Scand. J. of Economics 112(3), 570–594, 2010 DOI: 10.1111/j.1467-9442.2010.01617.x

Price of High-quality Daycare and Female Employment Marianne Simonsen∗ Aarhus University, DK-8000 Aarhus C, Denmark [email protected]

Abstract Using local variation between municipalities, I analyze the degree to which the price of high-quality publicly subsidized childcare affects female employment following maternity leave. Importantly, prices are income dependent and thus likely endogenous, yet by exploiting information on minimum income compensation during non-employment, I bypass this problem. The results show that the price negatively affects employment. A price increase of €1 per month decreases employment by 0.08%, which corresponds to a price elasticity of −0.17. These effects prevail during the first 12 months after childbirth. I also find that availability of childcare increases employment. Keywords: Quasi-experiment; municipality-level variation; maternity leave JEL classification: J 13; J 22; J 38

I. Introduction A well-established result in the literature on the female labor supply is that costly childcare works as a barrier to employment for women in the US. Examples are Heckman (1974), Connelly (1992), Ribar (1992), Kimmel (1998), and Powell (2002), who all find negative childcare price elasticities of employment. In addition, using an IV approach, Gelbach (2002) finds that providing free kindergarten increases the employment of mothers of five-year-old children. Using waiting lists for childcare subsidies, Berger and Black (1992) find that a reduced price increases female employment. Similarly, Averett et al. (1997) find that government subsidies for childcare increase female employment. Finally, a recent study by Baker et al. (2008) considers a large-scale change in the childcare system in Quebec, Canada. This policy change implies that all five-yearolds have access to full-time kindergarten and that the out-of-pocket price for childcare cannot exceed C$5 per day. Exploiting the before–after, ∗

I thank the Danish Social Science Research Council (FREJA grant) for financial support, and appreciate helpful input from Mette Ejrnæs, Nabanita Datta Gupta, Siv Gustafsson, Helena Skyt Nielsen, Inga Persson, Lars Skipper, Jeff Smith, Michael Svarer, two anonymous referees, and numerous seminar and conference participants. The usual disclaimer applies.  C The editors of the Scandinavian Journal of Economics 2010. Published by Blackwell Publishing, 9600 Garsington Road,

Oxford, OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

Price of high-quality daycare and female employment 571

Quebec-versus-other-regions variation, the authors find that the effects on the female labor supply of the transition to a regime with universal, highly subsidized childcare are clearly positive. US politicians have recently paid much attention to the possible role of expanded childcare subsidies and direct provision of public preschool; see Gelbach (2002). In this light, the Nordic countries make for an interesting case, because high-quality publicly funded daycare and parental cost subsidies are already in place.1 As such, the countries provide a laboratory in which to investigate the effects of such childcare policies.2 A similar argument for Europe in general is made in Ruhm (1998) regarding parental leave schemes. Additionally, Gustafsson and Stafford (1992) argue that because quality of childcare in the Nordic countries is more homogeneous compared to, for example, the US,3 estimates of price responsiveness are more easily uncovered and not as biased by product heterogeneity. Despite these observations, results on effects of childcare on the female labor supply in general in the Nordic countries are sparse. Using a smaller survey, Gustafsson and Stafford (1992) find that high-quality public childcare in Sweden encourages labor market activity of women with preschool children. Furthermore, by constructing a measure of rationing, they find that when spaces are not limited, a lower price encourages use. In this paper, I analyze the Danish set-up using data from 2001.4 As opposed to Gustafsson and Stafford (1992), I bypass the problem of endogenous childcare prices. Estimations are performed on a rich, Danish, register-based dataset consisting of 10% of the population that includes high-quality information on demographics and income for both spouses on a yearly basis. In addition, individual event history in terms of employment, unemployment, retirement, maternal leave, publicly subsidized child-rearing leave, education, and the residual category of non-participation is known on a monthly basis. A common problem in the literature is identification of prices of childcare, since these are often not directly observed. In the Danish case, the daycare prices faced by the parents vary deterministically with income, place of living (municipality), and number of siblings in daycare. Yet the problem of identifying prices remains; I do not observe labor income and 1

See OECD (2001) for an assessment of the quality of Danish childcare. A factor that might affect the results for the US and Nordic countries is the time at which the parents return to the labor market after giving birth. My calculations show that 96% of all mothers take all the leave possible (28 weeks in 2001), whereas working mothers in the US return much earlier after giving birth. 3 Also, Ruhm (2004) points out that the average quality of daycare in the US is extremely low. 4 According to Jaumotte (2003), Denmark has by far the highest level of spending on childcare within the OECD countries, amounting to 2.7% of GDP in 1999. 2

 C The editors of the Scandinavian Journal of Economics 2010.

572

M. Simonsen

therefore prices for a non-employed woman in the counterfactual case of employment. To overcome this problem, I exploit the fact that families with incomes above a certain threshold face prices that are independent of income. Using information on minimum compensation levels in the nonemployment state, I restrict the sample to consist of women for whom prices are independent of income regardless of employment status. This group includes 59% of all families, representing a wealthier part of the population. Thus, importantly, the estimated parameters are local in the sense that they only hold true for this smaller part of the population. I find that the price effect is significantly negative. An increase in the price of childcare of €1 per month will decrease the female employment rate by around 0.08%. This effect prevails during the first 12 months after childbirth. Furthermore, availability of childcare also seems to affect female employment. The paper is organized as follows. Section II presents the institutional settings, Section III discusses the identification and estimation strategy, Section IV presents the data, and Section V presents the results. Finally, Section VI concludes.

II. Institutional Settings Daycare in Denmark Daycare in Denmark (along with other Nordic countries) is characterized by high expenditure levels per capita compared to other countries within the European Union, and usage is high. According to the Society of Daycare Teachers (2004), about 75% of all children aged 6–9, 96% of children aged 3–5, and 61% of children aged 0–2 attended public daycare in 2002. Here, I consider care for the group of 0–2-year-olds. For the most part, daycare is publicly provided and organized within Denmark’s 271 municipalities.5 On average only 3.3% of registered care slots are privately provided6 and thus do not generally receive direct public funding. However, municipalities may decide to subsidize part of the price should the parents choose a private alternative. Unfortunately, information about this is not available for the period under consideration. The regulations of childcare institutions are described in the Law of Service (Serviceloven). All children are eligible for municipality childcare, including children born to unemployed parents. The only exception occurs if one 5 The Danish population is approximately 5.3 million in 2001. Hence, on average, each municipality has 20,000 inhabitants. The largest municipality is Copenhagen, with about 500,000 inhabitants. 6 Of course, this excludes the informal sector: for example, grandparents.

 C The editors of the Scandinavian Journal of Economics 2010.

