Seminar
When do mothers have the first child? Selection into (timing of) motherhood, earnings trajectories and earnings accumulation in Sweden

Stefano Cantalini (University of Milan)

16 October 2017, h. 11.00 - 13.00
Seminar Room (via Conservatorio side)
Department of Social and Political Sciences
via Conservatorio 7 - Milan

Empirical research has shown that mothers earn less than non-mothers. Some of these studies claimed that part of this 'motherhood penalty' is due to self-selection into parenthood, i.e. mothers and non-mothers differ according to unobserved characteristics, such as ambition and career orientation. However, works using (normal) fixed-effects methodology can only control those biases from time-constant unobserved heterogeneity, assuming that unobserved characteristics do not change over the career.
In this paper, co-authored with Juho Härkönen and Johan Dahlberg (Stockholm University), we study the consequences of motherhood on earnings trajectories over the life-course and analyse the impact of (timing of) parenthood on cumulative earnings in Sweden. By using a novel methodological approach, based on fixed-effects growth curve models with individual slopes, we are able to control the unobserved heterogeneity bias derived from time-varying characteristics and the self-selection bias operating not only on earnings level but also on earnings growth. In other words, we can investigate if women decide to become pregnant when their earnings are upwardly or downwardly growing and if motherhood has long-term impacts on earnings trajectories and accumulation.
We analyse data from Swedish population registers, using the cohorts of women born 1967-1981. Preliminary results show that (normal) fixed-effects models tend to underestimate the motherhood penalty and that Swedish women decide to have the first child when their earnings are increasing. From a life-course perspective, mothers' earnings dramatically decrease after the first birth and then increase after few years, making them to catch up with non-mothers in the long-run. However, this catching-up effect is primarily given by bias from time-varying unobserved heterogeneity.

This seminar is part of the ResFron ESLS Cycle of seminars - 2018 Edition