Child Marriage, Education, and Agency in Uganda

ABSTRACT

This contribution relies on four different approaches and data sources to assess and discuss the impact of child marriage on secondary school enrollment and completion in Uganda. The four data sources are: (1) qualitative evidence on differences in community and parental preferences for the education of boys and girls and on the higher likelihood of girls to drop out of school in comparison to boys; (2) reasons declared by parents as to why their children have dropped out of school; (3) reasons declared by secondary school principals as to why students drop out; and (4) econometric estimation of the impact of child marriage on secondary school enrollment and completion. Together, the four approaches provide strong evidence that child marriage reduces secondary school enrollment and completion for girls with substantial implications for agency.

KEYWORDS:

• Child marriage
• education
• out of school children
• Uganda

JEL Codes:

• I21

ACKNOWLEDGMENTS

This paper was prepared for a research project on the economic cost of child marriage managed jointly by the International Center for Research on Women and the World Bank. The project benefits from generous support from the Bill and Melinda Gates Foundation and the Children Investment Fund Foundation. The opinions expressed in this paper are those of the authors only and need not represent the views of the World Bank, its Executive Directors, or the countries they represent, nor the views of the foundations that funded this work.

NOTES ON CONTRIBUTORS

Quentin Wodon is Lead Economist in the Education Global Practice at the World Bank, where he currently leads work on equity and inclusion in education, out-of-school children, and child marriage. He holds graduate degrees in business engineering, economics, and philosophy and PhDs in Economics and in Theology and Religious Studies.

Minh Cong Nguyen is Economist in the Poverty Global Practice at the World Bank. His research focuses on household poverty, inequality, and human development, as well as firm entry and exit, growth and productivity, using both household and enterprise surveys. He holds a PhD in Economics.

Clarence Tsimpo is Economist in the Poverty Global Practice at the World Bank. He is finalizing a PhD in Economics from the University of Montreal and has 15 years of experience in policy research and dialogue on applied development economics, with a focus on poverty and social impact analysis.

Notes

1 See the Convention on the Rights of the Child, the Convention on the Elimination of All forms of Discrimination against Women, the Universal Declaration of Human Rights, and resolutions of the UN Human Rights Council.

2 On the measurement of child marriage, see Minh Cong Nguyen and Quentin Wodon (2012).

3 In an estimation with pooled DHS data for Africa as a whole, Nguyen and Wodon (2015b) find endogeneity.

4 For two-stage regression models where the first stage is an OLS and the second stage is a probit, we are not aware of readily available tests to assess the validity of the instruments. But we tested the validity of the instruments based on a two-stage OLS estimation. The results suggest that the number of years of child marriage is weakly endogenous (Durbin–Wu–Hausman statistics over 1, p=0.3). This is also suggested by the statistical significance at 10 percent level of the residuals in the second stage regression. The joint F-test on the instrument variables is above 10, suggesting joint significance of the instruments. Stock and Yoyo's test suggests that the test statistics is much higher than the critical value (21 vs 3.8), indicating that the instruments are not weak. The over-identification test returns a Sargan–Basmann statistics at about 10, which is not statistically significant (p=0.4), suggesting that we can reject the null hypothesis of over-identification of an incorrectly specified structural equation.

5 We know to which PSU households belong, so we can compute mean values for the share of girls marrying at various ages in each PSU, both contemporaneous and in the past. The PSU leave-out-mean variables capture social norms at the PSU level as well as other factors that may affect child marriage. In order to avoid endogeneity, we compute leave-out-means for those variables, where the term leave-out-mean indicates that the PSU level variables are computed for all girls except the one considered in the regression. That is, for each girl/woman, the variables are computed among all the other girls/women living in the same PSU.