Price of high-quality daycare and female employment 573

of the parents takes formal, publicly supported maternity or childcare leave, as described below. Municipalities provide both nursery centers (these may be integrated with kindergartens for 3–5-year-olds) and family daycare for children within this age group, and the local government is free to decide on the distribution of these two types of care within the municipality. Similarly, opening hours may vary across municipalities, but must “cover local need”. In general, opening hours during weekdays are between 6.30am and 5.15pm. Nursery centers may be owned by the municipality. No matter the owner status, the municipalities are required by law to monitor the institutions closely with respect to educational content as well as safety and hygiene. Evaluation of the former amounts to ensuring that the personnel have the necessary qualifications, and the latter includes accident-prevention measures, playgrounds, transport, sleeping facilities, toys, hygiene, and insurance schemes. The bulk of the costs of running nursery centers go to teacher wages. In a typical nursery center, around 80% of the budget is allocated to wages, 10% to rent, heating, and water, while the remaining 10% is operating costs (food, toys, sand, etc.). Teacher wages are centrally negotiated; that is, all teachers earn the same basic wage conditional on experience.7 On top of the basic wage comes a regional-specific premium, which is dependent on costs of living in the area. In contrast, family daycare takes place in private homes, and the carers are directly employed by the municipality. Again, the municipalities must approve the facilities and the qualifications of the carer. There may be up to five children in each home, and in some municipalities the carer’s own children under the age of three are allowed to enter into the total number of children in the family daycare. The carer will then receive compensation from the municipality for taking care of his or her own children. Costs of running family daycare are, not surprisingly, also dominated by wages to carers. Again, carer wages are centrally negotiated and consist of a basic wage and a regional-specific premium.8

Policies Table 1 summarizes the most important Danish policies in this area. It is seen that price variation stems from three sources: (1) municipalities, (2) parental income, and (3) the presence of other siblings. Municipalities can choose to provide guaranteed access to daycare (henceforth GADC), either in nursery centers or in family daycare, for all preschool children older than six months, though the parents cannot 7 8

Management responsibilities are rewarded with a (centrally negotiated) wage premium. Carers may receive a smaller qualification bonus if they have special training.  C The editors of the Scandinavian Journal of Economics 2010.

574

M. Simonsen

Table 1. Relevant family-related policies, Denmark 2001 Municipality-specific Childcare Policies Guaranteed Access to Guarantee of access to either family daycare or nursery center within Daycare, GADC municipality. Parents cannot choose a specific institution. Valid for children of six months or older. Prices, P Price maximum of 33% of total costs in case of GADC vs. 30% with no GADC. Prices reduced with lower income (see Table 2); 50% further price subsidy for additional siblings in public care. National Leave Policies Maternity Leave Childcare Leave

All mothers have the right to 28 weeks of job-protected leave. Compensation depends on union status and sector of employment but minimum €1,800 per month. All mothers have the right to 26 weeks of job-protected leave following maternity leave. Compensation €1,100 per month.

themselves decide on a specific institution. When providing GADC, the municipality is required by law to pay a minimum of 67% of the total costs per child. However, if the municipality chooses not to provide the guarantee, it is required to pay at least 70% of the total costs per child in daycare. If a municipality is unable to honor the guarantee, it must immediately switch to the higher-cost subsidy. A significant number of municipalities do not, as a matter of fact, provide GADC. In 2001, this amounted to 19.3% of all municipalities (affecting 25% of the women in the estimation sample; see Table 3). Thus, the incentive scheme seems too weak to secure complete immediate publicly provided daycare for all children. In case of waiting lists, open slots in childcare are allocated according to age and length of time on the waiting list. Depending on household income, municipalities take on an even larger part of the financing than the minimum 67%/70%. Table 2 shows further municipality subsidies for a family with one child. The critical income levels are increased by €1,000 for each additional sibling under the age of 18. Moreover, parents only pay the full price of daycare for the child Table 2. Daycare subsidies beyond minimum subsidy, family with one child Further subsidy (% of total costs per child)

Family income (2001) €16,215 or below €16,215 to €16,573 €16,573 to €50,300 €50,300 or above

30% 28.5% Subsidy reduced by 0.285 percentage points when income is increased by €337 0%

Note: Critical income levels increased by €1,000 for each additional sibling under the age of 18. Source: http://www.retsinfo.dk.  C The editors of the Scandinavian Journal of Economics 2010.

Price of high-quality daycare and female employment 575 Table 3. Moving patterns in estimation sample Share who moves, 1999–2001 0.072 Share who moves, 2000–2001 0.137 Among 2000–2001 movers: Share that stays within the region 0.488 Among 2000–2001 movers, share that move to municipalities where: Price is higher by 0%–10% 0.245 Price is lower by 0%–10% Price is higher by 10%–20% 0.111 Price is lower by 10%–20% Price is higher by 20%–30% 0.033 Price is lower by 20%–30% Price is higher by more than 30% 0.007 Price is lower by more than 30%

0.363 0.142 0.061 0.025

placed in the most expensive type of public childcare. For other siblings in public care, they pay 50% of the price. Thus the effective price of enrolling the youngest child in care depends on whether older siblings are enrolled. If at least one older sibling is enrolled, the incremental cost of enrolling the younger sibling is the price (given family income) minus half of the price of the cheaper type of care. Having older siblings enrolled in care therefore reduces the level as well as the variance of the effective price of care for the younger child. The average price of daycare without any further subsidies amounts to €280 per month and varies between €160 and €420. In my estimation sample, 0.9% of the women receive full coverage six months after giving birth, while 17% receive some further subsidy, and 82% receive no further subsidy. In the period under consideration, mothers had the right to a maximum of 28 weeks of job-protected maternity leave, and the vast majority of mothers took full advantage of this; see footnote 2. The degree of compensation while on maternity leave varied with sector of employment and union membership, but with a legally ensured lower bound on benefits received amounting to 100% of unemployment insurance benefits, or approximately €1,800 per month in 2001. In addition to maternity leave, parents had the right to 26 weeks of paid childcare leave before the child’s first birthday. While on childcare leave, parents received 60% of unemployment insurance, or about €1,100 per month in 2001. Some municipalities provided a limited supplement to these benefits amounting to a maximum of €400 per month. Unfortunately, only information on whether the supplement is provided, and not the exact amount, is available.9,10 Thus, potential provision of GADC 9 Including a dummy for leave supplement in the estimations below does not change the conclusions regarding the other parameters of interest in the paper. Results are available on request. 10 Childcare leave from employment is job-protected. Childcare leave during unemployment prolongs the period of unemployment insurance benefit entitlement and temporarily delays workfare participation.

 C The editors of the Scandinavian Journal of Economics 2010.

576

M. Simonsen

coincided with the expiration of formal maternity leave, but mothers of children who were not granted a slot in daycare had the possibility of taking childcare leave.

