Skip to main content

Understanding the effects of siblings on child mortality: evidence from India

Abstract

Given the intrinsically sequential nature of childbirth, timing of a child’s birth has consequences not only for itself but also for its older and younger siblings. The paper argues that prior spacing and posterior spacing between consecutive siblings are thus important measures of intensity of sibling competition for limited parental resources. While the available estimates of child mortality tend to ignore the endogeneity of sibling composition, we use a correlated recursive model of prior and posterior spacing and child mortality to correct it. There is evidence that uncorrected estimates under-estimate the effects of prior and posterior spacing on child mortality.

This is a preview of subscription content, access via your institution.

Notes

  1. Although there are younger women in our sample who have not completed fertility, our estimates seem to be robust. Not only we include mother’s age at birth as a control variable in the spacing equation, but also our analysis focuses on middle-order children born to these women. In particular, we use hazard equations to determine prior and posterior spacing; oldest children are censored in the prior spacing equation, while youngest ones are censored in the posterior spacing equation.

  2. Rosenzweig (1986) noted the problem of estimating the health production function, given that spacing could be correlated with child-specific unobserved characteristics. Accordingly, he compared OLS estimates with fixed effects (FFE) and lagged instrumental variable fixed effects (LIFE). These results, as summarised in Table 3 of his paper (p. 69), suggest that there is a big difference between OLS and the other methods. FFE and LIFE estimates are qualitatively similar, although there are some evidences that the FFE may under-estimate the effect of spacing. Unfortunately, we do not have any suitable instrumental variables including the kind used by Rosenzweig. We note however that our correlated estimates are similar to the fixed effects estimates suggesting that the birth spacing effect is robust.

  3. Wolpin (1984) develops a finite-horizon dynamic stochastic model of discrete choice with respect to life-cycle fertility in a world where infant survival is uncertain and offers results for the number, timing and spacing of children for exogenous child mortality. We however choose to focus on Rosenzweig (1986) because of its direct relevance for our purpose.

  4. In other estimations, posterior spacing with respect to other subsequent siblings was never significant, and we do not therefore consider it in this analysis (see also Makepeace and Pal 2006 for a review).

  5. The number of sisters (or brothers) depends on the choice of family size and is therefore endogenous. Larger families tend to have more girls because fertility is endogenous with respect to child’s sex—families who have a target number of boys continue to have more children if they have girls early on but stop if they have boys. Thus, the probability of having a sister increases with the number of siblings. However, the gender of the first child cannot be correlated with the gender and other aspects of the second child although it is correlated with the number of children of a particular gender and can therefore be used as an instrument.

  6. The second NFHS undertaken in 1998-99 was designed to strengthen the database further and facilitate implementation and monitoring of population and health programmes in the country. Though some additional information (e.g., height and weight of all eligible women, blood test for women and children) were collected, the information that we use remained very similar. Our preliminary analysis also yielded similar results as reported here.

  7. We have also tried to include all children in our estimation. In this case, prior spacing for oldest child was estimated by the time between the mother’s age at marriage and the birth of the first child, while posterior spacing for the youngest child was the time elapsed between the birth of the child and the time of the survey (for non-sterilised couple) or the time the couple was sterilised. However, the log-likelihood function would not converge probably because of the poor quality of the available information (age at marriage, number of marriages or time of sterilisation), which in turn resulted in rather sporadic distribution of prior/posterior spacing of the oldest/youngest children in our sample. Note that the estimation of prior and posterior spacing hazard equations indirectly takes account of firstborn and youngest children as censoring variables.

  8. A short note on this is available from the authors on request.

  9. One, however, needs to be careful about the treatment of the twins and the corresponding birth order because birth order in our dataset is recorded in a continuous fashion without taking account of the twin birth. In this study, we have given the second born twin the same birth order as the firstborn.

  10. We however cannot analyse the effects of specific health inputs (e.g. prenatal care, hospital delivery or child vaccination) on child mortality (e.g. Maitra 2004) because these information were only collected for children born in the last 3 years (this holds for both rounds of NFHS).

