Additive modeling of zonal level crop production in Ethiopia

Abstract

Abstract Crop production is the main source of food security and income for smallholder private farmers in Sub-Saharan countries. To have a sustainable source of food security and economy, it is important to identify covariates that affect crop production linearly and nonlinearly. Annual agricultural sample survey data of eight Meher seasons, from 2012/13 to 2019/20, is used in this study with the main objective of identifying the set of covariates that have linear and nonlinear effects on crop production and estimating their effects using an additive mixed effects model. The minimum, mean, and maximum crop production across the country for the study period were 1.616, 8.693, and 147.843 quintals, respectively, and 50% of the farmers produced less than 6.95 quintals. The histogram, kernel density, and P-P plots suggested that log-transformed crop production is approximately normally distributed. From the competing models’ summary statistics, information criteria values, and analysis of variance tests, relaxing the linearity assumption and including a random effect in the model has improved model performance, suggesting the additive mixed effects model best fits the data on hand. Gambella, SNNP, and Oromia regions have significantly different overall mean crop production than the reference in Dire Dawa town. Covariates like year, proportion of female farmers, household age, and UREA fertilizer used have a significant nonlinear effect, while covariates like proportion of educated farmers, area used, and proportion of farmers who received credit service have a significant linear effect on log crop production. The basic model assumptions are not violated, so the final additive mixed effects model can be used for prediction and inference purposes. We recommend farmers use more croplands, indigenous seeds, and UREA fertilizer; practice pure agriculture; and participate in local farmers associations. Policies regarding the participation of female and educated farmers, the implementation and effectiveness of credit services and extension programs, private investors to participate in crop production, and the provision of farm inputs to the elderly farmers should be revised.

Keywords

• Nonlinear effect
• Smooth terms
• Additive mixed models
• Random effect
• Crop production