The effect of maternal education on early childhood development in Uganda: an instrumental variables approach

• In Event: Using Quasi-Experimental Approaches to Study Child Development in Low and Middle-Income Countries

Abstract

Theory and research in child development consistently link higher levels of maternal education to better developmental outcomes for children (e.g., Harding, Morris, & Hughes, 2015). Developmental science also often considers the importance of maternal education for child development as “universal,” or relevant across diverse cultural settings (M. Black et al., 2017). Despite this, most research on maternal education suffers from methodological problems that make it difficult to mitigate issues of selection bias (S. E. Black, Devereux, and Salvanes, 2005). Indeed, multifactorial inputs to child development impose significant challenges to estimating the causal effect of maternal education on child development, and controlling for observed covariates is not enough to discard issues of selection bias. Moreover, most prior research identifying positive associations between maternal education and child development has focused on samples taken from high-income countries (Duncan et al., 2015), and it is unknown whether these findings generalize to other settings with lower levels of education, different returns to education, and, in general, different institutions and cultures.

The objective of the current study is to estimate the causal impact of maternal education on children’s development in Uganda, a country characterized by low levels of formal education. A secondary aim is to examine potential mechanisms through which maternal education may impact child development, including maternal care, stimulation, and discipline and children’s enrollment in early care and education programming. To do so, I employ nationally representative data for 4,874 children taken from the 2016 Demographic and Health Survey (DHS). The DHS includes the Early Childhood Development Index and other information on parenting and children’s exposure to learning opportunities. To identify unbiased estimates, I use an instrumental variables approach to exploit exogenous cross-cohort variation in maternal education caused by an educational reform that abolished school fees for primary education in 1997. The overall idea behind the empirical strategy is simple: some mothers in the DHS sample were just young enough to have benefited from the educational reform when they were children, whereas others were just too old to be affected by the reform (see Figure 1 for trends in educational enrollment after the reform). Therefore, the reform works as a natural experiment, as it causes natural variation in education attainment that was very likely unrelated to other characteristics of mothers.

The findings suggest that maternal education had a small but statistically detectable positive effect on children’s development. Moreover, additional years of maternal education also increased maternal engagement in stimulating activities, children’s attendance in early childhood education programs, and reduced harsh corporal punishment (see Figure 2). Additional sensitivity and falsification tests assessing the internal validity of the results reveal the robustness of the estimates and further reinforce the argument that maternal education had a causal impact on children’s development in Uganda. I conclude by providing recommendations for policy, as well as by discussing how the use of secondary data and quasi-experimental frameworks may allow developmentalists to conduct rigorous research in under-represented contexts to better understand child development and parenting globally.

Group Authors

Jorge Cuartas (Harvard University)

Author

• Jorge Cuartas, Harvard University
  Presenting Author