  11. Information about the father was collected from the woman concerned. First, there were lots of missing as well as inconsistent values for father’s age. Second, most fathers were literate, which was in turn causing problems of convergence when included. Hence, we could not include comparable characteristics of the father as we did for the mother.

  12. The observations are grouped by mother, so the factor is strictly speaking mother specific. However, family breakups are extremely rare, so we interpret this more broadly as a family-specific effect.

  13. While we could treat mother’s age at first birth as an exogenous variable (evidence suggests that the use of contraception is almost non-existent before the birth of the first child), we could not ignore the element of simultaneity in mother’s age at birth of each child.

  14. Analysis of NFHS 1992–1993 data (see Appendix Table A1.2) suggests that compared to Hindus, a significantly larger proportion of Muslim couples use no contraception. In particular, compared to Muslim couples, more than double of the Hindu couples are sterilised, while use of modern (use of pills, IUD/copper, injections and condoms) or traditional (abstinence/withdrawal) non-terminal methods of contraception remains rather comparable among these two religious groups in our sample. Also see Pal and Makepeace 2003.

  15. To check the robustness of our correlated estimates, we have also re-estimated the correlated model after dropping the household-specific variables in case there is a correlation between household-specific observable variables and the unobserved heterogeneity terms. These new estimates remain rather similar to the ones presented in Table 4. For brevity, these estimates are not presented in this study but are available on request.

References

  • Becker G (1991) A treatise on the family (enlarged edition). Harvard University Press, Cambridge, MA

    Google Scholar 

  • Behrman JR, Pollak RA, Taubman P (1982) Parental preferences and provision for progeny. J Polit Econ 90(1):52–73

    Article  Google Scholar 

  • Benefo K, Schultz TP (1996) Fertility and child mortality in Côte d'Ivoire and Ghana. World Bank Econ Rev 10(1):123–158

    Google Scholar 

  • Brien MJ, Lillard LA (1994) Education, marriage, and first conception in Malaysia. J Hum Resour 29(4):1167–1204

    Article  Google Scholar 

  • Butcher C, Case A (1994) The effect of sibling sex composition on women’s education and earnings. Q J Econ 109(3):531–563

    Article  Google Scholar 

  • Chamberlain G, Griliches Z (1975) Unobservables with a variance components structure: ability, schooling and the economic success of brothers. Int Econ Rev 16(2):422–449

    Article  Google Scholar 

  • Choe MK, Diamond I, Steele F, Kim SK (1998) Son preference, family building process and child mortality’ in too young to die: genes or gender?’ ST/ESA/SER.A/155, pp 208–222. United Nations, New York

    Google Scholar 

  • Curtis SL, Diamond I, McDonald JW (1993) Birth interval and family effects on postneonatal mortality in Brazil. Demography 30(1):33–43

    Article  Google Scholar 

  • Dasgupta M (1997) Socio-economic status and clustering of child deaths in rural Punjab. Popul Stud 51(2):191–202

    Article  Google Scholar 

  • Drèze J, Sen A (1995) India: economic development and social opportunity. Clarendon Press, Oxford, UK

    Google Scholar 

  • Garg A, Morduch J (1998) Sibling rivalry and the gender gap: evidence from child health outcomes in Ghana. J Popul Econ 11(4):471–493

    Article  Google Scholar 

  • LeGrand TK, Phillips JF (1996) The effect of fertility reduction on infant and child mortality: evidence from Matlab in rural Bangladesh. Popul Stud 50(1):51–68

    Article  Google Scholar 

  • Lillard LA, Willis RJ (1994) Intergenerational educational mobility: effects of family and state in Malaysia. J Hum Resour 29(4):1126–1166

    Article  Google Scholar 

  • Maitra P (2004) Parental bargaining, health inputs and child mortality in India. J Health Econ 23(2):259–291

    Article  Google Scholar 

  • Makepeace G, Pal S (2006) Effects of birth interval on child mortality: evidence from a sequential analysis. World Health Popul January:1–14

  • Pal S (1999) An analysis of childhood malnutrition in Rural India: role of gender, income and other household characteristics. World Dev 27(7):1151–1171

