The objective of this dissertation is to present empirical evidence and analysis of three key issues in the Mexican education system: 1) school accountability, as reflected in a particular state innovation pursued by the state of Colima in 2009 to identify and address the problems of low-performing schools, 2) age delay and the effects of a national reform introduced in 2006-2007 that modified the first grade entry-age across all Mexican states, and 3) the educational disadvantages of indigenous peoples in México and their consequences, as determined from recent data which allows identification of this population. First, the dissertation evaluates the impact of a targeted state-sponsored intervention program known as Programa de Atención Específica para la Mejora del Logro Educativo (PAE) designed to provide low-performing schools with remedial resources in Colima, México. The research analyzes the effect of this compensatory program in terms of standardized test scores among 108 participating schools having the lowest learning outcomes in 2009. The results of this "natural experiment" confirm that intervention in the form of the PAE program had a positive impact on average test scores in poorly performing Colima schools. By exploiting PAE's eligibility rules, a regression discontinuity method is used to estimate the impact on subsequent learning outcomes. Schools that participated in the program and a valid comparison group were followed for three years in order to compare their performance. The fact that the program was halted after only one year meant that the only realized interventions were those related to the program's preparation, which revolved around notifying schools as low-performing, identifying a school's main academic problems and devising a development plan to address those challenges. Yet, after only one year, test scores in PAE schools increased by 0.13 standard deviations vis-à-vis non-PAE schools and in fact, after three years, differences between the two groups of schools were no longer significant. Second, the dissertation explores the impact of exogenous variation in the age at which students enter school on education outcomes. Prior to the 2006-2007 school year, the cut-off day for school entry in Mexico had been September 1st. Since then, however, pupils aged 6 by as late as December 31 could start public school. Data related to this cut-off transition are reviewed and analyzed using a regression discontinuity method so as to estimate the causal effect of delayed school enrollment on math test scores. A two-stage least square (TSLS) estimator is used wherein the source of identification is the variation in 1st grade entry ages which resulted solely from differences in dates of birth. The results indicate that older students scored higher than younger students. The reform impacted the discrepancy between those regulated by the new cut-off dates and those regulated by the old cut-off date(s) by 0.30 s.d. (comparing the 1998-1999 cohort which entered school before the reform with the 2002-2003 cohort, which entered afterwards). The results also suggest age effects on education outcomes that are stronger for recent generations than for generations entering first grade prior to the reform. Because math scores have increased by 0.95 s.d. since the first administration of ENLACE in 2006, this result suggests that, at a minimum, moving the cut-off date by four months to December 31 did not have an adverse effect on mean math test scores. Finally, a sobering analysis of the educational outcomes of indigenous populations is conducted using data from Encuesta Nacional Ingresos y Gastos de los Hogares, ENIGH) which, for the first time in 2008 and then 2010 identified indigenous populations. The research finds that although the percentage of families in extreme poverty residing in municipalities where indigenous populations are concentrated dropped between 1992 and 2010, the gap in poverty rates between the municipalities where indigenous people concentrate and others remains huge, with extreme poverty in the former equal to 51.9% in 2010 and in the latter 16.9%. Because rates of return to education are estimated in this dissertation to be high in Mexico (around 10%, including those for indigenous populations), education is found to be essential in reducing the gulf in poverty levels by ethnicity. But the study shows that gaps in educational outcomes between indigenous and non-indigenous populations remain wide, whether in terms of average educational attainment, participation in Kindergarten, the percentage of students who are overage, and the average student achievement as measured by a variety of tests.

