Modeling predictors of poverty in agricultural households in Uganda: application of multilevel and interaction methods

Abstract

This study aimed to model predictors of poverty in agricultural households in Uganda using a multilevel and interaction regression model. The research utilized data from the Uganda National Household survey (UNHS 2019/20) data from Uganda Bureau of Statistics. A logit model was used in the estimation and estimates were provided using Multilevel methods approach. Key findings suggest that poverty in agricultural households was positively and significantly influenced by gender of the household head, marital status of the household head, income stability of the household, age of the household head and livestock ownership. Additionally, regional differences accounted for 17.9 % of the variations in poverty levels in Uganda and understanding such regional differences and their influence on poverty levels can assist policymakers and organizations in designing targeted interventions and policies. Such measures can address the specific challenges faced by different regions and promote more equitable development across Uganda. However, poverty in agricultural households was negatively and significantly influenced by residence status, saving accounts ownership and household size. Based on the study's findings, the key policy recommendations were; government should continue implementing gender-focused interventions to address gender disparities especially women empowerment programs such as access to resources including land, equal access to employment opportunities and equal access to education to reduce poverty among women. Regarding income instabilities in agricultural households due to price fluctuations, government should empower famers to form farmer groups where they can collectively increase their bargaining power to avoid price fluctuations.