Gross effects Net effects
Odds Reduction Odds Reduction
Variables ratio in LRX2 P ratio in LRX2 P
Length of preceding birth interval < 19 (months) and maternal age at
childbirth (years) 44.2 0.001 56.8 0.001
<24 1.966 2.030
24-33 1.231 1.379
33 + 1.486 1.203
Length of preceding birth interval 19-36 (months) and maternal age at childbirth (years)
<24 1.224 1.306
24-33 0.923 0.950
33 + 0.918 0.857
Length of preceding birth interval 36 + (months) and maternal age at childbirth (years)
< 24 0.529 0.596
24-33 0.729 0.739
33 + 0.695 0.634
n=7076 reduction in model LRX^ and (d.f) =108.99 (92) P=0.109 overall odds =0.100
Notes: Estimates are net of birth cohort, sex of child and birth order of child. The joint variable has 8 degrees of freedom.
Source: 1986 NFFS data tape.
The odds ratios in the net effects columns of Tables 3.19 and 3.20 reveal an inverse association between mother's age at childbirth and infant's survival after controlling for the preceding birth interval of under 19 months. The same is true for the length of preceding birth interval of 19-36 months except for infants bom to mothers in the 24-33 year age groups as compared with those bom to mothers over 33 years of age (Table 3.19). However, the difference is very small. The results thus suggest that infants bom less than 19 months after the birth of the preceding sibling and 19-36 months after the birth of the preceding sibling have a lower risk of death with the increase in mother's age at childbirth. However, the results for births more than 36 months after the preceding birth seem to show the risk of dying rising with advancing maternal age at delivery, perhaps reflecting delivery
difficulties associated with aging. In the case of younger mothers spacing seem to be associated with lower risk of dying for their children.
3.10
Further discussion of the results
During their reproductive life, women in Nepal on average tend to bear six children of which 12 per cent die before one year of age; 9 per cent of those who survive their infancy die before they turn five years of age (Tables 3.1 and 3.2). Thus only 79 out of 100 children born in Nepal survive their first five years of life, confirming the persistence of poor child survival prospects. However, the results based on birth cohort analysis suggest an optimistic scenario for systematic decline in infant and child mortality.
A higher risk of death to male infants due to the biological advantage of female infants, and the higher risk of death to female children was attributed to the parent's favouring of male children in India (Dasgupta, 1987: 88-96; Bourne and Walker, 1991: 207-218) and in Bangladesh (D'Souza and Chen, 1980: 259-268). Unlike in India and Pakistan, in this study the effect of sex of the child shows no statistically significant influence either on infant or child mortality. A similar result between sex of child and infant and child mortality was also found in an earlier study in Nepal based on the 1976 NFS data (Gubhaju, 1991). At first sight, this phenomenon is very surprising in a country where sons are preferred over daughters for several reasons: the high economic value of sons, old age security, religious significance and labour-intensive agriculture.
However, there are some plausible explanations for the evident insignificant influence of sex of child on infant and child mortality in Nepal. First, as forced dowry or groom price5/ in most areas of Nepal is not a cultural practice, parents may not consider female children as an economic burden. Second, a son occupies a prominent place in Hindu religion because a son is believed to be the key to heaven and a grandson is believed to be the permit to stay in heaven for an unlimited period. However, a grandson bom to a
5/ Groom price (Tilak) is an amount paid to the groom by the parents of the bride for accepting their daughter. This is a cultural practice only among some castes or communities. Dowry is the gifts given to the bride both in cash and kinds by the bride's parents.
son and a grandson bom to a daughter possess equal status (Shrestha, 1986: 15). Third, marrying off a daughter is considered to be the gift of a virgin (Kannya-dan) which is perceived to bring religious merit (Punnya) to parents. All these factors could tend to equalize the importance of sons and daughters in Nepalese societies explaining the absence of difference in influence of the sex of child on infant and child mortality. On the other hand, there can be persistent son preference in the society when birth intervals are discussed.
