Causal analysis of education and birth inequalities through a latent class structural equation model

We investigate in a causal perspective whether the maternal education and other social characteristics, such as marital status, can influence birth inequalities, mainly birthweight and gestational age. We base our causal analysis on a structural equation model that takes explicitly into account the possible existence of unex- plained heterogeneity through a latent background variable, which plays the role of confounder. The proposed model is a special case of mixture or latent class structural equation model, based on a suitable number of recursive equations in which (i) unobserved heterogeneity is represented by a discrete latent variable, (ii) maternal education and marital status may depend on the discrete latent variable and on other covariates, (iii) birthweight and gestational age depend on maternal education and marital status, on the discrete latent variable and on other covariates. The study is based on the population of singleton born in Umbria (Italy) from 2007 to 2009.