As part of the complex but intriguing question of what defines a highly qualified early childhood teacher (Kagan, Kauerz, & Tarrant, 2008), there has been a heated policy debate over whether to make a bachelor's degree (B.A.) the minimum education requirement for preschools' lead teachers in publicly funded programs (Zigler, Gilliam, & Barnett, 2011). A mixed and non-causal research base of the effect of a B.A. on preschool-teacher performance and child development outcomes is a partial source of the controversy (Kelley & Camilli, 2007; Early et al., 2007). Particularly, no experimental or quasi-experimental studies have been conducted for this topic (Barnett, 2011b). To fulfill the need for better causal inference, this dissertation first uses a nationally representative sample of American children born in 2001 who attended a preschool in 2005, to estimate the effect of having a lead teacher with a B.A. in preschool on the children's development outcomes assessed at aged 4, based on data from the Early Childhood Longitudinal Survey, Birth Cohort (ECLS-B). The term preschool is an umbrella term for all types of center-based programs. Based on three rigorous quantitative methods, including ordinary least squares with rich controls (OLS with rich controls) and two quasi-experimental methods (propensity score matching (PSM) and instrumental variables (IV)), this study finds: (1) In the model of OLS with rich controls, four of the eight comprehensive child development outcome constructs at age 4 are affected by teachers' B.A. status. Children with B.A. teachers are shown to exhibit higher early reading and math skills and fewer parent-reported internalizing behavior problems than children with non-B.A. teachers. No effects are found for story-telling skills, color recognition, parent-reported externalizing behavior problems or approaches to learning skills. Yet the children in the treatment group are reported by parents to have lower social competence. In PSM, B.A. positively predicts math skills and negatively affects social competence. In the IV estimates, a B.A. effect is only found for reducing parent-reported externalizing behavior. Comparatively, the PSM and IV estimates tend to be less statistically significant than the OLS estimates. This difference may be attributed to either bias or heterogeneity, given that the PSM and IV estimate may have removed some endogeneity of the treatment in a better way than OLS but they cannot represent the whole sample---the PSM estimate is for those matched and the IV estimate is only local to compliers. Further, when comparing teachers who have just a B.A. (as opposed to a B.A. or higher) with teachers who have an associate's degree (A.A.), the B.A. is found to have fewer statistically significant effects in the model of OLS with rich controls. Significant effects are found for two outcomes: Having a teacher with a B.A. increases math skills and reduces internalizing behavior problems. (2) There has not been much evidence of differential effects by preschool type, and the B.A. effects are no larger for children from low socioeconomic status (SES) families. Neither does the specialized education in early childhood education (ECE), as measured by whether a teacher has a degree in ECE or a related field and the number of college courses in ECE, interplay with the B.A. effect. (3) The supplemental analysis that uses two steps regression to link B.A., teacher-child interactions and child outcomes also returns some interesting findings. The treatment B.A. is found to increase the frequency of several classroom activities and the quality of teacher-child interactions (i.e., being more sensitive, less harsh, less detached and less permissive); but the two steps of the analysis only provide slight evidence for the mediating role of teacher-child interactions. Overall, there is some positive evidence of B.A. effects on children's early reading, math, the reduction of parent-reported internalizing behavior problems, the reduction of internalizing behavior problems and positive teacher behavior for the center-attending children in the ECLS-B dataset. Still, the evidence is not very strong given the inconsistency of findings across models and the negative effect of B.A. on parent-reported social competence. Such findings identified by rigorous methods in this study speak directly to the B.A. debate by adding a new piece of empirical information for a new generation of children and teachers; it adds some positive evidence to the pro side. Still, for future research and practice that aim to elevate quality, a full picture of cost-effectiveness and cost-benefit of the B.A. threshold policy is recommended and other teacher quality components should be considered.

Recent research suggests that the female advantage in educational attainment is driven in part by the differential effect of family background characteristics on the noncognitive skills of males relative to females. Building on this research, this study provides new evidence that links family characteristics and gender differences in noncognitive ability and school performance. Data are drawn from the NLSY79 Child and Young Adult Surveys. Multilevel modeling is used to examine how family context relates to gender differences in adolescent externalizing behavior, and how family context relates to gender differences in externalizing behavior and high school grades. Results indicate a strong relationship between externalizing behavior and grades that is not explained by the female advantage in grades. Results also indicate that males are differentially affected by family context and suggest that the pathways through which family structure, noncognitive ability, and school performance operate are different for boys relative to girls. A primary conclusion is that boys’ externalizing behavior is more dependent upon family background characteristics. Findings suggest the need to address both the school and family environments by formulating policies that promote the development of noncognitive skills in school as well as those that remedy family disadvantage in the home.

