Trends in Women's Labour Force Participation in Australia: 1984 - 2002
Melbourne Institute Working Paper No. 23/04
Date: September 2004
Author(s):
Abstract
Women's workforce participation increased strongly over the 1980s and 1990s, with the increases being generally larger for married than non-married women, and with the increases being especially large in middle age, as shown by ABS data. Multivariate analysis of IsssA data covering this period shows that there is actually rather little time trend per se. Instead, underlying the apparent shift over time, there are large compositional changes in the female population and there is a strong "birth cohort or "vintage" effect such that succeeding cohorts of women have higher propensities to work throughout their lives than did their predecessors. Among the compositional changes, the strong rise in women's educational attainments and the large decline in fertility both exert substantial influences elevating women's workforce participation and hours worked. There were no evident time effects associated with particular policy initiatives, but some of these are too colinear with time to analyse separately. We tested many interactions with time to assess in particular whether the effects of education and of family situation are declining over time, but no significant interactions with time were found. Thus, for example, there are now many more highly educated women, leading to higher rates of women's employment overall, but the relative importance of education has not changed significantly. Similarly, there are now more childless women, and women with children have fewer of them, so declining fertility has elevated employment, but the impacts of childlessness and of diverse family sizes have not changed, according to these models. Finally, note that we tested a number of potential effects of the family of origin, but none was significant, suggesting that analyses of women's labour force participation and hours worked using datasets that lack these variables probably do not suffer from omitted variables bias.