Abstract:
The last decade, more industrialized countries and international organizations have shown interest in different types of government intervention in the market for child care. Among these is the OECD, and in his most recent State of the Union address, President Obama proposed to make “high-quality preschool available to every single child in America” (Obama, 2013). Similar moves towards child care reform have been made in Germany and other European countries.
Such interventions are usually claimed to have at least one of three effects. First, they are said to affect fertility and combat low fertility. Economists have documented a negative relationship between income and fertility (Jones et al., 2010; Jones and Tertilt, 2006), and high-income countries like Germany have seen universal subsidized child care as a tool for increasing below-reproduction fertility. Second, universal child care is said to benefit child development and equalize differences in initial endowments. Although the evidence on the effects of child care on development is inconclusive, some economists have suggested to use early child care intervention as a cost-efficient way to combat social reproduction (Currie, 2001). Last, and most importantly for this thesis, universal child care is claimed to be an efficient tool for increasing female labor force participation by reconciling work and family responsibilities (OECD, 2006).
In Norway, female labor force participation and gender equality in the labor market is an important goal for the universal child care system and welfare state in general. The government has a clear goal of full child care coverage, claiming among other things that universal child care is “central to parents’ labor market inclusion” (Kunnskapsdepartementet, 2007-2008). Subsidized, universal child care has been an integral part of the Norwegian and Nordic welfare states since the second World War, and so these countries provide natural cases for evaluating the efficiency of such measures. Until recently, however, the availability of such child care has been more or less restricted to preschoolers, children of 3 to 6 years. Several scholars have evaluated the effect on labor supply of child care for these children, among them Havnes and Mogstad (2011) using a 1975 reform.
For younger children such research is scarce. There are several reasons to believe that the labor market response of mothers of toddlers (1- and 2-year olds) differ from mothers of older children. For one, the alternative mode of care may differ: If formal child care is not available, mothers of toddlers might take care of the children themselves while mothers of preschoolers find informal solutions enabling them to work. Second, the preferences of the mothers might differ so that they are more or less responsive to the availability of care.
The purpose of this thesis is to evaluate the third claimed effect of child care: Is universal child care an efficient tool for increasing the labor force participation of mothers of toddlers?
To investigate causal effects, we need a valid estimation strategy that can disentangle causality from correlation. Simply comparing the labor supply of mothers in municipalities with high child care coverage to that of mothers in low coverage municipalities is not likely to yield unbiased estimates of the true effects.
A reform from 2003 leading to large increases in child care for toddlers in Norway provides a natural way to evaluate the impact of toddler care. The reform increased government subsidies to investment in and running of child care institutions, and generated large variation in expansion rates between municipalities and over time. This variation provide a unique possibility to evaluate the effects of toddler care, and as will be shown, there are reason to believe that the changes in child care coverage following the reform can be regarded as exogenous.
To solve the unobserved heterogeneity problem, we make use of high quality Norwegian registry data in a fixed effects method, explaining changes in labor supply by changes in child care coverage. As long as the composition of unobserved determinants of labor supply is constant, this approach will control for them. Since the main regression uses fixed effects at the municipality level, we also control for a range of individual characteristics that might affect labor supply decisions. These include education of the parents, immigrant status, household characteristics, age and family structure.
In a non-rationed market, changes in both demand and supply will affect the observed coverage rates and we might worry for reverse causality: That increases labor supply lead to increases in coverage rates, rather than the opposite. As will be argued, the child care market following the reform is severely rationed. In a rationed market, the observed child care coverage rates will be driven by changes in supply only until the rationing is lifted. The rationing in the child care market therefore strengthens our empirical approach and reduces problems of reverse causality.
Using this method, we find that a full scale expansion of child care for toddlers will lead to an 11 percentage point increase in the share of working toddler mothers. This constitutes around 16% increased participation compared to the mean over the period. For full-time employment, the effects are smaller at around 5 percentage points or 13% increase in mothers working full time. These effects are significantly different from zero at any conventional significance level. We also find some evidence of persistency in the labor supply response, indicating that child care for toddlers have effects on labor supply that last several years after the toddler period. This improves the cost efficiency of the reform as it increases the tax base over several years.
We also find slight evidence of a reverse response to child care among fathers: Fathers seem to reduce their labor supply as a response to increased child care coverage, or possibly increasing labor supply when child care is not available to support a mother not working. Although these estimates are imprecise, the effect of this is to lower the cost efficiency of the reform.
These estimates are larger than other estimates in the literature, particularly Havnes and Mogstad (2011), who find effects around half the size for mothers of 3- to 6-year olds following a similar reform in 1975. Although preferences of mothers of older children might be different and preferences may also change over the three decades between the two reforms, our main explanation of this difference is that the alternative mode of care for toddlers and preschoolers differ. The results from this thesis indicate that the alternative mode of care for toddlers when formal care is not available to a larger extent is parental care than informal care such as relatives or unregistered child minders. This makes a child care reform for toddlers more efficient than for other children when considering the effect on female labor supply only.
Although a cost-benefit analysis of the child care reform is not the purpose of this thesis, the economic implication of these results is that the government must construct around 9 slots in child care to induce one more mother to enter the labor market. These slots are costly, considering the large government subsidies involved, and these are not covered by the increased tax income from one more working mother.
We might worry that our results are driven by self-selection into or out of treatment. This can happen if work-prone mothers migrate into or out of municipalities with large expansion of child care. If there is selective migration, the effects we have found might simply be the result of mothers interesting in working more sorting into municipalities with high expansion. To investigate this, we perform several robustness tests where this sort of selective migration is ruled out, and find that the results are stable to these tests.
Our approach is based on an assumption that is very similar to the common trend assumption of Difference in Differences setups: Conditional on the control variables and with no changes in child care coverage, the municipalities would follow a common trend in labor supply. To test this assumption, we perform two tests where we relax this assumption using separate time trends for each municipality and time shocks that vary by municipality characteristics. These tests also support our approach.
We last perform an overall placebo test explaining the labor supply of mothers of older children by the child care coverage rate for toddlers. This should not have any explanatory power, as these mothers no longer have toddlers. If they do, we would worry for misspecification of our model. This test shows little evidence of misspecification. These and more robustness tests lend support to our empirical approach, indicating that we have identified a causal effect of child care on female labor supply.
The results from this analysis are highly relevant to current political debates in Norway, considering that the demand for child care for older children is more or less covered. The findings in this thesis therefore speak to the efficiency of further expansion. They will also provide evidence for other countries and governments considering a move towards universally accessible, subsidized child care for young children. Universal child care expansion for toddlers are not necessarily money for nothing as Havnes and Mogstad (2011) find, but still a rather cost-inefficient way to increase female labor force participation.