Prediction of sustainability of various homegardens in Lampung Province, Indonesia using AHP [Analytical Hierarchy Process] and logit models

The results revealed that the majority of the respondents were Javanese (42.93%) and that the respondents were generally in the productive age (38.70 years old); 50.70% were male and 49.30% were female; and have attended elementary school. Most of the respondents (81.39%) have no problem in labor adequacy in each of the activities in the pekarangan. More than one-fifth (29.28%) have two activities in their pekarangan and 26.05% have three activities. Pekarangan crops are harvested monthly and all harvests are used by 32.26% of the respondents for their daily needs. Pekarangan's Social Acceptability Index was found to be moderate level and Farming Index indicated low level. All equations in binomial logit and three equations in multinomial logit were found to be effective. Binomial models developed for each district were all effective since all models yielded at least 90% correct classifications of the sustainability level of the pekarangan units under each district. On the other hand, the majority number of data in equation of multinomial logit were correctly predicted as low except Tanjung Baru village, North Lampung district which is moderate. The significant variables were ranked using the Analytical Hierarchy Process. The ranking of significant variables were based on coefficient and standard error. The most significant variables based on binomial logit that affected pekarangan sustainability were number of agroforestry activities in pekarangan (Act), Social Acceptability Index (SAI), Farming Index (FI), and Decision Making (DM). In multinomial logit y=1 (moderate level), the most significant variables were FI, ethnic group (Et), soil pH, Occupation (Op), Education (ES), soil organic matter content (OMC), soil texture (Tx), Act, and number of family members (Fm). In multinomial logit y=2 (high level), FI, Act, OMC, SAI, soil aggregate (Ag), and Et were identified as most significant variables