Abstract: With the counter-factual causal inference framework，this paper employs various propensity score matching methods to explore the effect of selectivity of migration on the fertility level of floating population in China，and estimates and compares migrant and non-migrant fertility levels after controlling for the selectivity bias. Both lifetime fertility( children ever born) and period fertility( births in the year proceeding the census) are examined using data from the 2000 census. Heckman 2-stage model shows that the observed difference between migrant and non-migrant fertility is biased due to selectivity of migration. The consistent results from three types of propensity score matching methods demonstrate that migrant fertility is markedly lower than non-migrant fertility，and a larger proportion of the difference is explained by the selectivity. Anyway，the study suggests that migration does reduce fertility.

Keywords: Migration, Floating Population, Fertility Level, Selectivity, Propensity Score Matching

摘要： 文章借助反事实因果推论框架，利用倾向得分匹配方法以克服传统人口统计学的弱点，使用2000 年第五次人口普查数据，以曾生子女数和普查前一年的生育作为生育水平指标，尝试讨论流动人口的选择性对估计其生育水平的影响作用; 并在控制选择性后，比较流动人口与非流动人口的生育水平差异，以期为深入讨论人口流迁与生育的因果关系提供基础。Heckman 二阶段模型表明，观测到的流动与非流动人口生育水平差异由于选择性而存在偏差。三种倾向得分匹配方法得到一致
结果: 剔除人口流动的选择性后，流动人口的生育水平显著低于非流动人口; 且两者的差异大部分可以被选择性所解释。从某种意义上说明人口流动有助于降低生育水平。

关键词: 人口流动, 生育水平, 选择性, 倾向得分匹配