2.1. Data Collection To collect data on fertilizer application methods, we opted for anonymous, semi-structured, personal interviews with farmers. The method we approached is effective (face-to-face, focus group discussion, and phone interviews) for data collection, as semi-structured interviews give farmers an opportunity to explain, increasing the response rate by finished questionnaires. We conducted the interview with respondents selected by two-stage cluster sampling. Our questionnaire consisted of 18 questions categorized into four parts: (i) farmers’ personal data (for example, sex, age, education, income, family members, and family bearing capability); (ii) farmers’ knowledge level about environmental sustainability (for example, their understanding of environmental sustainability, fertilization methods, training, and amount of fertilizer application); (iii) farmers’ capital (land ownership, land rental status, amount of fertilization in rental lands, availability of fertilizers); and (iv) fertilizer governance (for example, the role of agricultural extension agents and availability of soil testing). A copy of the interview questions appears at the end of this study in the supplementary materials. The questionnaire was prepared in non-scientific language for clarity for the farmers. The set of questions was developed upon consultation with a number of researchers and experts to can maintain scientific ethics and be easier for the farmers to complete. A pilot study was conducted with eight farmers before the full study to prove its quality.
2.1.1. Sampling and Study Sites Bangladesh is located in South Asia at 23.8o N and 90.3o E. In Bangladesh, agriculture is one of the most important sectors and is characterized by a large number of small farmers. The agricultural sector contributes 21% to gross domestic product (GDP) and employs about 50% of the labor force. Rice is the staple food for a population of about 160 million and provides over 70% of direct calorie intake in the country. Four out of 64 districts in Bangladesh were selected for data collection by two-stage cluster sampling. Cluster sampling assumes that selection of a household within a cluster is not independent of the selection of other households; members of a cluster are therefore likely to be similar. A household was defined as a family with one head of household eating and sleeping under the same roof. Two-stage cluster sampling aims to minimize survey costs and control the uncertainty related to estimates of interest; a simple case of multistage sampling, it selects cluster samples in the first stage and then selects a sample of elements from every sampled cluster.
The four districts include Manikganj, Noakhli, Dinajpur, and Moulvibazar. The villages Dhakaijora, in Manikganj, and Sondolpur, in Noakhali, were randomly picked for data acquisition. Shibganj was selected from Dinajpur and Baraoura from Moulvibazar. We interviewed 61 farm households in Dhakaijora, 54 in Sondolpur, 46 in Dinajpur, and 50 in Moulvibazar following simple random sampling for a total of 211 farm households. The districts chosen are agriculturally important and diversified in terms of agricultural patterns, farmers’ agricultural practices, topographical difference, and nature of fertilization; however, the demographic features of all clusters are homogeneous. For the diversification of these agricultural areas, the credibility and reliability of the data can be considered high, as there is an established relationship between researchers and farmers.
2.1.2. Data Process and Analysis Data, collected from interviews was input directly into SPSS for Windows (Version 16.0., SPSS Inc., Chicago, IL, USA,) for statistical analysis. Descriptive statistics have been divided into line graphs, for showing comparative fertilizer consumption between the world average and Bangladesh, and tabular statistics showing the basic socioeconomic and demographic features of farmers. Percentile, standard deviation, and variance options were employed in our description. To estimate the correlation and effect of the variables, we applied binary logistic regression.