Does access to services have a causal impact on children’s education in Peru? Evidence from panel data analysis

This paper investigates the impact of access to services on children’s education over a period of 15 years. Access to services is understood as “a measure of the household’s ability to meet functional requirements of sound shelter” (Briones, p10, 2017). It is a measure constructed as an equally weighted average of access to electricity, safe drinking water, safely managed sanitation and adequate fuel for cooking. The amount of time spent on studying both at school and outside school is used as a proxy for child’s education.

This paper analyses data from a longitudinal cohort study, Young Lives (2002-2016), that tracked 3000 children from two different cohorts in Peru over a period of 15 years. The study restricts itself only to younger cohort (as the older cohort has already started school) as the study patterns can be captured effectively right from the initial stages. Around 2,052 children aged 6-17.9 months were enrolled in the main (younger) cohort of the Young Lives study in Peru to be followed across 5 waves (Escobal and Flores, 2008). Young Lives employed a multistage, cluster-stratified random sampling to select households randomly from different sites (Escobal and Flores, 2008).

It is argued that more rigorous testing of causal relations can be carried out though a panel design and therefore are closer to experimental designs in the way in which they test causality than many other observational designs (Finkel, 1995). Therefore, this paper uses the rich panel data to exploit the variation across individuals and most importantly, the overtime ‘within’ individual variation to estimate the impact of ‘access to services’ on child’s education after controlling for other demographic characteristics.

Random effects and fixed effects models are constructed, along with year/round fixed effects, which show a high significant positive impact of access to services. The models were put to Hausman test to choose between the fixed effects and random effects estimators (Baltagi, 2003). As the hausman test favoured the fixed effects estimators, the paper includes discussion on the findings from the fixed effects model.

The next section of the paper also includes discussion on the consistency and unbiasedness of the estimators in light of potential threats to exogeneity, specifically omitted variable bias. In order to push for a more robust causal inference, instrumental variable approach is used to tackle the endogeneity problem. Household shocks, which have a completely exogenous character, are used as an instrument for access to services. Results demonstrate a very strong causal link. The evidence and analysis presented in the paper make a strong case from a policy perspective to prioritise the need for providing access to basic services which would improve children’s education over a sustained period of time.