Data to Decisions: leveraging penalized maximum likelihood estimation in agriculture

Abstract

Modelling agricultural participation is crucial for a comprehensive understanding of the agricultural sector, particularly in least developed economies where a large proportion of households rely on smallholder farming for their livelihood. Due to its importance, many households participate in agriculture in various ways along the stages of the value chain, and success often relies on several determinants. Climate change, rising input costs, alternative livelihoods, and changes in labour availability rank among the common predictors. This study utilises data from the household budget survey to explore these dynamics through Penalized Maximum Likelihood Estimation. We approach agricultural participation by examining various dimensions of agricultural output, which include output sold and household consumption, as well as production for processing and livestock feed consumption. Additionally, we consider factors such as land acquisition and farm asset ownership to provide a more nuanced understanding of participation. Our findings highlight the heterogeneity of agricultural participation, which is shaped by geographic location, household size, and income level. Importantly, the influence of these variables on participation evolves over time and differs across various forms of engagement, underscoring the need for tailored interventions to foster agricultural involvement. This holistic perspective reveals not only the multifaceted nature of agricultural participation but also the potential for diverse strategies to enhance engagement in the sector.