IS CHILDCARE LEAVE EFFECTIVE IN RAISING FERTILITY IN JAPAN?

Abstract

This paper estimates the effect of childcare leave on married women's fertility in Japan, based on data from the 2007 National Survey on Work and Family. The analysis takes into account how childcare leave influences fertility through its intermediate effects on women's selection into the labor market, job tenure, wages and the opportunity cost of children. Results indicate a strong effect of childcare leave on years of continuous job tenure with the same employer, and on predicted wages for full-time working women. Taking childcare leave for the first child increases the percentage progressing from first to second birth by six percentage points. There is also clear evidence that lowering the opportunity cost of children increases fertility, net of the effect of childcare leave, which affects fertility via the opportunity cost of children.

Keywords:

• fertility
• childcare leave
• job tenure
• opportunity cost of children

Acknowledgements

Research for this paper was funded by a grant from Nihon University Population Research Institute, the ‘Academic Frontier’ Project for Private Universities, a matching fund subsidy from the Ministry of Education, Culture, Sports, Science and Technology in Japan. The comments by Pau Baizan and a referee are greatly appreciated.

Notes

1. The use of somewhat arbitrary values of tenure and firm size for non-working women is a fairly minor problem, compared with the time inconsistency problem, namely, that predicted wage is estimated using current wage of full-time working women rather than their historical wages, for which information is lacking in our data set.

2. If childcare leave and wages are negatively related, then childcare leaves does not reduce the opportunity cost of children. It increases it.

3. This is calculated as follows. Take the derivative with respect to T (tenure) of both sides of the wage equation and multiply through by the differential dT. The result is dW/W=(b +2cT)dT, where W denotes wage, b is the coefficient of T, c is the coefficient of T-squared, and dT is a small increment in T. Set T equal to its mean value in the sample on which the probit regression is run, and set dT=5. Substituting in these numbers and the numbers for the coefficients b and c for full-time workers in column two of Table 2 this becomes dW/W=[(0.085)+(2)(−0.002)(12.3)](5) = 0.18 or 18 percent.