PROJECT 3:Unequal Effects? The Distributional Consequences of Edul. Interventions

过去20年是美国教育政策的实验期，旨在实现提高成绩和缩小弱势学生与其他学生之间的成绩差距的双重目标。几乎所有评价都侧重于估计这些政策对学生成绩的平均影响;有些评价还研究了这些平均影响在按种族和家庭社会经济地位等特征界定的学生分组之间的差异。正如在项目I-II中详细论述的那样，发展理论表明，学生环境与干预计划设计之间的良好和不良匹配将产生异质性计划影响，并且理解这种异质性的性质对于优化计划设计至关重要。项目l-ll提出了儿童/政策拟合假设的测试与互动模型，允许不同的计划影响，在人口亚组的基线特征定义。本项目开发和实施 测试儿童/政策拟合假设的补充方法：分位数治疗效果估计和其他分布估计。它将这些技术应用于从幼儿园到高中的四种不同教育环境的数据。在“先声夺人影响研究”和一个让贫困儿童能够进入纽约市私立学校的代金券项目中，我们检验了一些假设，这些假设表明，对于具有不同学术技能和行为特征的儿童，适度的总体影响掩盖了系统性影响差异。第三个分析将测试有关政策的影响，鼓励中学生参加代数对学生的动机和成绩的分布假设。 我们的最后测试是高风险的退出考试以及低风险（对学生）责任测试对努力和成就的影响分布。我们项目的第五年将涉及旨在促进在教育评价中更广泛地使用这些技术的外联活动。

The last two decades have been a period of experimentation in education policy in the United States, aimed at achieving the twin goals of boosting achievement and closing achievement gaps between disadvantaged and other students. Nearly all evaluations have focused on estimating the average impacts of these policies on student outcomes; some have also examined how these average impacts vary across subgroups of students defined by such characteristics as ethnicity and family socioeconomic status. As argued at length in Projects l-ll, developmental theory suggests that good and bad matches between student circumstances and intervention program design will produce heterogeneous program impacts, and that understanding the nature of this heterogeneity is essential for optimizing program design. Projects l-ll propose tests of child/policy fit hypotheses with interaction models allowing for differential program impacts across population subgroups that are defined by baseline characteristics. The current project develops and implements complementary approaches to testing child/policy fit hypotheses: quantile treatment effect estimation and other distributional estimators. It applies these techniques to data from four different education settings that range from pre-kindergarten to high school. In the case ofthe Head Start Impact Study and a voucher program to enable poor children to attend private schools in New York City, we test hypotheses suggesting ways in which modest overall effects mask systematic impact differences for children with different academic skills and behavioral characteristics. A third analysis will test hypotheses regarding the effects of policies encouraging middle schoolers to enroll in Algebra on the distribution of student motivation and achievement. Our final tests are of the distribution of impacts on effort and achievement of high-stakes exit exams as well as low-stakes (to students) accountability tests. The fifth year of our project would involve outreach activities designed to promote the use of these techniques more broadly in education evaluations.