2.1 Study Area Selection CA technologies have been implemented or are being practiced in seven Upazilas in four districts of Bangladesh namely Rajbari, Thakurgaon, Rajshahi and Mymensingh. Considering project resources, logistic support and CA technology adoption, three Upazilas namely Durgapur and Godagari Upazilas of Rajshahi district and Sadar Upazila of Thakurgaon district were purposively selected for the study.
2.2 Sampling Design and Data Collection The households were selected considering the level of adoption of CA technologies. At first, a complete list of farmers adopted CA technologies (i.e. minimal soil disturbance, crop residue retention, and suitable crop rotations) was prepared with the help of personnel from DAE and CA project. Then, a total of 135 CA farmers taking 45 farmers from each Upazila were selected randomly for this study. Again, a total of 270 non-adopting farmers were randomly selected for this study as control. Thus, the total sample size was 405. Data and information were gathered from selected farmers using a pre-tested interview schedule. Data were collected during January-February, 2017.
2.3 Analytical Techniques Collected data were edited, scrutinized, summarized and analyzed using computer software. Descriptive statistics were mostly used to present the results of the study. Moreover, the following Logit model was used to identify factors of CA technology adoption at farm level. According to Gujarati [10], the Logit model guarantees that the estimated probabilities lie in the 0-1 range and that they are not linearly related to the explanatory variables. In addition, it is easier and more convenient to compute than the Probit model. Since the dependent variable is dichotomous, OLS cannot be used. MLE method was followed to run the Logit model using STATA software (Version 12). The specification of the model was as follows:
Logit {P(Y=1)} = log{P/(1-P)} = α + β1X1 + β2X2 +…….. + βKXK
Where, Y is a categorical response variable with 1= adopters and 0 = otherwise; α is the intercept; β1, β2.... βk are coefficients of independent variables X1 X2... XK; P is the probability of adopting CA technology, and (1-P) is the probability that a farmer does not adopt CA technology.
The empirical Logit model was as follows: Y = α + β1X1 + β2X2 + β3X3 + β4X4 + β5X5 + β6X6 + β7X7 + β8X8
Where, Y= Dependent variable (1= Adopter, 0 = Non-adopter), X1 = Farmer’s age (year), X2 = Education (year of schooling), X3 = Family size (No./HH), X4 = LnFarm size (decimal), X5 = Availability of VMP (score), X6 = Societal membership (wt. score), X7 = Innovativeness (wt. score), X8 = Extension contact (wt. score), α = Constant, β1 β2 β3 β4 .................. β8 are the coefficients to be estimated.