摘要
基于Alkire和Foster (A-F)提出的多维贫困指数框架理论,运用中国家庭追踪调查数据(CFPS),分别采用等权重法、变异系数法和BP神经网络法选取多维贫困指标体系中各指标权重,测度和分解我国农村地区的多维贫困指数,并采用加总误差法建立权重优劣的评判标准。研究结论表明:三种权重选取法所测度的多维贫困指数差异较大,其中多维贫困指数从大到小是按变异系数法、BP神经网络法和等权重法所测得的,说明不同的权重法对多维贫困指数具有不同强度的敏感性;同时,多维贫困指数分解结果显示,不同权重法所得各指标对多维贫困指数的贡献率不同,但按变异系数法和BP神经网络法所得指标贡献率的排名基本一致,而按等权重法所得指标贡献率排名与另外二者相差较大。此外,相较于另外两种权重法,运用BP神经网络法测度的多维贫困指数更精确。本文为测度多维贫困指数时指标权重选取提供了新思路。
Based on the multidimensional poverty index( MPI) framework theory proposedby Alkire & Foster( A-F),using the China Family Panel Studies( CFPS) data,respectively,by equal weight method,variation coefficient( VC) approach and BP neural networks method to select the weight of each index in the MPI system,it measures and decomposes the MPI in rural China and evaluation criterion of weight is established by error method. The research results show that three kinds of weight selection method to measure the MPI differences,the MPI from large to small according to weight VC method, BP neural network,that is,the MPI has different intensity sensitivity to different weights. Meanwhile,the MPI decomposition results show that each index from different weights of the MPI contribution rate is different,but according to the VC method and BP neural networks method the contribution rate index ranking is basically the same,and according to the weight of the index contribution rate ranking and the other two are significantly different. In addition,compared to the other two weight methods,the multidimensional poverty index measured by the BP neural network method is more accurate. This paper provides a new idea for the selection of index weights in MPI.
出处
《数量经济研究》
CSSCI
2018年第2期47-60,共14页
The Journal of Quantitative Economics
基金
国家社会科学基金重点项目“巨灾风险管理机制设计及路径选择研究”(12AGL008)
西南大学决策咨询项目“社会化预期与公共情绪管理”(2016SWUJCZX09)
中央高校基本科研业务费专项资金项目“社会资本与农户多维贫困:作用机制与影响效应”(SWU1709413)的联合资助