Parental Education and Increased Child Survival in Madagascar: What Can We Say?

ABSTRACT

This paper investigates the relationship between parental education and child mortality in Madagascar. Until recently, most research linking parental education and child mortality had overlooked the case of Sub-Saharan Africa, despite the region having a high childhood mortality rate and a low association between parental education and child survival. Adopting a careful empirical strategy based on availability of schooling infrastructure and internal instruments, this paper contributes to the literature by analyzing the role of both the father’s and mother’s education as well as different educational levels. The results demonstrate that children’s survival probabilities increase when they have a mother with at least primary schooling. Controlling for wealth reduces the effect of mothers’ education by only one-third. In contrast, fathers’ education does not play a significant role in child survival.

HIGHLIGHTS

• Parental education is strongly associated with improvements in child health in many countries.

• Father’s education is not a strong determinant of child survival in Madagascar.

• Higher levels of maternal education increase child survival in Madagascar.

• Wealth only accounts for one-third of the total effect of maternal education.

• Increasing education levels especially for women will likely reduce child mortality in future generations.

KEYWORDS:

• Child mortality
• mother’s education
• instrumental variables
• Lewbel instruments
• Sub-Saharan Africa
• Madagascar

JEL Codes:

• I12
• I2
• J1

ACKNOWLEDGMENTS

This work has greatly benefited from the comments and suggestions of my colleagues and former colleagues at Monash Business School, at GATE L-SE CNRS Lyon 2, and Thema. I would like to thank Olivier Bargain, Nicole Black, Sonja de New, Sylvie Demurger, Olivier Donni, Christelle Dumas, Eliane El Badaoui, Flore Gubert, Rachel Knott, Estefania Santacreu-Vasut, Karinna Saxby, and Philippe De Vreyer for useful comments on this manuscript and earlier versions.

SUPPLEMENTAL DATA

Supplemental data for this article can be accessed at https://doi.org/10.1080/13545701.2021.1937265

Notes

1 The survey was designed to evaluate a multi-million project of community development funded by the World Bank. The survey collection was organized by the FID (Fonds d’Intervention et de Dévelopment, http://www.fid.mg/) and the INSTAT (Malagasy National Institute of Statistics, https://www.instat.mg/). See Online Appendix 1 for more details.

2 A study on the quality of self-reported childhood health status and living conditions shows that individuals (older than those in this study) reported these characteristics well (Havari and Mazzonna 2015).

3 Only parents with a child in the household were asked about their childhood (that is, our instruments). To avoid over-selecting first-time mothers who did not lose their child, the sample is restricted to women with at least two live births. The description for women who ever had a live birth is similar, except that the probability of experiencing child loss is 3 percentage points lower.

4 This does not preclude differences in the original family.

5 Additional data description including perinatal care information are provided in Tables A2–A3.

6 The linear probability model (LPM) was chosen for the binary outcome because it captures the conditional expectation function and in the case of IV regressions, marginal effects from non-linear models are similar to LPM (Angrist and Pischke 2009). The comparison with the second outcome becomes easier as both outcomes use linear estimations.

7 Peter Glick, Christopher Handy, and David Sahn (2015) find no direct effect of education on age at first birth in Madagascar, although it delays marriage age, which delays first birth. Age at first birth may not be a mechanism, and could be driven by preferences. If so, it will threaten identification. Therefore, the next section presents instrumental variables.

8 Table A4 reports within and between standard deviations.

9 Older mothers are as strongly affected by the instruments as younger mothers (see Online Appendix).

10 Both the White test and the Breusch-Pagan test reject homoscedasticity at 1 percent in all specifications.

11 See Stata ivreg2h (Baum and Schaffer 2012).

12 The partial R (Shea 1997) was too low to simultaneously instrument both parents’ education.

13 The coefficient for mother’s education in panel A Table 7 decreases from column 1 – which does not control for wealth – to column 3 – which controls for wealth – by one-third : 1−(−0.007/−0.011)=0.36.

14 This holds whether or not wealth is controlled for. When wealth is controlled for (column 3), adding litter away (not shown) does not significantly change the coefficient for mother’s education and the coefficient for litter away is still not significant. The introduction of the variable no toilet in column 4 reduces the coefficient for mother’s education by less than one-tenth. It is approximately one-twentieth if wealth is also controlled for (not shown).

15 Censorship stems from the sample design and should be orthogonal to fertility decisions. Controls for mother’s age capture age-related patterns, including the fact that fertility is better captured by the number of children as age increases. The appendix presents results for younger and older mothers separately.

16 Table A2 presents between and within community variation.

17 Table A8 displays how maternal education predicts perinatal care variables and other potential mechanisms.

18 See Table A8-panel C.

19 Excluding them avoids considering that they experienced no child death when it is actually unknown whether their child will survive to age 5. However, excluding them is not random. It removes women with healthy recent births whereas women who recently lost a child cannot be removed.

Additional information

Notes on contributors

Samia Badji

Samia Badji is Research Fellow at the Centre for Health Economics at Monash Univeristy. She studies human capital investment, in particular education and health, with a focus on the most vulnerable populations. She is currently looking at health and healthcare use of people with disability.