Birth-spacing, fertility and neonatal mortality in India: Dynamics, frailty, and fecundity☆
Introduction
Interest in the determinants of child mortality has recently surged, with the inclusion of targets for child mortality amongst the Millennium Development Goals (Lawn et al., 2005; UNDP, 2003), and short birth-spacing and high fertility are widely regarded as among the most important causes of early childhood death. However, reproductive behaviour is endogenous to mortality and both are influenced by characteristics and choices of families, some of which are difficult to observe. For these reasons, there is limited evidence of the true causal associations of these variables.
In developing countries, 30% of deaths are of children under 5, compared to less than 1% in rich countries (Cutler et al., 2006). Almost half of child deaths are in the neonatal period, the first month of life, when the tie between mortality and fertility is closest (Cleland and Sathar, 1984). About 4 million neonates died in 2000, 99% of them in developing countries, and 27% in India. The proportion of neonatal in under-5 deaths has increased, since interventions like immunization, control of acute respiratory infection, or oral rehydration have had more of an effect on post-neonatal death (Lawn et al., 2005). It is thus important to focus attention on the causes of neonatal death.
Despite a long-standing interest of economists and demographers in the relation between childhood mortality and reproductive behaviour, the literature is scarce in a complete microdata analysis of all inter-relations of these variables (Wolpin, 1997). The main contribution of this paper is to use panel data based on retrospective fertility histories to estimate causal effects of birth interval length on subsequent neonatal mortality risk and of neonatal mortality on subsequent birth interval length, controlling for unobserved heterogeneity in both processes (referred to as frailty and fecundity, respectively). It also provides estimates of the effects of expected mortality (hoarding) and realized mortality (replacement) on fertility. Third, we model mortality dynamics within families, estimating both the extent to which observed persistence in death risk is explained by state-dependence, and the contribution of endogenously determined birth-spacing to state-dependence effects. Other contributions are methodological, relating to the way in which we deal with right-censoring of birth intervals and with the initial conditions problem that arises in dynamic models with unobserved heterogeneity.
Understanding the way in which biological and behavioural factors shape the family-level relation between reproductive behaviour and childhood mortality is crucial to understand the demographic transition that has historically preceded economic growth (Kalemli-Ozcan, 2002), and the endogenous processes by which societies evolve past the Malthusian spectre (Galor and Weil, 2000). Time-series analyses of historical data for today's industrialized countries suggest that declining mortality stimulated fertility decline (e.g. Ben-Porath, 1976; Eckstein et al., 1999), and a similar tendency can be seen in recent data for developing countries (e.g. Nyarko et al., 2003). Cross-sectional studies using household survey data have emphasized the reverse direction of causation, namely high fertility, associated with close birth-spacing or an early start to childbearing, causes an increase in childhood mortality (e.g. Cleland and Sathar, 1984).
In families with multiple children, there is a recursive bi-causal relation of these variables. The death of a child is often followed by a shorter interval to the next birth, which may be explained either by volitional replacement (e.g. Olsen, 1988) or by the fact that the mother stops breastfeeding, enabling her to conceive the next child sooner than otherwise (e.g. Chen et al., 1974). A short birth interval, in turn, increases the mortality risk of the next child in the family, possibly because the mother has not recuperated from the previous birth (e.g. DaVanzo and Pebley, 1993). Thus vulnerable families are caught in a death trap, creating persistence in death risk within families. This mechanism operates by the endogenous shortening of intervening birth intervals. Of course a birth interval is only observed if the mother has another birth, and this fertility decision is also influenced by whether her previous birth survived or not. While these relationships have each been studied, their interactions have rarely been studied jointly, and unobserved heterogeneity, another potential source of correlation of death risks within a family, is often ignored.
