The Effect of School Safety on Academic Achievement: Evidence from Rwanda, Tanzania and Zambia

• In Event: Evidence-based knowledge to improve quality education in the Middle East and sub-Saharan Africa

Proposal

Introduction: In recent years, school safety has gained increasing attention from policymakers and academics. A safe learning environment is a place where structured learning is free from environmental, internal, and external threats to learners and educators’ well-being; where both the infrastructure of the organization and the people within that environment are deemed safe (USAID, 2016). Safe learning environments can be threatened by internal threats, such as bullying, corporal punishment, and gang recruitment, external threats, such as attacks on schools, and environmental threats, such as natural disasters. All these threats have the potential to significantly decrease students’ academic performance. While a growing body of research points to a connection between school environments and student outcomes, much remains unknown about the effect of perceived school safety on learning. Most evidence originates from middle and high-income countries and focuses on educational outputs, such as attendance and retention, rather than educational achievement. More quantitative analysis of the relationship between school safety and student performance in developing countries is needed yet absent. To such this literary gap, the objective of this study is to identify the causal direction and magnitudes of student and teacher perception of safety on learning outcomes in Rwanda, Tanzania, and Zambia through a quasi-experimental analysis.

Data: Data for this study is tabulated from USAID’s Global Reading Network: EdData Initiative. The EdData assessment collects primary surveys from developing country households, schools, and communities on issues of access to education, quality, and management. In this study, we gather information from three EdData assessments: Early Grade Reading Assessment (EGRA), the Early Grade Mathematics Assessment (EGMA), and the Snapshot of School Management Effectiveness (SSME). Overall, we synthesize information on 3,711 primary school students in Rwanda, Tanzania, and Zambia, collected between 2011 and 2013.

Limitations: The data set consists of student and head teacher reported perceptions of school safety. In Rwanda students responded if they felt safe in school; while in Zambia and Tanzania head teachers reported school safety on behalf of students. The data set does not offer consistent and comparable variables across all countries, which does not accommodate comparisons across countries.

Variables: The key dependent variables in this study are standardized learning outcomes in English reading fluency and math addition problems. The key independent variables are the perception of school safety from students in Rwanda, and from head teachers in Tanzania and Zambia. The datasets also consist of control variables encompassing students, teacher, school, and family specific attributes. Table I summarizes the variables:


Methods: We initiate our estimation by conducting an Ordinary Least Square (OLS) analysis using alternate models encompassing student, teacher, school, and family specific characteristics. The OLS estimation provides an understanding of linkages between school safety and learning outcomes. Subsequently, we use a quasi-experimental design to estimate the effect of school safety on learning outcomes. Specifically, we apply Propensity Score Matching (PSM) and Doubly Robust Estimator (DRE) methods. In these matching mechanisms, we use the perception of school safety as “treatment” and match students based on their propensity to be selected into the treatment given background covariates, such as personal, family, and school-specific attributes. We match each participant in the treated group with similar participants in the control group and estimate the Average Effect of Treatment on the Treated (ATT) as the difference in mean outcomes between the two groups.

Findings

Student reported school safety: We find negative effects of an unsafe school environment on learning outcomes for reading and math in all aforementioned estimation procedures of Rwandan students, who self-reported their perception of school safety. Results show that for 6th grade math evaluations, a student who feels unsafe solves seven fewer addition problems correctly per minute (36% deviation from average performance), compared to peers who feel safe at school. For 4th grade math evaluations, the difference is about two problems per minute (20% deviation from average performance) when compared to students of similar characteristics, who only differ through the perception of school safety. Also, 6th grade English reading fluency is significantly affected, with students who feel unsafe reading about five fewer words per minute than comparable peers (12% deviation from average performance).

Head teacher reported school safety: Negative effects are found both in regression and quasi-experimental estimation for Tanzania, where head teachers reported perceptions of school safety among students. Specifically, quasi-experimental estimation shows that Tanzanian students who reportedly had an unsafe learning environment solved 0.7 fewer problems correctly per minute in math addition (7% deviation from average performance). Also, English reading fluency is significantly affected, with students who had an unsafe environment reading about eight fewer words per minute than comparable peers (33% deviation from average performance). It is worth pointing out that an average Tanzanian student loses one-third of his/her reading efficiency (the mean being 24 words per minute) solely due to school safety issues. Particularly in the 2nd grade regression analysis for Tanzania, we find that female students perform worse than their male peers in English reading assessments, when both girls and boys are facing unsafe school environments. Also, the presence of security guards in unsafe school environments is found to reduce the negative consequence on 2nd grade math outcomes. For Zambia, where a head teacher reported safety, we do not find statistically significant effects of school safety, except for 3rd grade English reading. The quasi-experimental estimations show an 11% reduction from the average performance in English evaluations.

Conclusion:

We draw three major and consistent conclusions from our analysis. First, the effect of student-perceived school safety on academic performance in Rwanda is significantly negative. Second, similar with the findings in Rwanda, the effect of head teacher-perceived students’ school safety on academic performance in Tanzania is significantly negative. However, in Zambia, where head teachers report the perceived school safety for students, we do not find a statistically significant effect of school safety on most learning outcomes. Third, there is a need for consistent data collection that would enable researchers a. to offer comparable analysis across countries and b. to conduct longitudinal and/or experimental studies.

Authors

• shahriar kibriya, Texas A&M University

• Song Zhao, Texas A&M University

• Naureen Fatema, McGill University