III. Identification and Estimation Strategy In this section, I will discuss the objectives of the econometric analysis and notation. I then consider the parameters of interest along with the identification and estimation strategy. The aim of the evaluation is to measure the impact of a given treatment on an outcome variable. Here, the treatment is childcare prices, and the outcome of interest is female employment status in the period after giving birth. Let Y ∗ be the underlying utility of employment net of costs. I follow Connelly (1992), Ribar (1992), and Kimmel (1998) with respect to the behavioral model forming the basis for the empirical analysis. That is, Y ∗ is linear and additively separable: Y ∗ = β0 + β1 P + γ X + u,

(1)

where P is the price to the parents of placing a child in daycare, and this price P is a function of municipality policy, parental income, number of older siblings, and whether siblings are placed in public daycare. I do not have information on whether children are actually placed in daycare. Therefore, to construct the price for a child, I assume that parents do not place the youngest in daycare and at the same time keep older preschool siblings at home after birth. This assumption does not seem too strict given the share of children in public daycare, as shown earlier. X is a set of conditioning variables including GADC, regional dummies, labor market experience information, labor market interruption information, woman’s education, husband’s income, and number of siblings in different age groups. Furthermore, even though operating procedures for publicly provided childcare are highly regulated, municipalities do have some discretion. To account for potential quality differences, I therefore condition on the child–teacher ratio in nursery centers on the municipality level.11 See Table 3 for a detailed description of the variables used in the estimations. I do not explicitly include expected wages in the participation equation, since the parameters of interest are related to childcare policies and not wages.12 However, I do include determinants of the wage.13 I assume u to 11

Some of the variables in the conditioning set are potentially endogenous. For example, mother’s leave-taking as well as child–teacher ratios may be problematic. In the sensitivity analysis below, I re-estimate my models excluding these variables. 12 Estimating a Heckman selection model a` la Kimmel (1998) does not change the conclusions from the analysis presented below. Results are available on request. 13 This means, for example, that the estimated parameter from level of experience and education includes both a wage effect and a participation effect; see also Tekin (2004).  C The editors of the Scandinavian Journal of Economics 2010.

Price of high-quality daycare and female employment 577

be a standard normally distributed random variable, and the parameters of (1) can then be estimated using a probit where  1 if Y ∗ > 0; Y= 0 otherwise. Clearly, to avoid selection bias in the estimation of the effects of childcare policy, there must be no dependence between the explanatory variables and u. In other words, there must be no unobserved factors that explain both labor force participation and policy. Conditioning on X is meant to capture such correlation between policy variables and u; the variables in X are not of interest by themselves. Furthermore, there cannot be any settlement effects; that is, couples must not move because of lower childcare prices. Is this likely and, if so, is it possible to counterbalance such effects? First, note that according to Norstrand and Andersen (2002), Danes mainly settle in connection with their choice of educational institution. After this initial settlement, they rarely move between municipalities and hardly ever between regions; see Dilling-Hansen and Smith (1996). Table 3 shows that approximately 7% of the women in my sample move between 2000 and 2001, and 14% move between 1999 and 2001. Of the women who move between 2000 and 2001, 49% stay within the region (see Table 5 for definitions of regions), suggesting strong geographical ties. Furthermore, moves are most often associated with smaller changes in childcare prices: 61% move to a municipality where the 2001 price lies within 10% of the 2001 price of the old municipality, and 86% move to a municipality where the price lies within a 20% bandwidth. Second, there is municipalityspecific variation in prices over time. A couple therefore cannot be sure that a municipality will not change its policy. Table 4 documents this. Third, it is unlikely that the childcare price is the main driver for settlement when compared to job opportunities and prices of real property. A back-of-the-envelope calculation shows that the gain from a 20% price Table 4. Price variation in municipalities, 1997–2003 Share of muncipalities with

Year

Average price (€) (municipality level)

Average growth

1997 1998 1999 2000 2001 2002 2003

206 215 224 235 253 269 283

· 0.041 0.045 0.048 0.079 0.062 0.053

for 7 years for 6 years for 5 years for 4 years for 3 years for 2 years for 1 year in 0 years

Higher than average price

Higher than average growth

0.163 0.059 0.089 0.056 0.107 0.083 0.130 0.300

· 0.004 0.052 0.230 0.311 0.274 0.111 0.018

 C The editors of the Scandinavian Journal of Economics 2010.

578

M. Simonsen

Table 5. Detailed description of variables Variable Employment, month 5–15 GADC P Child–teacher ratio Husband’s income Own income Labor market experience Region 1 Region 2 Region 3 Region 4 Region 5 Region 6 Region 7 Region 8 High school Short further education Medium further education Long further education Long-term unemployed Medium-term unemployed Long-term leave Medium-term leave No. of siblings, 0–2 years No. of siblings, 3–6 years No. of siblings 7–9 years No. of siblings 10–14 years No. of siblings 15–17 years Unemployment rate

Description Employed in month 5–15 following childbirth. Includes part-time employment, excludes self-employment, Jan.–Dec. 2001 Whether a municipality expects to provide guaranteed access to daycare in 2001 Monthly family-specific € price of family daycare in municipality, 2001 Child–teacher ratio in nursery centers, 2001 Yearly € income of husband in 2000 Yearly € income in 2000 Actual years of experience prior to birth Residing in county of Copenhagen, 2001 Residing in counties of Frederiksborg and Roskilde, 2001 Residing in counties of Western Sealand and Storstroem, 2001 Residing in county of Funen, 2001 Residing in counties of Southern Jutland and Ribe, 2001 Residing in counties of Vejle and Ringkoebing, 2001 Residing in counties of Aarhus and Viborg, 2001 Residing in county of Northern Jutland, 2001 10–12 years of education 13–14 years of education 15–16 years of education

Source Statistics Denmark

Ministry of the Interior Ministry of the Interior Ministry of the Interior Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark

More than 16 years of education Unemployed more than 26 weeks in year prior to giving birth Unemployed 13–25 weeks in year prior to giving birth On publicly funded leave more than 26 weeks in year prior to giving birth On publicly funded leave 13–25 weeks in year prior to giving birth No. of siblings age 0–2 years prior to birth

Statistics Denmark Statistics Denmark

No. of siblings age 3–6 years prior to birth

Statistics Denmark

No. of siblings age 7–9 years prior to birth

Statistics Denmark

No. of siblings age 10–14 years prior to birth No. of siblings age 15–17 years prior to birth Share of unemployed among women in municipality, 16–49 years, 2001

Statistics Denmark

Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark

Statistics Denmark Ministry of the Interior Continued

 C The editors of the Scandinavian Journal of Economics 2010.