Education is often perceived as a key to development and growth, consequently, in the last decades, many countries have increased education coverage in all education levels. The creation of international education quality measurement programs, such as OECD’s Programme for International Student Assessment (PISA) or the Trends in International Mathematics and Science Study (TIMSS), have further led to a focus on improving education quality. For these reasons, the last decades have seen an increase in the importance of education in the political debate. This has been particularly relevant in Latin America, where, additionally, education has been used, sometimes, to fight traditionally high levels of income inequality, with a significant rise in education expenditure and coverage. The evaluation of education policies that aim to increase education coverage, quality or equity is, however, generally difficult. Many education policies are large-scale policies and are likely to affect all students or workers in the population, even those not directly benefiting from the policy. For example, students not participating in some education policy could still experience changes in their classmate characteristics that could affect their achievement. The presence of possible spillovers may change the direction of the effects of large-scale education policies when all the population is included in the analysis. Therefore, analyzing solely the effects on students participating in the policy may not give a complete picture of the effects of large-scale education policies. This dissertation focuses on the effects that three large-scale education policies that aimed to improve education equity, quality and coverage, respectively, had on students and workers affected differently by the policies. Particularly, each chapter analyzes the aggregate effects for the population of each education policy and decomposes these effects on the impact suffered by different groups of students or workers. In Chapter 1, I analyze the effects on test scores of a policy that aimed to increase education equity in Chile. I study the effects of an increase in school choice for low-income students by examining a 2008 reform that made the value of Chile’s (previously flat, universal) school voucher a step function of student income. This policy increased the proportion of private schools that low income, eligible children could access free of charge from 0.5 to 0.7. In order to identify aggregate effects and the impact within groups of students, I combine the introduction of the policy with variation from a date of birth enrollment cutoff for 1st grade. I show that the differentiated voucher lowered the probability that students used public schools by a small fraction and that these students shifted out of low achievement public schools to enroll in low achievement private schools. Nonetheless, private schools where these students enrolled had better test scores and socioeconomic composition at baseline, and less experienced teachers and smaller class sizes than public schools where they would have enrolled in the absence of the program. Despite the improvement in some school observable characteristics, I do not find any increase in test scores for students more likely to move to private schools. Further analysis suggests a rise in test scores for students most likely to stay in public schools. These results suggest that the policy had an overall modest positive effect on test scores, but that this positive effect was caused by responses from public schools instead of by students responding to the increase in school choice. In Chapter 2, I study the impact on test scores of a policy that aimed to improve education quality by increasing transparency of school performance in Chile. Particularly, I look at the effects of the distribution of school performance information to all families in Chile in 2011. Since I am interested in identifying effects for different groups of students, I define a control group within each group of students by using variation in enrollment year. Due to the presence of a date of birth enrollment cutoff for 1st grade enrollment there is variation in enrollment year for students born a few days apart. I combine this variation together with the timing of the distribution of information to identify the effects of the policy. I show that the distribution of information increased enrollment in high-performing schools, particularly for students in the third quartile of the municipality socioeconomic distribution. Thanks to this policy, students in the third quartile were exposed to a better socioeconomic composition of peers. Test score results suggest that there was an overall positive effect on verbal test scores, particularly for students in the third quartile, seemingly caused by an improvement in peer characteristics. However, there does not seem to be any significant change in test scores for students less likely to change enrollment decisions in response to the new information. Finally, Chapter 3, examines the effects of a policy that increased tertiary education coverage in Colombia on wages. I identify the effects on the distribution of wages using two different empirical strategies: the DiNardo, Fortin and Lemieux (1996) reweighting method and a differences-indifferences strategy. My results suggest that the overall distribution of wages remained constant, once labor demand shifts and productivity changes are taken into account. In contrast, wages increased for workers that were not at the margin of studying tertiary education, workers with primary education or less, and the density of wages at high levels of the distribution decreased for high school and tertiary education graduates. However, there were no effects on average wages for workers with any of the education levels. These results suggest that the policy had heterogeneous effects within the wage distribution and between education levels that were not captured by changes in average wages. These three chapters show that large-scale education policies can, sometimes, have effects on achievement or wages of students that are not participating in the policy, and that these effects are not always visible in the aggregate effects. Therefore, policy-makers and researchers should take into account the presence of spillovers or strategic responses when designing or analyzing large-scale education policies.

