Identifying motherhood and its effect on female labour force participation in South Africa.

Abstract

The objective of this thesis is to investigate the relationship between motherhood and women's labour force participation in South Africa. The key problem in estimating this relationship is the endogeneity of motherhood/childbearing with respect to women's labour force participation. Childbearing behaviour and decisions to participate in the labour force are jointly determined; and unobservable characteristics which influence childbearing behaviour are also correlated with women's labour force participation. This thesis shows that the definition of motherhood can exacerbate these sources of endogeneity bias. International studies typically identify mothers as women with biological children aged 18 years or younger who are co-resident with at least one of their children. In South Africa, however, a sizeable sample of women is not co-resident with their children. The remaining sample of co-resident mothers are a non-random sample of all mothers who are less likely to participate in the labour force than all mothers. Placing a co-residency restriction on motherhood therefore biases the relationship between motherhood/childbearing and labour force participation. In particular, it overestimates the negative relationship. In the international literature instrumental variable (IV) estimation has been used to disentangle these causal mechanisms. This thesis also considers an application of same sex sibling composition, first introduced by Angrist and Evans (1998), as a strategy to identify the exogenous effects of childbearing on women's labour force participation in South Africa. Little or no research has investigated this relationship in South Africa. One possible explanation for this is that studies on female labour force participation in South Africa have not been able to match women to their children with the datasets that have been analysed: most nationally representative household surveys in South Africa do not contain detailed birth history information. The first part of this thesis analyses what data are available to identify women with children and the quality of these data; it also outlines four different methods to match women to their children using these data. The second part of this thesis investigates the relationship between motherhood/childbearing and women's labour force participation in South Africa.