School Meals and Educational Outcomes in Rural Ethiopia

Abstract

We investigate the relationship between providing school meals programmes and educational outcomes in Ethiopia. Using data from school catchment areas across rural Ethiopia, the paper examines the role played by programme modalities and their implementation. The results indicate that supplementing on-site school meals with take-home rations can be beneficial for concentration, reading, writing and arithmetic skills. The timing of the distribution of school meals is also found to play an important role.

Acknowledgements

We thank the editors and two anonymous reviewers for helpful comments and suggestions. We thank conference participants at the German Economic Association’s Research Committee on Development Economics (AEL) Conference 2012 and the 2012 CSAE Conference, Oxford University, and seminar participants at the UN-WFP Country Office in Addis Ababa, the Ethiopian Development Research Institute (EDRI) and University of Mannheim for valuable comments on earlier versions of the paper. Robert Poppe also had extensive and valuable discussions with Nzinga Broussard while on secondment at EDRI, which helped to improve the paper. Markus Frölich acknowledges support from the special research area SFB 884 of the German Science Foundation (DFG). Data and STATA files to replicate the empirical results will be made available on request.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Other interventions to attract children to school that have been found to increase school enrolment and attendance include deworming (Miguel & Kremer, 2004), provision of additional teachers (Duflo et al., 2008) and conditional cash transfers (Behrman, Parker, & Todd, 2009).

2. A body of literature investigates the impact of school meals on (short-term) cognitive development, focusing on the specific micronutrient content of school meals. Although the empirical evidence is mixed, there appears to be a consensus on the importance of animal source food. For example, Whaley et al. (2003) explore the effect of three different diets (meat, milk, and energy), suggesting that animal source food has greater impact on cognitive function. Similarly, Gewa et al. (2009) investigate the effect of different school meals comprised of exclusively vegetarian meals, milk, or supplemented with meat; results show that the meat variant is relatively more important in terms of improving cognitive function among school-age children. However, most of these studies are conducted in a laboratory setting, which limits their external validity.

3. The Afar and Somali regional states were not included in the surveys due to security and logistical challenges at the time.

4. In Tigray and Amhara WFP’s school meals programme is operational in highland areas only. Additionally, there was a smaller subsample of phased-out programme school areas, that is of schools which had received meals only in the past but not now. These schools are not included in the analyses of this paper since the information on the timing and modalities of the school meals programme was rather scarce. Furthermore, in this paper we focus on the link between current school meals and learning outcomes. For phased-out schools the treatment status is imprecise since some children might have received some meals in the past but there is no precise measurement of such partial treatment status and its timing.

5. For a more detailed description of the survey design see Haile, Poppe, and Frölich (2011).

6. The survey adapted tests on reading, writing and arithmetic skills from the Young Lives project, a longitudinal study conducted in four countries (Peru, Ethiopia, India and Vietnam). http://www.younglives.org.uk.

7. For a description of the test design see Poppe (2014).

8. The livestock index is a weighted index using tropical livestock units (TLU) as weights as follows: cattle are weighted by 0.7 TLU, donkeys or horses are weighted by 0.3 TLU, goats or sheep are weighted by 0.15 TLU and poultry are weighted by 0.05 TLU.

9. This index is defined as the sum (range 0–4) of whether sanitation facilities are available, school buildings are in a good condition, the school compound is fenced and classrooms have glass windows.

10. Robust standard errors are used to adjust for school catchment area cluster.

11. In an earlier version of the paper we had also estimated random-effects models where random school service area effects were included in the linear model. Overall, the main results were similar with, as expected, somewhat larger precision. Due to space constraints, these estimates are not reported here.

12. In the regressions, we control for whether a child is aged between 7 and 10 years because the survey administered different tests for younger and older children – except for the Raven’s test and the concentration test which were administered irrespective of age – as in small samples the distribution of younger children might be unequal across programme status.

13. We use normal kernel, with logit specification of the propensity score and the inbuilt bandwidth-choice algorithms.

14. We also used the random-effects model. Because the random-effects model produces similar results, only results from OLS and PSM are presented.

15. In the sample of programme school catchment area boys the score has a mean of 136.7 and a standard deviation of 56.6. See Table A.1 in the Supplementary Materials.

16. In the sample of programme school catchment area girls, the score has a mean of 12.8 and a standard deviation of 3.5. See Table A.1 in the Supplementary Materials.

Additional information

Funding

This work was supported by the German Science Foundation (DFG).