Essay 1: Determinants of NCLEX-RN Success Beyond the HESI Exit Exam: Performance in Nursing Courses and Academic Readiness Abstract: Every year, nursing students are not allowed to receive their degrees or sit for NCLEX-RN (national licensure exam) because of their performance in standardized exit exams like the HESI (Health Education Systems, Inc) or ATI (Assessment Technologies Institute). These exits exam have been found to be highly predictive of NCLEX-RN success but not failure. Consequently, some have argued that progression rules in nursing programs should not be based on a single metric but on a broader assessment of student readiness to pass the NCLEX-RN. In this study, I examine whether demographics, pre-college academic readiness measures, and performance in nursing courses improve the correct identification of students who will pass/fail above and beyond the HESI exit exam. I find that their inclusion can improve the identification of those who will fail but not those who will pass. Essay 2: The Impact of Remediation on NCLEX-RN Success: Positive, Neutral, or Negative? Abstract: Nursing programs assign students to NCLEX-RN remediation - based on the results of the HESI exit exam- to improve their probability of passing the NCLEX-RN on the first attempt. Previous correlational studies have found NCLEX-RN remediation based on HESI exit exam scores to be effective. In this study, I use two nationally representative samples to explore the impact of required remediation on passing the NCLEX-RN on the first attempt using regression discontinuity design both as local randomization and as continuity at the cutoff using. As the former, I find some evidence that remediation has a negative impact on passing the NCLEX-RN on the first attempt. As the latter, I find limited evidence of positive treatment effects. Both sets of statistically significant findings, however, are sensitive to bandwidth, kernel functional assumptions and/or sample trimming. Overall, RD evidence suggests that remediation may not have an impact on NCLEX-RN outcomes. Essay 3: Testing a Rule of Thumb: For STEM degree attainment, More Selective is Better Abstract: The supply of STEM workers depends to some degree on the ability of post-secondary institutions to keep those students already interested in STEM engaged and, to a much lesser extent, to generate interest among those initially not interested. The institutional attributes which may exert positive or negative influences on STEM degree attainment are many and students and parents may not be able to assess the status of each factor or a bundle of factors for specific institutions in their college choice sets which may maximize the probability of STEM degree attainment. In this essay, I test the rule of thumb that, for STEM students, attending a highly selective institution instead of a moderately selective institution improves the probability of obtaining a STEM degree at the first attended institution among those interested in STEM among and among those who are not initially interested. Overall, I find that highly selective institutions have a comparative advantage in producing STEM graduates among those already interested in STEM but not among those initially not interested in STEM. These findings hold true also for female students but to a different degree.

It is challenging for central authorities to change the nature of teaching and learning despite great efforts to do so through both command and commitment-style approaches, as well as through recent reforms aimed at teacher quality. Capacity-building initiatives, particularly those that engage educators in structured collaboration within and across schools, hold promise for school improvement. These approaches engage educators to develop contextualized solutions to the specific obstacles facing their students. However, without certain conditions in place, particularly a strong professional community that prioritizes continuous improvement, collaborative activities are unlikely to yield positive outcomes, and instead produce variable results based on schools’ capacity pre-intervention. This dissertation explores New York City’s attempt to build schools’ capacity for improvement through the Learning Partners Program (LPP), a program developed under the Chancellor Fariña administration in 2014. LPP combines interschool collaboration, inquiry, teacher leadership, and teaming, thus providing an opportunity to explore how a district can foster the prerequisite conditions necessary for school improvement. Drawing on qualitative data from 3 years of implementation, I provide a detailed portrait of the elements of the program in practice, and describe the processes and conditions that allowed some schools in the program to implement coordinated changes that moved them towards their improvement goals. These descriptions can provide lessons to other districts interested in taking on similar “learning organization” reforms, or reforms that generally aim to shift the teaching profession towards one that fosters self-examination and continuous improvement.

