流动人口调查抽样的实践与思考

Reflections on Sampling of Migrants Surveys

文章对流动人口调查中抽样框的编制、重复抽选及抽选后合并、目标与抽样总体关系的复杂化等问题进行了描述和讨论。实践表明,在编制流动人口调查抽样框时,研究人员不仅需要得到相应数据库的支持,而且还应特别注意偶然性因素的影响;对规模过大或过小的抽样单位,重复抽选或抽选后合并是比较简便的方式;此外,在以"登记户籍人口"作为目标总体的调查中,受"跨界流出人口"的影响,特定地区目标总体与抽样总体的关系将表现为"一对多"的形式,数据处理时必须进行"事后加权"。为保证"事后加权"有效性,需要考虑不同省(区、市)人口流动的特点,特别是样本中"省际"和"省内"流动人口的规模。"2013年全国生育意愿调查"数据处理的过程表明,不同省(区、市)"事后加权"的分组变量及分组数量并不完全相同,具有一定的灵活性。

This paper describes and discusses issues on the establishment of sampling frame, replicate selection and linking after selection, the relation between target population and frame population in migrants surveys. Experience shows that when preparing the sampling frame of migrants surveys, researchers not only need the support of relevant database, but also should pay particular attention to the effects of accidental factors. For oversized or undersized sample units, a relatively simple approach is to replicate selection, or link sample units after selection. In addition, in a survey whose target population is the ＂Registered-household Population＂, the relation between target population and frame population in specific areas will appear as ＂one-to-many＂ as a result of impacts of cross-border migrant outflows. Thus, post-weighting is necessary when processing data. To ensure the effectiveness of the postweighting, the characteristics of population flows in different provinces should be taken into account, especially the size of floating population both across and within province in the survey sample. Data processing of 2013 National Fertility Desires Survey indicates that the grouping variables and number of groups of post-weighting in different provinces are not exactly the same, but subject to some flexibility.