Inference for Average Treatment Effects

评估政策干预（如职业培训计划）的因果效应是许多应用经济学研究的一个重要目标。通常一个合理的起点是假设对治疗的分配是随机给出的足够详细的观察到的预处理变量。在这个假设下，我们可以确定总体平均效应。本研究将分三部分对这些假设下的平均治疗效果的推断文献做出贡献。首先，研究将发展匹配估计的大样本理论。通过匹配估计器，我们的意思是每个处理单元与一个或固定数量的控制匹配，并且每个控制与一个或固定数量的处理单元匹配的估计器。这种纯匹配估计具有相当的直观吸引力，在实践中得到了广泛的应用，除特殊情况外，其大样本理论尚未建立。结果应该是这种匹配估计的渐近理论，使研究人员能够在实践中使用这些估计。在第二部分中，研究将研究一些已经提出的平均处理效果估计器的高阶性质。其中许多估计量都有非参数分量。然而，除了速率条件外，大多数文献对平滑参数的实际选择保持沉默。这让它变成了。让实践者能够实际实现这些估算器。这里的计划是开发一个基于均方误差的准则，为平滑参数导出一个明确的数据驱动准则。在第三部分，研究将比较一些平均治疗效果的估计。到目前为止，已经提出了许多估计器，通常有一个小型的模拟研究来调查它们的性质。本研究完成的是对各种估计器的系统比较。在许多社会项目的研究中，如职业培训项目，观察数据被用来评估这些项目。这种评估的统计方法通常依赖于匹配类型的方法，将受训人员与相似的控制相匹配，即接受培训的个体与未接受培训的个体具有相似的背景特征和劳动力市场历史。目前使用了各种这样的方法，但这些方法的性质和可靠性通常是未知的。本研究探讨了这些方法的形式性质。此外，该研究将开发自动化程序来实施其中一些方法。目前，这些方法通常要求研究人员在实施过程中做出许多选择，这些选择可能会对最终结果产生重大影响，而没有太多的指导来指导这些选择。这将使这些方法更透明，更容易实现。最后，该研究将在已知正确答案的情况下比较一些这些方法，以便评估它们的性能并提出建议，以告知这些方法的未来使用。

Estimation of causal effects of policy interventions such as job training programs is animportant goal of much applied economic research. Often a reasonable starting point is theassume that assignment to the treatment is random given on sufficiently detailed observedpretreatment variables. Under that assumption one can identify the population averageeffect. This research will contribute to the literature on inference for average treatments effects underthese assumptions in three parts. First, the research will developlarge sample theory for matching estimators. By matching estimators we mean estimatorswhere each treated unit is matched to one or a fixed number of controls, and each controlis matched to one or a fixed number of treated units. Such pure matching estimators haveconsiderable intuitive appeal and have been used widely in practice, without their largesample theory having been established other than for special cases. The result should bean asymptotic theory for such matching estimators that allows researchers to use theseestimators in practice. In a second part, the research will investigate higher order propertiesof some of the estimators for average treatment effects that have been proposed. Manyof these estimators have a nonparametric component. However, most of the literature issilent regarding the actual choice of smoothing parameters, beyond rate conditions. Thismakes it di .cult for practitioners to actually implement thse estimators. Here the plan is todevelop a mean-squared-error based criterion to derive an explicit data-driven criterion forthe smoothing parameter. In the third part, the research will compare a number of theestimators for average treatment effects. So far a number of estimators have been proposed,often with a small simulation study to investigate their properties. What this research accomplishes isa systematic comparison of various estimators.In many studies of social programs such as job training programs observational dataare used to evaluate these programs. Statistical methods for such evaluations often rely onmatching type methods that match trainees to similar controls, that is individuals whoreceived the training to individuals who did not receive the training with similar backgroundcharacteristics and labor market histories. A variety of such methods are currently used, withoften the properties and reliability of such methods unknown. This research investigates theformal properties of such methods. In addition the research will develop automated procedures forimplementing some of these methods. Currently these methods often require the researcherto make a number of choices in the implementation that potentially affect the final resultssubstantially, without much guidance available to guide these choices. This should makethese methods more transparent and easier to implement. Finally, the research will compare a numberof these methods in settings where the correct answers are known so as to evaluate theirperformance and reach recommendations to inform the future use of such methods.