Due to the variety of factors that may affect student achievement, individual students often come to schools with different levels of academic preparation. These students from diverse academic background come with different learning needs. So, to better serve them, schools have adopted a variety of strategies, including increasing instructional time, reducing class sizes, providing differentiated curriculum and improving teacher quality through professional development trainings. My dissertation consists of three papers that examine several education interventions targeting at students with different academic abilities. These studies examine the design and current uses of several popular education interventions, and provide actionable insights on improving these interventions to enhance learning experiences for students at different points of the achievement spectrum. Chapter one evaluates a multi-subject remedial program that provided additional instruction on math and reading to under-performing students in a large metropolitan school district. Using a fuzzy regression discontinuity design and school administrative records, I find that double-dose remedial math courses improve math test scores by 0.21 standard deviations at the end of the school year. Yet, the effect of double-dose reading courses on reading achievement is small and statistically insignificant. In addition, the required extra classes in math or reading do not have crowd-out impact on instruction time and student performances on non-targeted core subjects. Finally, the study shows that students who receive treatments in multiple subjects do not necessarily have larger gains. While double-dose math courses may improve students’ math skills, the effect disappears for students who were taking double-dose reading courses at the same time. This finding suggests that two separate double-dose courses in different subjects may be ineffective in improving student achievements. To support students who struggle with more than one subject, schools should consider redesigning the double-dose courses with alternative curriculum and instructional strategies to integrate content of different courses and to increase student engagement. Chapter two examines effects of taking accelerated math courses under a subject-based acceleration program for middle school students. Students assigned to accelerated courses were exposed to more advanced curriculum and higher-performing peers. Using school administrative records and fuzzy regression discontinuity approach, the study finds null effects of taking accelerated math courses on students’ end-of-grade math test scores. Specifically, the effects are insignificant for students who took accelerated courses in both math and English Language Arts, and for those who took accelerated courses only in math. Also, the effects are insignificant for low-income and minority students. These findings are unexpected given the treatments provided by accelerated courses. The study provides possible explanations to the findings, and suggests directions for future research. Chapter three examines the sustainability of teacher knowledge gains from teacher professional development (PD) programs. Teacher PD is seen as a promising intervention to improve teacher knowledge, instructional practice, and ultimately student learning. While research finds many instances of significant program effects on teacher knowledge, little is known about how long these effects last. If teachers forget what is learned from the professional development program, the contribution of the intervention will be diminished. Using a large-scale dataset with 3,340 in-service teachers from 161 programs, this study examines the sustainability of gains in teachers’ content knowledge for teaching mathematics (CKT-M). Results show that there is a negative rate of change in CKT after teachers complete the training and that this estimated rate is relatively stable over time, suggesting that the average gain in test scores before and after the program is lost in just 37 days. There is, however, variation in how quickly knowledge is lost, with teachers participating in summer programs losing more rapidly than those who attend programs that occur during school years. The implications of these findings for designing and evaluating professional development programs are discussed.

The public school system in the Philippines adopted a multilingual policy starting in school year 2012-2013. Implemented as the Mother Tongue-Based Multilingual Education program, the policy localizes the language of early literacy and instruction, the content of the curriculum, and the support systems. This signals a shift away from the country’s long history of standardized curriculums with their one-size-fits-all approach and Western, colonial paradigms, and language policies with their preferential treatment of the colonizer’s language (English) and/or national language. This dissertation describes and assesses how the program was implemented in its first year. The focus is on resources or capacity to implement the program since the literature on implementation studies identifies the availability of these as a factor affecting the passage from policy to local practice. Using the case study design in qualitative research, one division and two elementary schools under it were purposively selected for this inquiry. All Grade 1 teachers in the two schools along with their curriculum consultants or instructional leaders (namely, their principal, district supervisor, and the division supervisor) comprised the study participants. Semi-structured interviews and document review were conducted over a period of one year to generate data. Theory-generated typologies from the conceptual framework as well as emergent themes or categories from real-life data guided the process of data organization and analysis. The findings suggest patterns that tended to reinforce institutional continuities rather than institutional shifts. Critical resources or capacities for change were not productively provided to the local implementers. Additionally, the resource or capacity gaps appeared to be underpinned by a lingering command or hierarchical structure. To get past the weight of this status quo, recommendations are offered. These include policies aimed at strengthening the practice of teacher expertise and instructional leadership, and at modifying the administration of the program to align the school language with the home language.

