Inherited dimensions of infant mortality. Detecting signs of disproportionate mortality risks in successive generations

Abstract

In an effort to unravel the diffusion and mechanisms of long-term fertility change, there is a growing body of literature on the intergenerational transmission of reproductive behavior. Several reproductive traits have already been studied in an intergenerational dimension, but the impact of infant deaths and their clustering within families on these intergenerational mechanisms of reproduction has been overlooked. This paper explores the continuity of infant deaths between successive generations. Data are from the COR*-database, a longitudinal multigenerational life course dataset covering the larger Antwerp area in the second half of the nineteenth century. We use the family as main unit of analysis, a perspective that proves a useful alternative to that of the individual child.

Keywords:

• infant mortality
• death clustering
• intergenerational transmission
• family effects

Notes

1. The Antwerp COR*-database is hosted and maintained by the research group Family and Population Studies from the Centre of Sociological Research at the Catholic University of Leuven (Parkstraat 45, post box 3601, 3000 Leuven, Belgium). For further information, please contact [email protected].

2. The ratio which is calculated in this paper is not exactly the same as the more frequently used infant mortality rate (IMR). The IMR leaves out stillbirths in both nominator (deaths) and denominator (births), where we included them. According to our sample, 128 children ( = 12 per cent of all infant deaths) are recorded as stillbirths, which is almost certainly an over-registration due to national regulations in vital registration. The Belgian law allows for a delay of maximum three days between the birth and the declaration of the birth in the city hall. Some of the children who died before their birth could be declared, for example during weekends, were declared as stillbirths, and recorded as such in the source material (for a discussion: see Poulain & Tabutin, 1977). Because it is impossible to differentiate real stillbirths from early perinatal deaths in the consulted Belgian sources, we avoided this registration issue by widening the definition of the IMR, including all births (instead of only live births) and all deaths (instead of only non-stillbirths).

3. Under the assumption of no clustering (no dependence between infants), the number of infant deaths K in a family can be seen as a series of independent Bernoulli trials, where every trial stands for one infant with a certain risk of dying, and is repeated for as many times as there are children in the family. K therefore follows a binomial distribution, defined by parameters n (number of trials) and k (number of successes). The proportion of families with exactly k infant deaths out of n births, in a population with an overall infant mortality rate of p (which equals 0.155 in the Antwerp case) is given by the probability mass function f:

This function, in combination with the minimum required number of infant deaths for each level of mortality by family size (Table 3) allows to calculate the proportion of families we expect to lose a high, a medium, or a low proportion of their infants.