生育机会成本及异质性——基于机器学习算法的发现

Childbearing Opportunity Costs and Its Heterogeneity:Findings Drawn from Machine Learning Method

生育机会成本对生育意愿和生育行为的重要性已被广泛认可,但量化研究仍处于探索阶段。把生育机会成本视为一种“反事实”的生育行为的收入代价,运用机器学习算法测量女性的生育机会成本并考察生育机会成本的异质性,通过回归模型识别生育机会成本对生育意愿和生育数量的边际效应。结果显示,生育男孩的机会成本低于女孩;收入水平越高、受教育程度越高的女性,生育绝对机会成本和相对机会成本越低,这种异质性与能否获得生育保险密切相关;职业替换门槛低的女性群体,生育保险的覆盖率更低,生育机会成本更高。进一步的研究发现,生育绝对机会成本上升1单位,女性生育二孩和多孩意愿分别下降1.6%和1.5%,实际生育数量下降1.1%。研究发现对认识生育率发展趋势具有参考价值。

The significance of childbearing opportunity cost(COC)on fertility preference and fertility behaviors has been recognized widely,but the quantitative analysis is still under-developed.In this paper,fertility opportunity cost is regarded as a counterfactual income cost of fertility behavior,and the machine learning algorithm is used to measure women’s fertility opportunity cost and investigate the heterogeneity of fertility opportunity cost.Through regression model,the marginal effect of fertility opportunity cost on fertility intention and fertility quantity is identified.The results show that the COC for male births is lower than that for female births.The women with higher education income levels have lower obsolute and relative opportunity costs of childbearing.This heterogeneity is closely related to the access to maternity insurance.Low-income female laborers are largely outside the coverage of maternity insurance and have higher COC.The further analysis reveals that one unit of increasement in absolute COC will lead to the intentions to have two or more children decrease by 1.6 percent and 1.5 percent,respectively,and the actual number of births decreases by 1.1 percent.The findings make good reference for understanding the future fertility trend.