New Jersey is one of only three states in the United States offering a paid maternity leave benefit beyond temporary disability insurance at the present time. Understanding the impact of state maternity leave policies on low-income mothers is of particular urgency, since previous research suggests this group is less likely to utilize paid leave than wealthier women. In addition, existing literature also suggests that in a poverty environment, with its already existing vulnerability to social, emotional, cognitive, and health impairments, rapid return to work postpartum may be particularly damaging to the physical and emotional health of both mothers and their babies. This study examines the effectiveness of the New Jersey law mandating payments to postpartum mothers who were employed before giving birth. Using a mixed methods approach, outcomes from high- and low-income mothers were compared regarding the usage and impact of Family Leave Insurance (FLI), New Jersey’s paid family leave policy. For the quantitative study, data from 497 postpartum mothers from the Center for Disease Control’s 2012 Pregnancy Risk Assessment Monitoring System (PRAMS) data set were analyzed. For the qualitative study, three high- and three low-income mothers from the same New Jersey county were interviewed in depth, and six narrative profiles were constructed. Data from the qualitative and quantitative analyses were combined. Consistent with existing literature, quantitative results suggested poor FLI utilization. However, there was no significant association (p > .05) between low-income status (household income under $22,000) and FLI usage. A statistically significant (p < .05) association between FLI usage and postpartum depression in low-income mothers was noted. Qualitative findings supplemented and explained the quantitative results. The qualitative data suggested policy underutilization stemmed from poor public awareness due to inadequate publicity, lack of community education, and poor advisement on the part of human resource personnel and New Jersey Department of Labor of Workforce Development call-in center advisors. Results also suggest that if parameters of FLI and the awareness of the policy remain the same, the benefits existing research associates with paid maternity leave will not be fully reaped by New Jersey families.

This dissertation addresses the question: How does an attempt to redesign instructional delivery using technology-based personalization affect the technical core of teaching, learning, and student outcomes? In recent years, many prominent educators, business leaders, and philanthropists have suggested that schools be redesigned to personalize students’ learning experiences using technology. However, the justification for these reforms remains largely theoretical. Empirical research on technology-based personalization is sparse, and what little research does exist focuses predominantly on macro effects rather than the specific school-level, class-level, student-level, and lesson-level mechanisms that contribute to overall student achievement. The absence of research that pushes inside the “black box” of implementation is particularly problematic given a century of failed attempts to reform the technical core of instructional delivery, with symbolic reforms typically withering in the face of institutional resistance. This study attempts to address that gap by examining the implementation of an innovative model for using technology-based personalization to deliver middle school math instruction. I draw upon theoretical tools from institutional theory, instructional improvement, and the history of educational reform to deepen our understanding of how technology-based personalization affects the role of students and teachers, the logistics of content delivery, and students’ learning outcomes. Unlike previous studies in K-12 settings, which typically use summative assessments and virtual control groups to estimate aggregate effects on student learning, this study examines the relationships among a diverse set of lesson-level variables, including instructional modality, instructional content, group size and composition, teacher characteristics, student characteristics, and learning outcomes. In doing so, this study contributes to our understanding of the on-the-ground processes and mechanisms by which technology- based personalization affects (or does not affect) student learning. Although the instructional model documented in this case study will remain anonymous, it is well known and respected among educators and philanthropists, and regarded as one of the most prominent and archetypical examples of technology-based personalization currently active in American schools. Using multiple methods, including novel applications of hierarchical linear modeling, cluster analysis, and heatmap data visualization, I explore: (a) the degree to which ground-level implementation of technology-based personalization represents an authentic departure from the traditional technology of schooling, and (b) the relationships among various elements of the model and student learning outcomes. I draw on longitudinal data from a full year of implementation in five schools, including the daily lesson assignments and assessment scores of 1,238 unique students supervised by 48 teachers. This study supports four main findings: (a) the program succeeds in altering the technical core of instruction in several fundamental ways; (b) policy and logistical constraints limit the program’s ability to reform the technical core of instruction to the degree that it aspires; (c) students who enter the program as already higher-performing are more successful on daily exit slips than students who enter the program with lower performance; and (d) the quantitative methods used in this paper represent useful and replicable tools for exploring the data produced by technology-based and personalized models.