如何准确地拟合居民的收入分布

How to Fit Residents' Income Distribution Precisely

中文关键词:  分布函数  城镇居民  收入分布  广义矩估计  极大似然估计  Hitomi估计

英文关键词:distribution function  urban residents  income distribution  Generalized Method of Moments (GMM)  Maximum Likelihood Estimation (MLE)  Hitomi estimation

中文摘要:

      采用分布函数拟合居民收入分布需要解决两个关键问题,即找到恰当的分布函数与合适的估计方法。文章基于国家统计局提供的2011年城镇居民收入分组数据以及安徽省统计局提供的微观城镇住户调查数据,运用常见的六种分布函数依次拟合了中国城镇居民的收入分布。随后,根据不同的检验标准,比较了新估计方法、极大似然估计以及广义矩估计的拟合效果。研究发现,广义第2类beta分布函数对中国居民收入分布的拟合效果最好,同时,极大似然估计方法的参数估计误差相对较小。不过,在选择具体方法时,还应考虑实际的数据结构以及对计算结果的偏好。

英文摘要:

      Employing distribution functions to fit real income distribution requires to deal with two key issues:appropriate distribution function and suitable method of estimation. Based on the grouped Chinese urban residents' income data in 2011 provided by National Statistics Bureau and the micro survey data of urban residents provided by Anhui Statistics Bureau, this paper uses 6 common distribution functions in turn to fit Chinese urban residents' income distribution. Furthermore, in light of different testing standards, it compares the fitting effects of Hitomi's new method of estimation, Maximum Likelihood Estimation (MLE) and Generalized Method of Moments (GMM). The research finds that the generalized beta distribution function of the second kind has the best fitting effect on Chinese resident income. At the same time, MLE has a relatively small error in parameter estimation. However, the actual data structure and the preference to calculation results should be considered in selecting specific methods. 