The analysis in this paper provides estimates, using survey data from India, of a dynamic panel data model that describes the complete process of child survival and birth-spacing (and thus fertility), allowing for endowment heterogeneity, input endogeneity, right-censoring and accounting for the initial conditions problem. We find evidence that childhood mortality risk is influenced by the pattern of childbearing, that is, by the timing and spacing of births, and that birth-spacing and fertility are, in turn, a function of realized mortality. We find a replacement effect of 0.37, in line with the few available estimates in the literature. The results suggest that the full impact of family planning interventions extends to reducing mortality and that mortality-reducing interventions like provision of piped water also affect birth-spacing and fertility. Our finding of causal effects of sibling mortality on both mortality and reproductive behaviour implies that interventions that reduce mortality or lengthen birth intervals will have multiplier effects.
The paper is organized as follows. Section 2 summarizes related research. Section 3 describes the data. The econometric model is presented in Section 4 and estimation and simulation results are reported in Section 5. Section 6 summarizes and concludes.
Section snippets
Related literature and contributions
Previous demographic research provides estimates of some of the main effects analysed in this paper, although not in a unified framework: for example, see Curtis et al. (1993), Madise and Diamond (1995), Hobcraft et al. (1985), and Whitworth and Stephenson (2002) for analysis of the effects of birth-spacing on mortality, and Olsen (1988), Zenger (1993) or Frankenberg (1998) for analysis of the effects of mortality on birth-spacing. The limitation of these studies is that their estimates cannot
Data and descriptive statistics
The data are from the second round of the National Family Health Survey of India (NFHS-II) which recorded complete fertility histories for ever-married women aged 15–49 in 1998–1999, including the time and incidence of child deaths.1 Mothers constitute the cross-sectional dimension of the data. As mothers are observed repeatedly, in relation to every birth, birth-order creates the time dimension of the (unbalanced)
The model
The model has a recursive dynamic structure: the risk of neonatal mortality depends upon mortality of the previous sibling and on the preceding birth interval, while the birth interval depends upon survival of the preceding sibling. Similarly, the probability of continuing fertility depends on previous mortality and parity. Identification of the main causal effects rests on exploiting the natural sequencing of the birth and mortality processes, avoiding the need for exclusion restrictions.
Results
This section presents the results of the complete “benchmark” model (Table 1).
Summary and conclusions
The main findings are as follows. The difference in neonatal mortality according to whether or not the preceding sibling died in the neonatal period is 14%-points. This is enormous, given that the average risk of neonatal death in our sample is 7.4%. We estimate that genuine state-dependence accounts for 37% of this, the remaining 63% being explained by inter-family heterogeneity. The analysis confirms that endogenously determined birth-spacing is a mechanism generating state-dependence in
Acknowledgments
Sonia Bhalotra acknowledges an ESRC research grant for work on sibling death clustering. Part of this work was carried out when Arthur van Soest was working at RAND and was funded through a RAND IR&B grant.
References (57)
Family planning, gender differences and infant mortality: evidence from Uttar Pradesh, India
Journal of Econometrics
(2003)- et al.
A quantitative analysis of Swedish fertility dynamics, 1751–1990
Review of Economic Dynamics
(1999) - et al.
Neonatal survival 1–4 million neonatal deaths: when? where? why?
Lancet
(2005) - et al.
Self-reported depression and negative pregnancy outcomes
Journal of Clinical Epidemiology
(1992) - et al.
Birth spacing, sibling rivalry and child mortality in India
Social Science & Medicine
(2002) A framework for estimating dynamic, unobserved effects panel data models with possible feedback to future explanatory variables
Economics Letters
(2000)- et al.
Ownership of stocks and mutual funds: a panel data analysis
Review of Economics and Statistics
(2004) - et al.
Children and their parents’ labor supply: evidence from exogenous variation in family size
American Economic Review
(1998) - et al.
Sibling death clustering in India: genuine scarring vs unobserved heterogeneity
Journal of the Royal Statistical Society, Series A
(2006) - Arulampalan, W., Bhalotra, S., 2006b. Persistence in infant mortality: evidence for the Indian states. Discussion Paper...
