Characterizing Weights in the Measurement of Multidimensional Poverty: An Application of Data-Driven Approaches to Cameroonian Data

The study seeks to compare multidimensional poverty indices in Cameroon generated by different multivariate techniques. After carefully exploring the theoretical and empirical review of the statistical methods of setting weights in the measurement of multidimensional poverty, the study employs three different statistical or data-driven methods – principal components analysis, multiple correspondence analysis, and fuzzy set approach to set weights in the aggregation procedure. Use is made of the 2001 Cameroonian household survey data to estimate the models. The poverty distributions obtained from the three approaches are submitted to stochastic dominance tests to investigate the sensitivity of the resultant poverty index rankings to changes in the weighting characterization. It comes out of the empirical analysis that the principal components analysis index distribution unambiguous shows less poverty than the multiple correspondence analysis and fuzzy set composite indices, while comparison of the two latter index distributions shows no clear dominance.

Citation: Njong, A.M. and Ningaye, P. (2008). 'Characterizing weights in the measurement of multidimensional poverty: An application of data-driven approaches to Cameroonian data', OPHI Working Paper 21, Oxford Poverty and Human Development Initiative (OPHI), University of Oxford.