Mapping Disaggregate-Level Agricultural Households in South Africa Using a Hierarchical Bayes Small Area Estimation Approach

Abstract

The first important step toward ending hunger is sustainable agriculture, which is a vital component of the 2030 Agenda. In this study, auxiliary variables from the 2011 Population Census are combined with data from the 2016 Community Survey to develop and apply a hierarchical Bayes (HB) small area estimation approach for estimating the local-level households engaged in agriculture. A generalized variance function was used to reduce extreme proportions and noisy survey variances. The deviance information criterion (DIC) preferred the mixed logistic model with known sampling variance over the other two models (Fay-Herriot model and mixed log-normal model). For almost all local municipalities in South Africa, the proposed HB estimates outperform survey-based estimates in terms of root mean squared error (MSE) and coefficient of variation (CV). Indeed, information on local-level agricultural households can help governments evaluate programs that support agricultural households.