Fertility response to child mortality: micro data from Israel
Journal of Political Economy
A prospective study of birth interval dynamics in rural Bangladesh
Population Studies
Maternal depressive symptoms and infant health practices among low-income women
Pediatrics
Fertility decisions when infant survival is endogenous
Journal of Population Economics
The effect of birth spacing on childhood mortality in Pakistan
Population Studies
Birth interval and family effects on postneonatal mortality in Brazil
Demography
The determinants of mortality
Journal of Economic Perspectives
Death clustering, mothers education and the determinants of child-mortality in rural Punjab, India
Population Studies
Failure to thrive in the term and preterm infants of mothers depressed in the postnatal period: a population-based birth cohort study
Journal of Child Psychology and Psychiatry
Uttar Pradesh: the burden of inertia
Indian Indian Development: Selected Regional Perspectives
The relationship between infant and child mortality and subsequent fertility in Indonesia, 1971–1991
Population, technology, and growth: from Malthusian stagnation to the demographic transition and beyond
American Economic Review
Birth intervals, gestational age and low birth weight: are the relationships confounded?
Population Studies
Use of sibling data to estimate family mortality effects in Guatemala
Demography
The contraceptive role of breastfeeding
Population Studies
Classical estimation methods for LDV models using simulation
Cited by (107)
The impact of women's political representation on child health outcomes during 1990–2020: Evidence from a global dataset
2022, Social Science and MedicineCitation Excerpt :For example, female politicians are more likely than male politicians to support public health improvements directed toward female constituents as they disproportionately bear the burden of weak reproductive and child health services (Cutler et al., 2006). In addition, women primarily bear the burden of care for the sick, witness the death of their children, undergo childbirth, and face the risk of mortality during childbirth (Bhalotra and Van Soest, 2008). In India, female political leaders were more likely to support a higher provisioning of public goods (Beaman, 2007) and policies that reflect women's needs such as drinking water and maternity pension than their male counterparts (Chattopadhyay and Duflo, 2004).
From fees to free: User fee removal, maternal health care utilization and child health in Zambia
2022, World DevelopmentCitation Excerpt :Such fertility decisions are likely to be endogenous. In particular, it may be influenced by unobserved characteristics at the household level including parents’ preferences, and motivated by a replacement effect in case of child death (Bhalotra & van Soest, 2008; Hossain, Phillips, & Legrand, 2007). Contamination effects – One concern is that some individuals living in control areas might have benefited from the policy if they seek care in a treated one.
Girls unwanted – The role of parents’ child-specific sex preference for children's early mental development
2022, Journal of Health EconomicsSaving for dowry: Evidence from rural India
2022, Journal of Development EconomicsReligion and abortion: The role of politician identity
2021, Journal of Development EconomicsCitation Excerpt :We show that there is no evidence of this in stated preferences for sons. We investigate this formally, using the same identification strategy to estimate the impact of legislator religion on fertility since continuation of fertility has been shown to be a substitute for sex-selective abortion as a means of achieving the desired gender mix of children (Clark, 2000; Gangadharan and Maitra, 2003; Bhalotra and van Soest, 2008; Almond et al., 2013; Rosenblum, 2013; Anukriti et al., 2021). We find that, in districts won by Muslim legislators, there is an increase in the probability of a third or higher order birth.
How does a ban on sex-selective abortions affect child health?
2024, Health Economics (United Kingdom)
- ☆
We are grateful to Wiji Arulampalam for helpful discussions, and to John Geweke, an Associate Editor and an anonymous referee for their suggestions. We also received helpful comments from Julie DaVanzo, Helene Turon, and Mike Veall. The paper has benefited from presentations made at the World Congress of the Econometric Society, the ESRC Econometric Study Group, the PAA meetings, RAND, Bristol, Tilburg and the Cambridge Panel Data conference.