akgunduz

Appendix 6C shows some of these alternative specifications. First, we fitted an alternative specific conditional logit model (ASCL) without random coefficients. Quality turns significantly negative in this case, suggesting that the heterogeneity of tastes towards child care quality plays an important role. Second, we omitted siblings from the model both in prediction equations and in the main model. The results remain similar as can be seen in Appendix 6C. Third, we used an unbalanced, larger sample of 1400 observations. Once again, the results shown in Appendix 6C are similar apart from stronger effects from wages. We also made a few minor specification tests by first omitting the male income variables from the main model which were not included in the prediction equations and then adding dummies for the second wave. The former test leads to slightly stronger effects from wages, while the latter slightly weaker. Price effect remains stable. Finally, to test the consistency of the main model, we calculated heteroskedasticity robust standard errors. Significance levels changed little.

6.5 Conclusions

Our analysis of Dutch women’s labor market behavior leads to the conclusion that quality as defined by developmental psychologists does not positively affect women’s labor supply behavior. The result is not in line with the standard economic theory that predicts a positive link between centers’ quality and center based child care use and labor supply. The lack of effects may be due to the difficulty parents have in observing centers’ quality. Alternatively, quality may be negatively correlated with other center characteristics such as flexibility that parents value. For the remaining variables, the effects are in the expected directions. Prices has the expected negative effect on labor supply while the wage elasticity of Dutch women with partners and young children is positive.

It is likely that research on ECEC quality and its determinants will continue regardless of its implications for labor supply. There is a growing policy and academic interest in ECEC quality and the benefits of high quality child care on future life outcomes. The impact of ECEC quality on development has already led to more investments in early childhood in both Europe and North America. An alternative reason to invest in child care quality could be effects on female labor supply, but we find no evidence of such effects. If we assume that the effects are lacking due to information asymmetry problems, more publicly information about the quality of

child care centers may increase the returns from policies designed to increase ECEC quality by influencing labor supply.

6.7 Appendix 6B

Table 6.9: OLS estimates for quality and price

*** p<0.01, ** p<0.05, * p<0.1. All observations indicates the number of observations in the selection equation (812) that also includes observations where price and quality are not observed.

decrease the opportunity cost of working for women. Despite generally relying on an unchanged static labor supply framework, the literature estimates for labor supply elasticity with regards to child care prices are varying. While some estimates imply substantial gains from child care subsidies, others find insignificant effects. Determining the reasons for the variance in the results and the settings in which elasticity estimates are smaller or larger is of substantial policy interest. If the elasticity estimates are large, child care subsidies may be a free lunch for governments that can recoup their expenditures through larger tax bases.

To understand the cause for variation among the elasticity estimates, this chapter reviews and analyzes the elasticity sizes using estimates from 39 studies. The chapter begins by reviewing the theoretical and empirical aspects related to participation elasticity with regards to the child care costs, paying special attention to sample characteristics, methodological aspects and institutional factors. We conclude by providing a meta-regression using control variables based on our review of the literature to explain some of the differences between the estimates.

Large differences in the target group used in the analysis or estimation methodologies seem to matter for elasticity sizes. The elasticity estimates tend to be larger for studies using subsamples of married women. More recent papers that use multinomial logit models tend to have smaller elasticity estimates than traditional probit or logit estimations. Most such models allow for substitution between the informal and formal child care which generally produces smaller elasticity estimates (Blau and Currie, 2006). Other improvements in methodology that are not explicitly controlled for may also account for the smaller elasticity estimates from studies published after 2000. There are many minor choices that differ across studies which we cannot control for such as how unobserved heterogeneity is modeled or whether choices regarding child care type and employment are simultaneously modeled.

The more striking finding is that the elasticity estimates are significantly correlated with labor market characteristics of the country that the estimation is from. Countries with very low participation rates tend to have smaller elasticity estimates. The effects may also diminish at very high participation rates. Countries with high working time flexibility through part-time employment also show lower elasticity estimates. Child care services do not seem to matter as much when part-time, informal substitutes are easily available. Although these findings are correlations rather than causal effects, they suggest that increasing child care subsidies in countries with very high female participation rates and opportunities for part-time employment do not