    Article  Google Scholar 

  • Pal S, Makepeace G (2003) Current contraceptive use in India: has the role of women’s education been over-emphasized? Eur J Dev Res 15(1):146–169

    Article  Google Scholar 

  • Panis CWA (1994) The piecewise linear spline transformation with an application to age at first marriage. Stata Tech Bull 18:27–29

    Google Scholar 

  • Panis CWA, Lillard LA (1994) Health inputs and child mortality: Malaysia. J Health Econ 13(4):455–489

    Article  Google Scholar 

  • Panis CWA, Lillard LA (1995) Child mortality in Malaysia: explaining ethnic differences and the recent decline. Popul Stud 49(3):463–479

    Article  Google Scholar 

  • Rosenzweig M (1986) Birth spacing and sibling inequality: asymmetric information within the family. Int Econ Rev 27(1):55–76

    Article  Google Scholar 

  • Rosenzweig M, Schultz TP (1982) Market opportunities, genetic endowments and intrafamily resource distribution. Am Econ Rev 72(4):215–235

    Google Scholar 

  • Sen AK, Sengupta S (1983) Malnutrition of rural children and sex bias. Econ Polit Wkly 855–864

  • Sengupta S, Gazdar H (1997) Agrarian politics and rural development in West Bengal. In: Drèze JP, Sen AK (eds) Indian development: selected regional perspectives, WIDER studies in development economics. Clarendon Press, Oxford, UK

    Google Scholar 

  • Wolpin K (1984) An estimable dynamic stochastic model of fertility and child mortality. J Polit Econ 92(5):852–874

    Article  Google Scholar 

Download references

Acknowledgement

The first draft of this paper was written when Sarmistha Pal was based in Cardiff Business School, and she wishes to thank the Cardiff Business School for providing a research grant for this project. We are most grateful to editor Junsen Zhang and three anonymous referees of this journal for their very helpful and insightful comments. We would also like to thank Stephen Jenkins and Stan Panis for their helpful suggestions on the earlier version of this paper. The usual disclaimer applies.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Sarmistha Pal.

Additional information

Responsible editor: Junsen Zhang

Appendices

Appendix

Appendix 1: Variable definitions and summary statistics

Variable definitions

The data are taken from the NFHS 1992–1993 household data for West Bengal.

The dependent variables are:

  • 1 if died at age 5 or less

  • Time to death (in months) censored at

Regression variables

  • Continuous variables

    • Mother’s age at birth of first child

    • Length of time (in months) since the birth of the previous child

    • Length of time (in months) to the birth of the next child

  • Binary variables

    Mother is literate:

    1 if the mother is literate

    Twin:

    1 if the child is a twin or a triplet

    First child is female:

    1 if the first sibling in the family is a female

    First child is dead:

    1 if the first sibling in the family died

    Delivery problem in the previous birth:

    1 if delivery problem in the previous birth

    Radio:

    1 if the household owns a radio

    Television:

    1 if the household owns a television

    Agricultural land:

    1 if owns land

    Pucca:

    1 if lives in a brick house

    Muslim:

    1 if the family is Muslim

    Rural:

    1 if the child lives in rural areas

    Male:

    1 if the child is male

Table A1.1 Sample characteristics—means and standard deviations
Table A1.2 Current contraception use amongst various religious groups

Appendix 2

Table A2.1 Mortality hazard estimates
Table A2.2 Mortality probit estimates with discrete spacing variables
Table A2.3 Fixed effects logit estimates of mortality

Appendix 3. Full results for Table 3

Table 3.1A Estimates of the mortality probit equation
Table 3.1B Correlated and uncorrelated estimates of prior and posterior spacing, male
Table 3.1C Correlated and uncorrelated estimates of prior and posterior spacing, female

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Makepeace, G., Pal, S. Understanding the effects of siblings on child mortality: evidence from India. J Popul Econ 21, 877–902 (2008). https://doi.org/10.1007/s00148-006-0123-6

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00148-006-0123-6

Keywords

  • Sibling rivalry
  • Birth spacing
  • Endogeneity bias

JEL Classification

  • D13
  • I12
  • O15