Tables 3.3 and 3.4 suggest that typical childbearing in Nepal starts before 20 years and is completed by about 34 years. During the 15-year span women end up with about six children suggesting an average birth interval length of about two years. As a result of rapid births, mothers are continuously disadvantaged in terms of their depletion in health, nutritional and reproductive efficiency which ultimately are likely to drive these mothers towards a higher rate of child loss. Thus as suggested by Hobcraft et al. (1985), the influence of mother's age at childbirth and birth order on infant mortality in this study is largely attributed to birth interval. Very large attenuation in the LRX2 value corresponding to maternal age at childbirth and birth order after including the interval variable in the model further supports this argument. At the same time, mother's age at childbirth might be partly influencing child survival through birth weight, which is not available in the data.
However, for child mortality the influence of mother's age at childbirth from both the 1976 NFS and 1986 NFFS becomes statistically insignificant after taking into account the preceding birth interval variable, so any age effect is entirely attributable to length of birth interval. This result supports the argument that sibling competition for mother's care, tangible resources and health care is the major cause of child mortality.
Length of preceding interval in this study emerges as the most important variable in influencing survival status of both infant and child. It is clearly evident in this study that short birth intervals lead to poor child survival prospects and increased birth interval between siblings provides better child survival prospects. Palloni and Millman (1986), in a study of 12 Latin American countries, noted this phenomenon in seven countries for child
mortality (one to five years of life), in nine countries for 6-11 months of life and in six countries for 1-2 and 3-5 months of life. Pebley and Stupp (1986) and Hull and Gubhaju (1986) also noted this phenomenon for Guatemala and Indonesia respectively.
In the literature the adverse effects of closely spaced births on infant survival are widely attributed to maternal depletion syndrome and gestational immaturity leading to low birth weight (Fedrick and Adelstein, 1973: 754-755; DaVanzo et al., 1983: 392; Cleland and Sathar, 1984: 402). In contrast, the adverse influence of closely spaced births on child mortality is attributed to sibling competition for their mother's attention and tangible resources (Palloni and Millman, 1986: 216). In this study the observed favourable child survival associated with long birth intervals could be due to mothers having adequate time to recover their strength after the preceding pregnancy and lactation before the conception of the following child. Moreover, the possibility of disease communication for children close in age is minimised for those who are bom after a longer interval.
This study has examined the influence of demographic variables of interest on infant and child mortality. Although the length of the preceding birth interval emerged as the major influential factor in determining child survival prospects, it was not possible to demonstrate its mechanism empirically because information on child care behaviour of mothers, mother's nutritional and health status was not collected in the survey. Moreover, length of birth interval is also likely to be correlated with other factors such as breastfeeding, gestational age of child and birth weight. Thus the association between length of interval and survival status of child is not entirely conclusive. As gestational age of a child was not collected in the survey and as breastfeeding information was available only for selected children, these variables are not explored in relation to child survival in this study. However, closely spaced births in Bangladesh and the Philippines (Miller et al., 1992: 315) and in Guatemala (Pebley and Stupp, 1987: 58) were noted to be associated with an elevated risk of deaths to infants even when potentially confounding effects of gestation were taken into account. Similarly, Cleland and Sathar (1984: 417) noted that the influence of the interval on child survival status in Pakistan was not the spurious consequence of early weaning. The same could be true for Nepal.
The association between survival status of children and length of interval is also argued to be the consequence of spurious association in those studies where the survival status of the preceding child is not taken into account (Winikoff, 1983: 232; DeSweemer, 1984: 50 and Cleland and Sathar, 1984: 401). This is because mothers who lose their children tend to conceive as soon as possible in order to recover the child loss. Households which lose an earlier child are also likely to lose subsequent children because the environment, resources and the socio-economic factors for the next child are also likely to be the same. Most of these hypotheses cannot be tested in this study because of the unavailability of the data. However, the effect of survival status of the preceding child along with other demographic, socio-economic and health-related variables on infant and child mortality is examined in Chapter Five.