in efficiency among farms, using a two-stage procedure. Although this exercise has been recognized as a useful one, the two-stage estimation procedure utilized for this exercise has also been recognised as one which is inconsistent in its assumptions regarding the independence of the inefficiency effects in the two estimation stages 3 (Coelli, 1996). Battesse and Coelli (1995) extended the stochastic production frontier model by suggesting that the inefficiency effects can be expressed as a linear function of explanatory variables, reflecting farm-specific characteristics. The advantage of Battesse and Coelli (1995) model is that it allows estimation of the farm specific efficiency scores and the factors explaining efficiency differentials among farmers in a single stage estimation procedure. The present paper utilises this Battesse and Coelli (1995) model by postulating a profit function, which is assumed to behave in a manner consistent with the stochastic frontier concept. This model is applied to a large sample of rice producers in three agro-ecological regions of Bangladesh, differentiated by variety and by season. Data and the Empirical Model Data Primary data for the study pertains to an intensive farm-survey of rice producers conducted during February to April 1997 in three agro-ecological regions of Bangladesh. Samples were collected from eight villages of the Jamalpur Sadar sub-district of Jamalpur, representing wet agro-ecology, six villages of the Manirampur sub-district of Jessore, representing dry agro-ecology, and seven villages of the Matlab sub-district of Chandpur, representing wet agro-ecology in an agriculturally advanced area. A total of 406 farm households from these 21 villages were selected following a multistage stratified random sampling procedure. Among these 406 farms, 380 farms produced modern varieties of rice and therefore taken as the final sample size. In analysing crop production, it is often the case that data is only available for the major inputs, such as land, labour, fertiliser, and animal power. However, crop production is affected by many other variables that play significant roles in explaining performance. In this study, an attempt was made to collect information on most of the inputs used for rice production. Thus, information on the use of seeds, pesticides, and farm capital assets was collected. This is expected to increase the explanatory power of the analyses significantly. It is often argued that seeds and animal power services are more or less used in fixed proportions, so their omission is not important (Hossain, 1989 and Hossain et al., 1990), but results here suggest that this is not the case.