Causal Inferences in Education Politics and Policy PDF Download

Author: Alexander Karl Mayer
Publisher:
ISBN: 9781267029133
Category :
Languages : en
Pages :

Get Book Here

Book Description
Education is among the most important and widely used explanatory covariates in the social sciences, yet we know little about causal effects related to educational advancement or educational policies. Deficits in our knowledge generally derive from the infrequent use of randomized controlled trials and the considerable challenges associated with causal inferences using observational data. This dissertation focuses on two causal questions whose answers have eluded social scientists for nearly fifty years. Both questions involve mechanisms to improve outcomes for the disadvantaged. The obstacles to causal inference for these questions, moreover, are common to questions pertaining to education. Research designs that overcome these obstacles may inform a variety of related research questions. The first question asks whether higher education causes political participation. The second asks whether Title I funding -- additional federal funding for K-12 schools with high concentrations of low-income students -- improves academic performance. Both of these questions have profound implications for politics and public policy in the United States. Individuals with higher education participate in politics at higher rates and consequently have greater influence over political outcomes. Household income correlates highly with children's educational success, and the United States currently spends over $13 billion annually on Title I programs to improve education for low-income students. Recent studies suggest that neither intervention is effective, but this dissertation demonstrates that they neglect to properly evaluate assumptions necessary to support such conclusions. The following chapters re-evaluate the research designs, identify flaws in the studies, demonstrate that the conclusions are premature, and provide evidence for positive effects in each case. Methodologically, this dissertation emphasizes matching and regression discontinuity (RD) designs for causal inference with observational data. These methods are becoming increasingly popular for causal inferences in the social sciences, yet they are not as generally applicable as recent use would imply. This dissertation reviews the identifying assumptions behind each method, and employs a variety of tests to examine the plausibility of these assumptions and provide a firmer foundation for causal inferences. It also identifies flaws in previous applications of these methods to the substantive questions considered here. Moreover, these flaws are not uncommon in the social sciences. In the case of RD applications, few studies that conclude interventions were ineffective consider the statistical power of their designs. Through a series of new analyses using data from three randomized controlled trials, this dissertation also demonstrates how failure to consider statistical power -- a common occurrence in RD applications -- can often lead to incorrect interpretations of null findings. Overall, this dissertation provides evidence for positive effects of higher education on political participation, and for Title I funding on academic performance. This dissertation also demonstrates limits of regression discontinuity designs that have been widely overlooked, and it presents rigorous examples of matching applications -- with detailed post-matching analyses -- to support causal inferences in two important substantive areas at the intersection of education and politics.