Sanzidur Rahman
Associate Professor in RuralDevelopment, School of Geography, Earth and Environmental Sciences, University ofPlymouth, Drake Circus, Plymouth, UK.
Bangladesh, Modern rice producers, Sample selection framework, Stochasticproduction frontiers, Technical efficiency.
Socio-economic and Policy
Studies analysing determinants of technology adoption Several studies have analysed the determinants of modern technology adoption by farmers in developing countries using simple ad hoc models. These are typically ordinary least squares (OLS), probit or tobit regressions of technology adoption on variables representing: (i) socioeconomic circumstances of farmers – such as, farm size, tenurial status, farmers’ education level, farming experience, family size and gender and (ii) institutional and bio-physical factors – such as irrigation, credit, extension contact, member-ship of organisations, and distance to market/bus stop/extension office (e.g., Hossain 1989; Nkamleu and Adesina 2000; Shiyaniet al.2002; Asfaw and Admassie 2004). Few of these studies outline the implicit theoretical under-pinning of such ad hoc modelling (e.g., Nkamleu and Adesina 2000), which is the assumption of utility maximisation by rational farmers. Furthermore, all of these studies ignored or omitted price factors (both input and output prices) as determinants of technology adoption, which has an important bearing on productivity and resource allocation decisions, and hence provides an incomplete picture of farmers’ decision-making processes. The model of technology adoption developed by Pitt (1983) explicitly takes into account price and non price factors in determining adoption while allowing for switching between varieties, but assumes farmers to be fully efficient in their production technologies. With the development of stochastic frontier analysis by Aigneret al. (1977), a large number of studies followed which typically place the farming efficiency of developing country farmers in a range of 60 per cent to 82 per cent (e.g., Ali and Flinn 1989; Wanget al.1996; Coelliet al.2002; Rahman 2003; Bravo-Uretaet al.2007). As a result, analysis of factors determining technology adoption under the assumption of the farmer being fully efficient inherently incorporates bias into the results. The contribution of this study to the existing literature on the economics of technology adoption, as well as efficiency analyses, is the extension of the model of technology adoption developed by Pitt (1983) to relax the restrictive assumption of fully efficient farmers. This approach is used to jointly address our three key research questions. 3. Theoretical framework The conventional approach to incorporating selectivity is the estimation procedure proposed by Heckman (1976), which involves the following two steps: •Step 1: Fit the probit model for the sample selection equation.•Step 2: Using the selected sample, fit the second step model (ordinary least squares or weighted least squares) by adding the inverse Mills ratio from he first step as an independent variable to correct for selectivity bias and test its significance. However, Greene (2006) claims that such an approach is in appropriate for several reasons in models that are not linear, such as probit, tobit and soforth. This is because:•The impact on the conditional mean of the model of interest will not necessarily take the form of an inverse Mills ratio. Such an adjustment is appropriate and is specific to linear models only.•The bivariate normality assumption needed to justify the inclusion of the inverse Mills ratio in the second model does not generally appear any where in the model.•The dependent variable, conditioned on the sample selection, is unlikely to have the distribution described by the model in the absence of selection (Greene 2006). Hence, Greene (2006, 2008) proposed an internally consistent method of incorporating ‘sample selection’ in a stochastic frontier framework which was adopted in our study and is elaborated as follows. Farmers are assumed to choose between modern and traditional rice varieties to maximise profits subject to a set of price and non price factors. The decision of the ith farmer to choose modern rice is described by an unobservable selection criterion function, Ii*, which is postulated to be a function of a vector of exogenous output prices, and factors representing farmers’ socio-economic circumstances, as well as bio-physical and environmental factors. The selection criterion function is not observed. Rather a dummy variable, I, is observed. The variable takes a value of 1 for modern rice farms and 0 other-wise. 4. Data and variables 4.1. Data This study utilizes cross-sectional primary data for the crop year 1996. The data were collected by a team of field researchers via an intensive farm survey coordinated by the author. Multistage random sampling techniques were used in selecting study locations as well as the sample farmers. Three agro-ecological regions of Bangladesh are represented in the dataset: the Old Brahmaputra Floodplain, the High Ganges River Floodplain and the Middle Meghna River Floodplain. Samples from 21 villages – eight villages of the Jamalpur Sadar subdistrict of Jamalpur, six villages of the Manirampur sub-district of Jessore and seven villages of the Matlab subdistrict of Chandpur –were used to represent these regions. Information was obtained on input and output quantities as well as prices, at the plot level. Additionally, socioeconomic characteristics of the farm families and village-level infrastructural development and soil fertility data were also recorded. The geographical dispersion of the sample plots and imperfections in input markets in Bangladesh ensure adequate variability in prices across the cross-section. A total of 946 observations (324 observations of traditional rice varieties and 622 observations of modern rice varieties) constitute the final sample.
The Australian Journal of Agricultural and Resource Economics, 55, pp. 273–290
Journal