Central Data Catalog

Citation Information

Type Working Paper - CRC Discussion Papers, Courant Research Centre Poverty, Equity and Growth, Universität Göttingen
Title An alternative strategy to identify deprivations in multidimensional poverty: A partial least squares approach
Author(s)
Issue 271
Publication (Day/Month/Year) 2020
URL http://hdl.handle.net/10419/219035
Abstract
This study determines data driven weights for the indicators in the multidimensional poverty index (MPI), based on partial least squares (PLS), using income as the outcome variable. Consequently, the resulting MPI is particularly useful to income related policy and research questions. An innovative data driven procedure is proposed to determine the first cut-offs of the MPI inside of the PLS algorithm, which provides an alternative to the first cut-offs based on researchers' judgement. Another adjustment to the PLS procedure enables the weights to respect the existing practice in the MPI literature, that health, education and living standard dimension are equally important. The new MPI can consider heterogeneous observations by means of interaction terms in the weighting structure. Using this approach, a new MPI is created considering the additional deprivation of the black population in South Africa, compared to other racial groups. This MPI shows different weights and first cut-offs than the old MPI. It suggests that the first cut-offs for the year of education indicator needs to be 12 years instead of 5 years to have a practical relevance for the South African context. Additionally, the weight of assets is important and electricity less so. The black population shows higher deprivation for all considered deprivation indicators within the MPI using interaction terms.

Related studies

»
»
»
Yoon, Jisu, and Atika Pasha. "An alternative strategy to identify deprivations in multidimensional poverty: A partial least squares approach." CRC Discussion Papers, Courant Research Centre Poverty, Equity and Growth, Universität Göttingen , no. 271 (2020).
Copyright DataFirst, University of Cape Town