Measuring Multidimensional Poverty in Kenya Across Time and Space

Abstract

This study aims to assess Kenya’s progress in the reduction of poverty in the context of achieving sustainable development goals (SDGs). We start with a standard set of indicators from the Alkire-Foster (AF) multidimensional framework which encompasses multiple facets of the SDGs but begin by interrogating their appropriateness. We start by assessing the reliability of the indicators using the item response theory and the reliability of the overall model using McDonald’s Omega statistic. We find that the overall model is not reliable. In particular, we find that the child mortality indicator is not reliable thus we drop it from the analysis. We introduce indicators measuring child schooling gaps, household overcrowding, and financial inclusion which make the poverty measurement model reliable. We assess the validity of the AF framework using the confirmatory factor analysis. We find that the most valid model is one with equal non-normalized weighting of the indicators which excludes the child mortality indicator. This is in contrast to the standard AF framework which uses equal weights per domain and equal weights for the dimensions within each domain. We thus estimate the multidimensional model AF framework with additional indicators, with equal weights and without the child mortality indicator. We use the finite mixture model, latent class analysis and negative binomial frameworks to estimate an optimal poverty threshold and find the negative binomial framework fits the data best. The optimal threshold classifies an individual as poor if they are deprived in 7 or more of the 12 indicators. This contrasts with the standard AF framework in which an individual is considered poor if they are deprived of a third or more of the weighted indicators. Using our index, we then go on to profile multidimensional poverty over the period 2014 to 2022. The results show that Kenya has experienced remarkable improvements in the poverty situation as shown by significant reductions in the poverty headcount and multidimensional poverty index (MPI) at the national, sub-group and spatial levels between 2014 and 2022. A decomposition of the MPI shows that the largest contributor to multidimensional poverty is the living standards dimension while the education dimension makes the least contribution. The provision of cheaper clean cooking fuel, electricity and low-cost housing is imperative since the lack of access to these indicators is the largest contributor to the MPI.