Evaluating Education: An Economic Analysis of Education in Rural Zimbabwe

Loading...
Thumbnail Image
Date
Authors
Nordstrom, Ardyn
Keyword
Education , Evaluation , Economics
Abstract
This thesis uses a combination of innovative and existing evaluation techniques to understand the barriers to education in rural Zimbabwe, and to determine the impact of interventions that targeted some of these barriers. Chapter 2 uses a randomized controlled trial combined with text mining analysis of qualitative interviews to demonstrate the impact of a large-scale community mobilization campaign. After three and a half years of exposure to the community mobilization campaigns, community attitudes towards girls' education improved by 0.56 SD, and struggling students performed 0.28 SD better on learning assessments. Using mediation analysis, I show that the increased support likely partially mediated the program's impact on education. Chapter 3 exploits variation from a natural experiment to determine the impact of a severe drought on girls' education outcomes. I show that the drought decreased the opportunity cost of education, leading to an increase in enrolment. However, this did not correspond with an increase in learning, with students from particularly vulnerable households performing worse on learning assessments after the drought. Chapter 4 provides a descriptive analysis of the most important factors associated with education outcomes. By using multiple machine learning methods, I show that lack of support and pregnancy are the biggest barriers to student advancement. These are also barriers to student learning, however, other factors such as self-confidence were relatively more important for learning outcomes. These findings describe the programmatic contributions of this thesis, which provides important insights into the types of barriers students experience, as well the effectiveness of specific education interventions. The second contribution is more methodological, where I show how natural language processing and machine learning methods can be used in future evaluation research.
External DOI