重新审视一项多站点早期儿童教育项目：渐进式扩面、生育间隔与父母投资

中文摘要

我们重新审视了一项多站点、随机化、高质量的早期儿童教育项目，该项目因以早产低出生体重儿童为对象而无意中过度抽取了双胞胎样本。我们发现，生育间隔决定了项目的有效性。对于单胎儿童，直至十八岁的结果改善在各站点间均可复制，且与佩里学前项目（Perry Preschool Project）等标志性项目所观察到的改善相当。对于双胞胎，至十八岁时改善转为负值。项目对父母投资（养育行为）的影响中介了结果差异；项目提高了单胎儿童的养育水平，却降低了双胞胎儿童的养育水平。一个家庭生产模型表明，向双胞胎联合供给投入的可能性解释了养育行为的下降。

Abstract

We reexamine a multisite, randomized, high-quality early childhood education program that unintentionally oversampled twins by targeting premature low-birthweight children. We find that birth spacing determines program effectiveness. For singletons, outcome gains up to age eighteen replicate across sites, and are comparable to gains observed for iconic programs like the Perry Preschool Project. For twins, gains become negative by age eighteen. Program impacts on parental investment (parenting) mediate outcome differences; the program increases parenting for singletons and decreases it for twins. A household production model indicates that the possibility of jointly supplying inputs to twins explains the decrease in parenting.