Socio-economic differentials in labour market attachment and its impact on childbearing behaviour: analyzing the transition to first, second and third births among Belgian women and their partners

David De Wachter, Vrije Universiteit Brussel
Karel Neels, University of Antwerp

Retrospective research for Belgium, based on the 1991 and 2001 census, has focused extensively on socio-economic differentials in cohort profiles of order-specific fertility. It appeared that higher educated women combine later ages at childbearing with more frequent transition to motherhood, whereas lower educated women combine a relatively young fertility schedule with less frequent transition to motherhood. Furthermore, higher educated women record the highest probability of having a second and even third birth, despite the fact they postponed fertility to ever later ages. In this paper we relate these socio-economic differentials in order-specific fertility to socio-economic differentials in female and male labour market attachment. Not only are there clear differences in occupational status according to educational level, with higher educated women typically moving into more stable labour market positions, also the effect of labour market attachment on childbearing behaviour differs between educational groups. For instance, unemployment results in lower first birth hazards, regardless of the educational level achieved. At the same time, the impact of being unemployed seems to affect higher educated women more severely. In this paper we intend to give a more comprehensive picture of the relationship between labour market attachment and childbearing behaviour in Belgium. First, we explore the correlation between educational attainment and labour market attachment and estimate the impact of occupational status on first, second and third birth hazards. Second, we expand the scope of the analysis by also controlling for the labour market attachment of the male partner. Using a prospective research design, and data from the 1991 and 2001 Belgian census, discrete-time event history models are estimated, where we stratify according to age-group and educational level.