Price of high-quality daycare and female employment 579 Table 5. (Continued) Variable Single-parent children Asylum seekers Third-world immigrants Social Democrats Conservatives Liberals Child families

Description

Source

Share of single-parent children 0–17 years old in municipality, 2001 No. of asylum seekers per 10,000 inhabitants in municipality, 2001 No. of third-world immigrants per 10,000 inhabitants in municipality, 2001 Largest party in 1997 municipality election is Social Democrats Largest party in 1997 municipality election is Conservatives Largest party in 1997 municipality election is Liberals Share of families with children among all households in municipality

Ministry of the Interior Ministry of the Interior Ministry of the Interior Statistics Denmark Statistics Denmark Statistics Denmark Statistics Denmark

decrease in the average price for a family with two children in public care for five years is just below the broker costs of selling an average house in Denmark (the actual costs of moving are more than simply the broker costs).14 I realize, though, that childcare policy is likely to be correlated with other municipality-specific characteristics. These may affect, on the one hand, the woman’s (or couple’s) decision about where to live and, on the other hand, the municipality’s capability of providing services in general. Short of perfect measures for this, the conditioning set is augmented with municipality characteristics (as in Gustafsson and Stafford, 1992), including the level of female unemployment rates for the age group under consideration, the share of single parents, the share of third-world immigrants, as well as the share of asylum seekers. Finally, to account for voter preferences for family-friendly policies, I condition on the outcome of the most recent municipality election (largest party being Social Democrats, Conservatives, or Liberals) as well as the share of families with children out of all households in the municipality. A similar issue is whether the most career-oriented families use their voice to affect childcare prices. One way to address this is to estimate conditional correlations between family types and prices. A negative correlation between measures of career-mindedness and prices may be indicative of such a problem. In particular, I run a simple regression of log prices (municipality level) on the full set of explanatory variables in my main analysis below. This regression suggests that neither level of education nor labor market experience correlate with prices. 14

Here I assume no discounting, and house prices are set equal to the tax authorities’ evaluation, which systematically lies below sales prices; see http://www.skat.dk. The average price in 2001 was about €190,000. Broker costs are assumed to be 4% of the sales price. This corresponds to the average in Denmark; see http://www.familieadvokaten.dk  C The editors of the Scandinavian Journal of Economics 2010.

580

M. Simonsen

The only exceptions are registered for individuals with more than 16 years of education (compared to less than 12 years), which is associated with a slightly higher price (2.4%) and husband’s income, which is associated with a 0.02% higher price for every additional €1,000 income. Important predictors of prices are the child–teacher ratio (+), largest political party, and the share of families with children in the municipality (−). A separate issue is that having older siblings reduces the effective price of childcare for the youngest child. The problem is that fertility is unlikely to be exogenous in this setting. For example, daycare arrangements may directly affect the number of children in a family. The optimal solution would be to have an instrument available. Since such a variable is not on hand, I pursue two other solutions. First, in my estimations, I condition on the number of older siblings in each of five different categories (a description of variables follows). This effectively steals all the price variation due to family composition such that the estimated price effects are only driven by price variation across municipalities. Second, I perform a sensitivity analysis, dividing my sample into families with only one child and families with more than one child to investigate whether the results are driven by differences in intra-family fertility. Unfortunately, it is impossible to identify P for all women in both counterfactual states, since (family) income depends on the employment decision. One strategy would be to perform a Heckman-type estimation to predict labor market income and thus P. This would, however, require an exclusion restriction in the employment equation to avoid strong reliance on functional form assumptions. Such a variable is not at my disposal; therefore, I rely on an alternative strategy. I limit the sample to a group of women for whom P is independent of income by excluding those whose compensation does not exceed the upper limit in Table 1 (€50,300 corrected for number of siblings), both when employed and not employed. This requires information about the woman’s compensation in both states, while the husband’s income, H INC, is taken as given. Unemployed individuals receive unemployment insurance benefits, UI, or approximately €21,800 per year in 2001. All the individuals in my sample will be eligible for UI benefits after the expiration of childcare leave. As described earlier, compensation during childcare leave amounts to 60% of UI,15 or approximately €13,100 per year in 2001. I assume that compensation in the state of employment always exceeds this lower bound. Therefore, the smallest level of compensation during non-employment is 60% of UI, and family income, F INC, will at least equal minimum income, MIN INC, for all mothers during both employment and non-employment: MIN INC ≡ H INC + 0.6 · UI  F INC. 15

See earlier discussion of childcare leave during unemployment.

 C The editors of the Scandinavian Journal of Economics 2010.

Price of high-quality daycare and female employment 581

For the subsample of individuals with MIN INC (and therefore family income) always above the appropriate upper limit, P is independent of income. In principle, more individuals belong to this group (82% of all mothers six months after childbirth). For example, some women can be observed in non-employment with a family income higher than the upper limit. Assuming that income in employment always exceeds income in non-employment, these women could be included. The problem, however, is that I do not observe income in non-employment for employed women. Such an inclusion will therefore treat non-employed and employed women asymmetrically, whereas imposing a common minimum income during non-employment will not. Obviously, the selection procedure means that the estimated parameters only hold true for this group consisting of approximately 60% of mothers following childbirth. I estimate female employment propensities on a monthly basis from five months after childbirth. Month five is included as a (weak) consistency check. Since daycare is not provided in any municipality before month six, a significant effect will indicate a misspecification of the model and/or that the analysis has excluded factors that simultaneously explain a municipality’s choice of policies in general and female employment.

IV. The Data I employ register-based data maintained by Statistics Denmark along with municipality information supplied by the Ministry of the Interior from 2001. The register dataset contains information on a representative sample of 10% of all Danish individuals in the age bracket 15–74. Information stems from several registers maintained by Statistics Denmark. The registers include yearly information on income and demographics. Furthermore, the individual event history in terms of periods of employment, unemployment, retirement, maternal leave, publicly subsidized leave (childrearing or sabbatical), education, and the residual category of non-participation is known on a monthly basis. A woman is coded as employed only if she is working. Importantly, a child register provides exact information about the date on which women in the sample give birth. This information allows for identification of the labor market status of a woman in each month after giving birth. The information from the Ministry of the Interior includes the female unemployment rates, the share of single parents, the share of third-world immigrants, the share of asylum seekers, the municipality-specific prices of childcare, and whether the municipality expects to be able to provide GADC. This last piece of information is reported to the Ministry of the Interior as part of the municipality budget requirements.  C The editors of the Scandinavian Journal of Economics 2010.

582

M. Simonsen

Table 5 gives a detailed description as well as the timing of measurement and the source of the variables used in the estimations. Self-employed individuals, women with no labor market insurance, women with early retirement, and women enrolled in the educational system are excluded from the analysis in order to secure a homogeneous estimation sample representing women for whom the choice of being employed is real. This excludes 10% of the full sample. To estimate the effect of childcare policies on the female labor supply five months after giving birth, I choose as my outcome variable the labor market status in January 2001 for women giving birth in July 2000, the labor market status in February 2001 for women giving birth in August 2000, and so forth. To estimate the effect of childcare policies six months after giving birth, I use labor market status in January 2001 for women giving birth in June 2000, labor market status in February 2001 for women giving birth in August 2000, etc. The estimation samples used to evaluate the decisions 7–15 months after childbirth are constructed analogously. Note that the monthly samples only partly overlap. As will become clear, it is possible to establish a distinct pattern of the effects of the policies within the first 15 months after childbirth. Effectively, considering only this age range also excludes the possibility of younger siblings, which would greatly complicate not only the modeling of mothers’ choices but also how to determine the price of childcare.16 Therefore, I limit the analysis to cover 16

