Essays on the economics of education
Chapter 1 estimates the impact of dual language and immersion education on student achievement. Dual language classrooms provide English Language Learners (ELLs) an opportunity to receive instruction in their native language as they transition to English fluency. This might allow ELLs to build a stronger foundation in core subjects and lead to better academic outcomes. Dual language and immersion education have also grown substantially in popularity among English speaking families across the U.S., as they present an option to learn content in, and presumably become fluent in, a second language. Despite the spike in practice, there is little causal evidence on what effect attending a dual language school has on student achievement. I examine dual language and immersion education, and student achievement using school choice lotteries from Charlotte-Mecklenburg School District, finding local average treatment effects on math and reading exam scores of more than 0.06 standard deviations per year for participants who were eligible for English second language (ESL) services or designated limited English proficient (LEP). There is also some evidence that attending a dual language school led to a lower probability of having limited English proficient status starting in third grade. For applicants who were not eligible for ESL services or designated as LEP, attending a dual language school has resulted in higher end of grade exam scores of about 0.09 and 0.05 standard deviations per year in math and reading, respectively.Chapter 2 builds upon recent research discussing apparent gender differences in returns to parental investments. Parental time investments are one potential mechanism, if correlated with household structure differentially by gender, that could help explain documented gender differences in non-cognitive skills, as well as the sensitivity of outcomes and behavior to family structure for boys. This paper investigates gender differences in parental time investments around changes in household composition. I find that, although both boys and girls experience decreases in parental time investments following a change in family structure from a two-parent to a single-parent household, the loss for boys is relatively large. The difference is strongest in paternal weekday investments, for which boys lose an additional 24 minutes per day, equivalent to roughly 35\% of average paternal weekday investments. There is no significant evidence that mothers compensate for the loss by increasing their investments to boys relative to girls.Chapter 3 discusses the construction of confidence intervals in teacher value-added (VA). The use of teacher value-added models to measure teacher effectiveness is expanding rapidly, with teacher value-added estimates being incorporated into teacher evaluation systems and potentially high-stakes decisions. However, we still know little about the precision of value-added estimates, or the performance of the resulting confidence intervals, which could play an important role in the decision-making of districts and policymakers. Our study aims to fill a gap in the literature by providing comparisons of confidence interval performance for the OLS-Lag (DOLS) value-added estimator. We use simulated student achievement data to study the behavior of standard errors and confidence intervals for teacher value-added estimates under several student grouping and classroom assignment scenarios. We propose a simple method for calculating confidence intervals, which includes a critical value adjustment, and compare it's performance to that of the confidence intervals you get from using a typical variance estimator with standard normal critical values. We find that this method generally leads to 95% confidence interval coverage rates near 95%, which we consider a desirable feature. In particular, the method has advantages over standard confidence interval calculations when value-added estimates are based on a small number of classrooms per teacher and students are grouped on unobservable heterogeneity. We then use student-level administrative data to compare standard errors and confidence intervals across the different methods. Our proposed method assigns 18.5% of teachers in the top decile of the value-added distribution confidence intervals with a lower bound above the 75th percentile of the distribution, a much lower percentage than that from ignoring correlation in unobservables, 60%, or using cohort-school clustering, 75.8%. These differences indicate that the policy conclusions drawn from using the value-added estimates may depend on the choice of confidence interval. Hence, knowledge of the reliability of value-added models could be a critical part of the decision making process of administrators and policy makers.
Read
- In Collections
-
Electronic Theses & Dissertations
- Copyright Status
- In Copyright
- Material Type
-
Theses
- Authors
-
Bibler, Andrew Jacob
- Thesis Advisors
-
Elder, Todd
- Committee Members
-
Dickert-Conlin, Stacy
Imberman, Scott
Mavrogordato, Madeline
- Date
- 2016
- Subjects
-
Immersion method (Language teaching)
English language--Study and teaching--Foreign speakers
Education, Bilingual
Parent and child
Scheduled tribes in India--Education
Education
Teacher effectiveness
Evaluation
- Program of Study
-
Economics - Doctor of Philosophy
- Degree Level
-
Doctoral
- Language
-
English
- Pages
- ix, 147 pages
- ISBN
-
9781369368451
1369368453
- Permalink
- https://doi.org/doi:10.25335/tcrn-zf52