The Study Of Weights Separating Comparison On Multidimensional Poverty Index Measuring

The study on poverty which measuring from multiple dimensions has been drawn more and more attentions than measuring from single income dimension.When computing the multidimensional poverty index,it is very important to take weights separating for each sub-indicators into consideration.Different weights separations will lead to different measurements of multidimensional poverty,therefore every measurement will lose some conviction inevitably.For this reason,a study on the comparisons of different weighting methods is rather important and meaningful.For comparing different weighting methods,this paper selected five methods applied in multidimensional poverty study including whole indicators equal weight,dimensions equal weight,principal component analysis,multidimensional correspondence analysis and fuzzy set analysis.After that,I computed the multidimensional poverty index for each 12 area with all five weighting method respectively based on the CHNS survey data 2011 edition.Then for each weighting method,ordering the multidimensional poverty index in ascend.So that producing five order sequences.Meanwhile,I applied a simulation experiment and take the stable mean of cumulative ordering to order the areas again,and take the consequent ordering sequence as the standard order.At last,I set an error function to compute each of five sequences' total errors by comparing with standard sequences.Obviously,the ordering sequence which has more total errors shows less conviction and therefore the corresponding weighting method is less recommendable.After all the study above had been done,it showed that fuzzy set analysis and multiple correspondence analysis received the biggest total error as 12.Meanwhile whole indicators equal weight had smallest total error as 4 and principal component analysis' s total error is 6.Based on the result above,this paper suggest not to use fuzzy set analysis or multiple correspondence analysis to separate weights when researching multidimensional poverty and use whole indicators equal weight or principal component analysis instead.