The limited available information on childcare policies restricts me from meaningfully estimating a discrete time duration model. Assume that I were to construct a flow sample to be used for duration analysis. The starting point (in real time) would be January 2001. Here, to consider the decision to return five months after childbirth, I could include children born in July 2000 and subsequently include mothers of children that flow into the sample (those with children born from August 2000 through June 2001). Therefore, the sample used to consider the decision to return to the labor market after five months would be the same in the two types of models (contributions from 12 calendar months in total). To consider the decision to return to the labor market six months after giving birth in the duration model set-up, however, I could only use children born from July 2000 through May 2001 (contributions from 11 calendar months) as opposed to the larger sample used in the paper. Including mothers of children born in June 2000 as well (or equivalently, to change the real-time starting point to December 2000) is not possible because the children would be six months old in December 2000, for which I do not have the necessary childcare policy information. To consider the decision to return to the labor market 15 months after childbirth, I would only be able to exploit information for mothers of children born in July and August 2000. In addition, in the discrete time duration model, only the at-risk population is used for identification. Thus the effective sample used to estimate the decision to return after 15 months would be even smaller. Moreover, apart from only using the at-risk population, a discrete time duration model where one allows the coefficient estimates of the explanatory variables to vary every month is actually very similar to the approach taken in the paper as it stands. Finally, Figure 2 shows that women enter employment relatively quickly after the expiration of formal maternity leave, indicating little risk of duration dependence. See also Lechner (2000) and Larsson (2003) for examples of papers evaluating period-specific effects.

 C The editors of the Scandinavian Journal of Economics 2010.

Price of high-quality daycare and female employment 583 Table 6. Comparison, excluded and included groups, six months after childbirth Excluded group

GADC Age No. of siblings 0–2 years No. of siblings 3–6 years No. of siblings 7–9 years No. of siblings 10–14 years No. of siblings 15–17 years Own income (€) Husband’s income (€) Childcare price, P—family-specific (€) Childcare price—municipality level (€) Child–teacher ratio High school Short further education Medium further education Long further education Long-term unemployed Medium-term unemployed Long-term leave Medium-term leave Single-parent children Asylum seekers (per 10,000) Third-world immigrants (per 10,000) Unemployment rate Social Democrats Conservatives Child families Observations Share of all

Included group

Average

Std. dev.

Average

Std. dev.

0.77 30.06 0.13 0.39 0.14 0.10 0.02 25,528 24,355 169 287 3.56 0.15 0.35 0.18 0.06 0.02 0.05 0.02 0.51 0.15 21 354 6.12 0.67 0.06 0.23

0.42 4.77 0.34 0.58 0.37 0.35 0.14 9,691 12,649 87 27 0.36 0.36 0.48 0.38 0.23 0.13 0.21 0.13 0.50 0.05 76 239 1.86 0.47 0.24 0.05

0.77 31.43 0.13 0.47 0.13 0.07 0.01 29,932 58,451 209 282 3.56 0.11 0.42 0.32 0.12 0.02 0.04 0.01 0.59 0.14 19 326 5.76 0.58 0.10 0.23

0.42 4.12 0.34 0.58 0.37 0.30 0.12 12,129 32,776 75 29 0.37 0.32 0.49 0.42 0.32 0.13 0.20 0.12 0.49 0.04 69 217 1.77 0.49 0.30 0.04

2,037 0.41

2,883 0.59

Note: Bold indicates a significant difference between the two groups at the 5% level.

this period. I have about 5,000 observations for each month. As explained in Section III, in order to identify P when a woman is employed and also in the counterfactual case, I am forced to consider only women for whom P is independent of income. This step also excludes single women (i.e., women who are neither cohabiting nor married) since 60% of UI obviously does not exceed the appropriate upper limit in Table 2. The final sample consists of 3,000 observations for each month. Table 6 compares the characteristics of the excluded and included groups of women six months after childbirth. Note that the price variation is still considerable within the included group.17 Clearly, the included group of The mean family-specific price in included families with only one child is €283, and the standard deviation is €30. The corresponding figures in included families with more

17

 C The editors of the Scandinavian Journal of Economics 2010.

584

M. Simonsen Employment propensity 0.8 0.6 0.4 0.2 0 5

6

7

8

9 10 11 12 13 14 15

Months after childbirth Fig. 1. Employment propensity after childbirth, estimation sample, 2001

women has different characteristics compared to the excluded group. In particular, husbands’ income is higher among the included women due to the selection criteria. In addition, the group of included women comprises individuals who are older, have slightly more children, have higher education, have higher income, are more likely to have been on leave, and live in municipalities with lower social costs (single-parent children, asylum seekers, and third-world immigrants) and unemployment rates. In this sense, the 41% of women excluded from the analysis have less favorable socioeconomic characteristics than those included. Therefore, their reactions to changes in policies may potentially differ from those of the included population. With the available data, however, this cannot be investigated. Figure 1 depicts female employment 5–15 months after giving birth for the included group. We see that only 4% of women are employed five months after giving birth.18 Hereafter, we observe a gradual increase in the female employment propensity. After 14 months, the profile levels out at 71%. This corresponds to the average propensity for Danish women. Table 7 gives information on regional variations in childcare policies.

V. Estimation Results This section presents selected results from the analysis discussed in Section III. The full set of estimation results conditioning on the set of variables described in Table 5 is shown in Tables A1–A4. The price of childcare seems to be very important for the female employment propensity than one child are €158 and €49. Clearly, part of the total variation in prices stems from differences in intra-family fertility, yet there is still substantial variation in the price after conditioning on the number of children. 18 During the first four months after childbirth, the share of women participating in the labor market is even smaller.  C The editors of the Scandinavian Journal of Economics 2010.

Price of high-quality daycare and female employment 585 Table 7. Regional variation in childcare policies

Region 1 2 3 4 5 6 7 8

Share of population

Share with GADC

Average familyspecific price, P (€)

Min. familyspecific price, P (€)

Max. familyspecific price, P (€)

0.23 0.12 0.09 0.09 0.09 0.12 0.17 0.18

0.43 0.58 0.69 0.95 0.99 0.88 0.92 1.00

190 198 169 160 173 182 178 182

0 0 0 0 0 0 0 0

356 414 299 302 325 321 299 316

Note: For definition of regions, see Table 3.

in the short term. Figure 2 shows the marginal effects (evaluated at the sample mean) of increasing monthly family-specific prices by €1. The marginal effects of a price increase on the female labor supply are significantly negative and vary between −0.0006 and −0.0012 8–12 months after childbirth. This should be seen in a context where the standard deviation in observed prices is €75 and the average employment rate in the period increases from 0.36 eight months after childbirth to 0.58 12 months after childbirth. The marginal effects correspond to average crossprice elasticities (the percentage change in the propensity for women to be employed as per a percentage change in the price of childcare) between −0.13 and −0.25. Interestingly, there is a negative effect of increases in prices during the months where women have the possibility of taking childrearing leave. This option makes it much less costly to temporarily opt out of employment because leave is job-protected. At the child’s first birthday, the only alternatives to employment are non-participation and (voluntary) unemployment. Hence, it makes perfect sense that women are less priceresponsive from 13 months after childbirth onwards. The (short-term) cross-price elasticities for Danish cohabiting mothers turn out to be in the lower end compared with cross-price elasticities found in the literature for the US. Yet the estimates remain close. Given the cross-country differences in childcare regimes and compensation schemes during parental leave, some disparity is expected. Among papers comparable in terms of the underlying model, Connelly (1992) finds a cross-price elasticity of −0.20 (a replication by Kimmel, 1998, including a richer set of conditioning variables results in an elasticity of −0.42), Ribar (1992) finds an elasticity of −0.74, and the results from Kimmel (1998) indicate a cross-price elasticity of −0.92 for married mothers. The other Nordic results differ to some extent from the findings of this paper. Gustafsson and Stafford (1992) consider families with precisely one child and find an  C The editors of the Scandinavian Journal of Economics 2010.

