Socioeconomic Differences and Family Clustering of Child Mortality: A Multilevel Analysis of Rural Southern Sweden, 1766-1895

Tommy Bengtsson, Lund University

This paper studies the determinants of child mortality during the first stages of the demographic transition. Previous analyses have shown that while socioeconomic differences in child mortality during the nineteenth century were modest but growing, there was a large amount of clustering in certain families. This paper examines the issue using multilevel Cox regression with shared unobserved factors at the family level included. The problem of comparing the impact of unobserved factors at family level with the effects of observed factors is solved by employing a new method for converting the variation stemming from unobserved factors at family level into Median Hazard Ratios (MHR), directly comparable with relative risks. The data come from the Scanian Demographic Database. The principal findings are that while socioeconomic differences in child mortality were greater than sex differences, differences between families, whether due to observed or unobserved factors, were even larger.