Individual and School Effects On Academic Performance of Youth in the Ghana Youthsave Experiment
Method: This study used the YouthSave baseline data. 6,252 youth attending 100 schools from eight of Ghana’s ten regions constituted the sample. The outcome variables are math and english examination scores. The highest possible score for both subjects is 50. Because of the hierarchical nature of educational data, we used hierarchical linear modeling (HLM) in the analysis (Raudenbush & Bryk, 2002). We conducted a two-level HLM to explore the between-school variability and the effects of individual and school characteristics on average student examination scores. One-way ANOVA with random effects was estimated to determine intraclass correlation. The intercepts-and-slopes-as-outcomes model was estimated to examine the effects of individual and school characteristics.
Results: The average Math and English scores were 22.53 (9.65) and 23.04 (9.88), respectively. Thirty-seven percent of the variation in math and 32% of the variation in english scores are between schools. The remaining variation is due to differences among students within schools. Four individual-level characteristics (gender, age, self-efficacy, and home parental involvement) are statistically significant (p < .05) predictors of math exam scores. Five individual-level characteristics (gender, age, ownership of household possessions, self-efficacy, and home parental involvement) are statistically significant (p < .05) predictors of english exam scores. None of the school-level variables are a statistically significant (p > .05) predictor of either math or english scores.
Conclusion and Implications: This study illustrates the role of individual- and school-level characteristics on academic scores of Ghanaian youth. The variance decomposition suggests that most of the variation is within schools; however the between-school variation is above 0.25 which indicates the importance of using HLM (Heinrich & Lynn, 2001; Guo, 2005). The study finds evidence to support that there are substantial associations between individual-level predictors and youth’s academic achievement. Some of the findings are consistent with prior studies in Ghana and other developing countries that have shown association between individual characteristics and academic performance (Etsey, 2005; Lockheed, Fuller, & Nyirongo, 1989). Contrary to prior studies in developing countries (Fuller and Clarke, 1994; Kasirye, 2009; Suryadarma, Suryahadi, Sumarto, & Rogers, 2006), we did not find statistically significant associations between school factors and academic performance.