586

M. Simonsen Marginal effects 0.0005 0.0000 –0.0005

5

6

7

8

9 10 11 12 13 14 15

–0.0010 –0.0015 –0.0020 Months after childbirth

Fig. 2. Marginal effects on female employment propensity of childcare prices after childbirth Note: With 95% confidence bounds (based on normal approximation).

elasticity of −0.06 for all individuals and −1.88 for those not subject to rationing.19

Sensitivity Analysis Extensive sensitivity analysis has been performed to check the validity of the results. I find that the parameters of interest are extremely robust to the choice of specification. However, the precision of the estimates is affected when considering smaller subsamples; see Table A5 for the results. First, I investigate whether the estimated parameters vary across families with different numbers of children (see previous discussion). Specifically, I re-estimate my model using only one-child families and families with more than one child. Also, because the price sensitivity may be different in municipalities with and without GADC, the model is re-estimated for GADC and non-GADC municipalities. Further, one might hypothesize that labor markets in larger cities are different from those of the provinces, and that this may affect childcare policies as well. To address this point, I re-estimate the model, first excluding the municipality of Copenhagen (the Danish capital and largest city, with 500,000 inhabitants), and next excluding both Copenhagen and Aarhus (the second-largest city, with 280,000 inhabitants).20 Finally, I exclude potential endogenous variables from my conditioning set. In particular, I drop (a) lagged endogenous variables, such as interruption information (unemployment and leave-taking), (b) interruption 19

In the following section, I re-estimate my model using only one-child families. The results are not sensitive to this. 20 This is more flexible than just including a dummy for living in a large city.  C The editors of the Scandinavian Journal of Economics 2010.

Price of high-quality daycare and female employment 587

information as well as labor market experience, (c) child–teacher ratios, and (d) all municipality-level information (including child–teacher ratios). None of the above sensitivity checks affect the main conclusions of the analysis, and none of the estimated marginal effects of price increases are statistically significantly different from the results presented above.21

Effects of GADC This subsection investigates whether availability of childcare leads women to return to the labor market earlier after giving birth; see also Gustafsson and Stafford (1992). In particular, I analyze the effects of providing GADC. The effect of living in a municipality providing GADC should be interpreted as the effect of having daycare available when the mother wants to return to the labor market compared to a situation where daycare is possibly not available until a later (undefined) point in time.22 Clearly, the effects of the policy depend on the length of the waiting lists in municipalities that do not provide the guarantee. See Simonsen (2005) for descriptive evidence on the samples of women living in the two types of municipalities. Figure 3 shows the distribution of prices in the two types of municipalities. We see that the distribution of prices in municipalities that do not provide GADC has fatter tails. More municipalities within this category have both very low and very high prices. Therefore, although municipalities with GADC have significantly lower prices of daycare, there is no systematic relationship between the provision of unrestricted access to daycare and price structure. It should thus be possible to identify both an effect of GADC and a price effect. Figure 4 depicts the marginal effects (evaluated at the sample mean) on the female employment propensity of providing GADC 5–15 months after childbirth. We see that the provision of GADC has a positive and significant effect on the female employment propensity 8–13 months after childbirth. The effect is relatively large. During this period, women in GADC municipalities have around a 6% higher employment propensity. Unrestricted access to daycare therefore seems to enable women to return to employment after childbirth. 21

It may be surprising that the price sensitivity does not vary between GADC and non-GADC municipalities. A possible explanation is that the effect of GADC is still relatively small, compared to the share employed in each month (e.g., 6.1 percentage points in month nine compared to an average employment propensity of 0.4), and any differences in price sensitivity must be driven by women affected by GADC. 22 The same is true, for example, for most studies attempting to evaluate the effects of labor market training: individuals do not participate in training in a given period but may participate at a later (undefined) point in time.  C The editors of the Scandinavian Journal of Economics 2010.

588

M. Simonsen

Fig. 3. Distribution of monthly prices of daycare (€) in municipalities with and without guaranteed access to daycare

Marginal effects 0.15 0.10 0.05 0.00 5

6

7

8

9 10 11 12 13 14 15

–0.05 Months after childbirth Fig. 4. Marginal effects on female employment propensity of providing guaranteed access to daycare after childbirth Note: With 95% confidence bounds (based on normal approximation).

Note that there is no statistically or economically significant effect of the policy before month six. Similarly, the effect of GADC dies out over time. If the effects of GADC did not die out with time, it could either be because of very long waiting lists in non-GADC municipalities, or because women who reside in municipalities that provide GADC have unobserved characteristics that increase their employment propensity in all time periods.  C The editors of the Scandinavian Journal of Economics 2010.

Price of high-quality daycare and female employment 589

VI. Discussion This paper presents empirical evidence on the effects of increases in the marginal childcare price following expiration of formal maternity leave six months after childbirth. The sample consists of women living in couples. I analyze the case of Denmark, where the quality of public daycare is considered high in an international comparison. I exploit price variation across municipalities. To overcome the problem of identifying childcare prices for non-employed women, I exploit the fact that families with incomes above a certain threshold face prices that are independent of income. The downside of this approach is that the estimated parameters only hold true for the wealthier part of the population. The results show that the price effect is significantly negative. An increase in the price of childcare of €1 per month will decrease female employment by around 0.08%, corresponding to a cross-price elasticity of −0.17. This effect prevails during the first 12 months after childbirth. That is, using data from a regime like the Danish with highly subsidized public childcare, to uncover effects of childcare on female employment reveals a considerable price sensitivity. Also, the size of the effects, though short-lived, is in the lower end, but still comparable to effects found in other papers for other countries. Also, availability of childcare seems to be important for women’s return to the labor market. An important policy question is, of course, whether public provision of high-quality highly subsidized childcare “pays off” in the sense that it improves welfare. Clearly, the analysis in this paper is not sufficient to answer this ambitious question, but it does, nonetheless, provide important input for a future cost–benefit analysis.

