ISER Working Paper Series 2006-46
Socio-economic differences in postponement and recuperation of fertility in Italy: results from a multi-spell random effect model
21 Sep 2006
The positive association between postponement in fertility and a general decline in total fertility rate has been accepted as an empirical regularity across cohorts and countries. One of the major findings from this literature is that women with higher education or higher earnings potential tend to postpone first birth. But the postponement of the first birth does not necessarily imply a lower completed fertility because it could be associated with recuperation in terms of the second and third birth. Unfortunately it is not straightforward to separately identify the postponement and the recuperation effect: they depend both on observed and unobserved characteristics.
Combining two different data sets (from the ISTAT Labor Force Survey and the Survey of Households' Wage and Wealth led by the Bank of Italy in 2002), we investigate socio-economic differences in delay and recuperation of childbearing in Italy, a country that is suffering from what is termed the lowest low-fertility. The approach is based on a multi-spell random effect hazard model, whereby estimation of first, second and third birth is done jointly. The model allows for unobserved heterogeneity and we test to what extent the assumption about its functional form has any impact on the overall parameter estimates.
An expected result is that women with high wages tend to delay the first birth: we expected this result because education is the best predictor for potential earnings. Thus for a woman with high wage it could be optimal to complete her education, establish herself in the labour market and then start their childbearing career. We also find a strong recuperation effect of a high earning woman in terms of the progression to the second birth. As a consequence, by the age of 40 high earning women have caught up with low earning women almost completely. However, socio-economic status has little effect on third birth. To overcome the estimation problems caused by women specific unobserved characteristics we control for sample heterogeneity. In particular we show that the introduction of the unobserved heterogeneity is important, but the assumption of its functional form is not. The case of a non parametric specification of the random effect identifies two different groups: the 'movers', i.e. women who make the transition faster than the 'stayers' and that are more family oriented.
In order to asses which is the wage level that captures the largest part of the variance of the unobserved heterogeneity, we interact the random effect with different socio-economic groups. Women with relatively low and high wages are the typology of women for which it is more difficult to explain their attitude towards fertility decisions without considering other unobserved characteristics.