Appendix A Table A1. Marginal effects and asymptotic standard errors from employment probit, 5, 6, and 7 months after childbirth 5 months Variable GADC P (€10) Child–teacher ratio Experience (1,000 years) Experience (10,000 years) squared Medium-term unemployed Long-term unemployed Medium-term leave Long-term leave

6 months

7 months

Marg. effect

Std. error

Marg. effect

Std. error

Marg. effect

Std. error

0.003 0.001 −0.020 −0.005 0.029 −0.007 · 0.016 0.001

0.009 0.001 0.006 0.003 0.014 0.017 · 0.036 0.008

0.011 −0.002 −0.033 0.014 −0.039 −0.179 −0.054 −0.122 0.048

0.020 0.002 0.015 0.006 0.034 0.029 0.038 0.039 0.017

0.018 −0.006 −0.034 0.026 −0.066 −0.147 0.020 −0.272 0.086

0.023 0.003 0.018 0.007 0.039 0.069 0.049 0.032 0.019 Continued

 C The editors of the Scandinavian Journal of Economics 2010.

590

M. Simonsen

Table A1. (Continued) 5 months Variable Husband’s income (€100,000) High school Short further education Medium further education Long further education No. of siblings 0–2 years No. of siblings 3–6 years No. of siblings 7–9 years No. of siblings 10–14 years No. of siblings 15–17 years Unemployment rate Single-parent children Asylum seekers (per 1,000,000) Third-world immigrants (per 1,000,000) Social Democrats Conservatives Child families

6 months

7 months

Marg. effect

Std. error

Marg. effect

Std. error

Marg. effect

Std. error

0.004 0.023 0.005 0.015 0.038 0.007 −0.010 0.006 0.005 0.007 −0.001 0.003 −0.005 0.002 0.003 −0.014 −0.055

0.009 0.020 0.013 0.016 0.022 0.015 0.014 0.009 0.011 0.037 0.003 0.002 0.007 0.003 0.009 0.010 0.152

0.007 0.037 0.024 0.030 0.049 −0.065 −0.008 −0.010 0.024 −0.020 0.006 −0.003 0.013 0.005 0.007 −0.032 −0.154

0.022 0.035 0.027 0.030 0.036 0.033 0.027 0.020 0.025 0.072 0.007 0.004 0.011 0.007 0.020 0.029 0.309

−0.071 0.032 −0.006 −0.002 0.059 −0.047 −0.095 −0.041 −0.023 0.086 0.001 −0.003 0.015 −0.002 0.003 0.019 0.348

0.033 0.038 0.031 0.033 0.040 0.040 0.033 0.023 0.030 0.095 0.008 0.005 0.013 0.008 0.024 0.036 0.357

No. of observations Employment propensity Pseudo-R 2

2,835 0.044 0.038

2,931 0.212 0.028

2,937 0.326 0.044

Notes: Regional dummies are included; bold coefficients are significant at the 5% level; Huber–White robust standard errors. Results robust to clustering at the municipality level.

Table A2. Marginal effects and asymptotic standard errors from employment probit, 8, 9, and 10 months after childbirth 8 months Variable GADC P (€10) Child–teacher ratio Experience (1,000 years) Experience (10,000 years) squared Medium-term unemployed Long-term unemployed Medium-term leave Long-term leave Husband’s income (€100,000) High school Short further education Medium further education Long further education No. of siblings 0–2 years No. of siblings 3–6 years No. of siblings 7–9 years

9 months

10 months

Marg. effect

Std. error

Marg. effect

Std. error

Marg. effect

Std. error

0.047 −0.006 −0.023 0.020 −0.040 −0.191 −0.016 −0.304 0.079 −0.128 0.029 −0.003 −0.011 0.067 −0.075 −0.109 −0.051

0.023 0.003 0.018 0.007 0.040 0.069 0.049 0.028 0.020 0.036 0.038 0.031 0.033 0.041 0.043 0.034 0.023

0.061 −0.008 −0.007 0.031 −0.089 −0.157 −0.103 −0.344 0.094 −0.110 0.049 0.029 0.016 0.134 −0.118 −0.135 −0.065

0.024 0.003 0.019 0.007 0.040 0.083 0.038 0.033 0.020 0.036 0.040 0.033 0.035 0.042 0.047 0.036 0.024

0.058 −0.010 −0.021 0.036 −0.106 −0.204 −0.134 −0.399 0.115 −0.083 0.020 0.034 0.025 0.185 −0.125 −0.169 −0.078

0.025 0.003 0.019 0.008 0.042 0.083 0.053 0.033 0.021 0.038 0.040 0.034 0.036 0.041 0.051 0.038 0.025 Continued

 C The editors of the Scandinavian Journal of Economics 2010.

Price of high-quality daycare and female employment 591 Table A2. (Continued) 8 months Variable No. of siblings 10–14 years No. of siblings 15–17 years Unemployment rate Single-parent children Asylum seekers (per 1,000,000) Third-world immigrants (per 1,000,000) Social Democrats Conservatives Child families No. of observations Employment propensity Pseudo-R 2

9 months

10 months

Marg. effect

Std. error

Marg. effect

Std. error

Marg. effect

Std. error

−0.046 0.068 −0.002 −0.003 0.009 −0.005 0.002 −0.007 0.515

0.031 0.092 0.008 0.005 0.013 0.009 0.024 0.036 0.365

−0.073 0.072 −0.002 0.003 −0.006 −0.015 −0.015 −0.029 1.179

0.032 0.091 0.009 0.006 0.015 0.009 0.025 0.037 0.395

−0.094 0.016 0.003 0.000 0.011 −0.005 −0.021 −0.016 0.883

0.033 0.091 0.009 0.006 0.015 0.009 0.026 0.038 0.403

2,929 0.343 0.051

2,964 0.398 0.061

2,944 0.451 0.073

Notes: Regional dummies are included; bold coefficients are significant at the 5% level; Huber–White robust standard errors. Results robust to clustering at the municipality level.

Table A3. Marginal effects and asymptotic standard errors from employment probit, 11, 12, and 13 months after childbirth 11 months Variable GADC P (€10) Child–teacher ratio Experience (1,000 years) Experience (10,000 years) squared Medium-term unemployed Long-term unemployed Medium-term leave Long-term leave Husband’s income (€100,000) High school Short further education Medium further education Long further education No. of siblings 0–2 years No. of siblings 3–6 years No. of siblings 7–9 years No. of siblings 10–14 years No. of siblings 15–17 years Unemployment rate Single-parent children Asylum seekers (per 1,000,000) Third-world immigrants (per 1,000,000) Social Democrats

12 months

13 months

Marg. effect

Std. error

Marg. effect

Std. error

Marg. effect

Std. error

0.061 −0.011 −0.031 0.039 −0.108 −0.243 −0.167 −0.390 0.127 −0.098 0.025 0.048 0.065 0.211 −0.207 −0.191 −0.096 −0.104 −0.024 −0.003 −0.002 0.026 0.002 −0.019

0.025 0.003 0.020 0.008 0.043 0.086 0.055 0.048 0.021 0.039 0.040 0.034 0.036 0.039 0.056 0.039 0.025 0.033 0.082 0.009 0.006 0.016 0.009 0.027

0.052 −0.008 −0.028 0.046 −0.123 −0.224 −0.170 −0.427 0.133 −0.110 0.023 0.044 0.084 0.198 −0.204 −0.165 −0.101 −0.124 0.010 −0.014 −0.005 0.020 0.009 −0.020

0.026 0.003 0.020 0.008 0.042 0.091 0.057 0.053 0.021 0.038 0.039 0.034 0.034 0.036 0.057 0.039 0.025 0.032 0.075 0.010 0.006 0.016 0.009 0.027

0.079 −0.007 −0.003 0.061 −0.174 −0.249 −0.167 −0.396 0.109 −0.124 0.021 0.049 0.108 0.222 −0.229 −0.162 −0.115 −0.131 −0.017 −0.010 −0.007 0.017 0.008 −0.006

0.026 0.003 0.020 0.008 0.042 0.094 0.057 0.066 0.021 0.037 0.038 0.033 0.033 0.031 0.060 0.038 0.024 0.031 0.072 0.009 0.006 0.018 0.009 0.027

Continued  C The editors of the Scandinavian Journal of Economics 2010.

592

M. Simonsen

Table A3. (Continued) 11 months Variable Conservatives Child families No. of observations Employment propensity Pseudo-R 2

12 months

13 months

Marg. effect

Std. error

Marg. effect

Std. error

Marg. effect

Std. error

−0.036 0.522

0.038 0.395

−0.005 0.462

0.038 0.401

0.018 0.315

0.038 0.386

2,980 0.489 0.087

2,961 0.545 0.100

2,925 0.612 0.129

Notes: Regional dummies are included; bold coefficients are significant at the 5% level; Huber–White robust standard errors. Results robust to clustering at the municipality level.

Table A4. Marginal effects and asymptotic standard errors from employment probit, 14 and 15 months after childbirth 14 months Variable GADC P (€10) Child–teacher ratio Experience (1,000 years) Experience (10,000 years) squared Medium-term unemployed Long-term unemployed Medium-term leave Long-term leave Husband’s income (€100,000) High school Short further education Medium further education Long further education No. of siblings 0–2 years No. of siblings 3–6 years No. of siblings 7–9 years No. of siblings 10–14 years No. of siblings 15–17 years Unemployment rate Single-parent children Asylum seekers (per 1,000,000) Third-world immigrants (per 1,000,000) Social Democrats Conservatives Child families

15 months

Marg. effect

Std. error

Marg. effect

Std. error

0.027 −0.006 −0.008 0.062 −0.162 −0.302 −0.179 −0.523 0.134 −0.157 0.023 0.024 0.119 0.201 −0.244 −0.122 −0.108 −0.100 0.040 −0.009 −0.003 0.015 0.002 0.017 0.007 −0.069

0.025 0.003 0.019 0.007 0.040 0.094 0.058 0.061 0.020 0.036 0.037 0.031 0.030 0.027 0.061 0.038 0.023 0.028 0.062 0.009 0.005 0.016 0.009 0.026 0.037 0.371

0.037 −0.003 −0.015 0.063 −0.179 −0.198 −0.142 −0.534 0.124 −0.140 −0.007 0.048 0.157 0.229 −0.227 −0.104 −0.092 −0.082 0.025 −0.010 −0.001 0.026 0.002 0.014 −0.027 −0.017

0.026 0.003 0.018 0.007 0.041 0.092 0.059 0.060 0.019 0.036 0.036 0.030 0.027 0.023 0.064 0.038 0.022 0.028 0.060 0.009 0.005 0.018 0.009 0.026 0.039 0.377

No. of observations Employment propensity Pseudo-R 2

2,899 0.669 0.166

2,868 0.676 0.166

Notes: Regional dummies are included; bold coefficients are significant at the 5% level; Huber–White robust standard errors. Results robust to clustering at the municipality level.  C The editors of the Scandinavian Journal of Economics 2010.

2,835

0.001 (0.001) −0.002 (0.002) −0.006 (0.003) −0.006 (0.003) −0.008 (0.003) −0.010 (0.003) −0.011 (0.003) −0.008 (0.003) −0.007 (0.003) −0.006 (0.003) −0.003 (0.003)

1,146

0.003 (0.003) 0.003 (0.005) −0.004 (0.006) −0.005 (0.006) −0.008 (0.006) −0.011 (0.007) −0.011 (0.007) −0.003 (0.007) 0.004 (0.006) 0.004 (0.006) 0.005 0.006

One child P (€10)

1,682

0.000 (0.001) −0.001 (0.003) −0.006 (0.004) −0.007 (0.004) −0.006 (0.004) −0.008 (0.004) −0.010 (0.004) −0.009 (0.004) −0.008 (0.004) −0.008 (0.004) −0.004 (0.004) 2,120

0.001 (0.001) −0.002 (0.002) −0.006 (0.003) −0.005 (0.003) −0.008 (0.003) −0.013 (0.004) −0.013 (0.003) −0.014 (0.004) −0.012 (0.004) −0.011 (0.003) −0.009 (0.004)

GADC municip. P (€10) −0.004 (0.004) −0.002 (0.005) −0.009 (0.006) −0.013 (0.006) −0.011 (0.006) −0.008 (0.007) −0.014 (0.007) −0.002 (0.007) 0.003 (0.007) −0.006 (0.007) 0.007 (0.007) 2,835

0.001 (0.001) −0.002 (0.002) −0.006 (0.003) −0.005 (0.003) −0.007 (0.003) −0.009 (0.003) −0.010 (0.003) −0.008 (0.003) −0.007 (0.003) −0.006 (0.003) −0.004 (0.003)

No GADC municip. P (€10)

699

Drop interruption variables P (€10)

2,835

0.001 (0.001) −0.002 (0.002) −0.006 (0.003) −0.005 (0.003) −0.007 (0.003) −0.009 (0.003) −0.010 (0.003) −0.008 (0.003) −0.007 (0.003) −0.006 (0.003) −0.004 (0.003)

Drop experience + interruption variables P (€10)

2,835

0.000 (0.001) −0.002 (0.002) −0.006 (0.003) −0.006 (0.003) −0.007 (0.003) −0.009 (0.003) −0.011 (0.003) −0.009 (0.003) −0.007 (0.003) −0.006 (0.003) −0.004 (0.003)

Drop child– teacher ratio P (€10)

2,835

0.000 (0.001) −0.002 (0.002) −0.007 (0.003) −0.006 (0.003) −0.008 (0.003) −0.010 (0.003) −0.011 (0.003) −0.009 (0.003) −0.007 (0.003) −0.006 (0.003) −0.004 (0.003)

Drop all municip. variables P (€10)

2,609

0.000 (0.001) −0.001 (0.002) −0.007 (0.003) −0.006 (0.003) −0.007 (0.003) −0.009 (0.003) −0.010 (0.003) −0.009 (0.003) −0.007 (0.003) −0.006 (0.003) −0.004 (0.003)

Drop Copenhagen P (€10)

2,451

0.001 (0.001) −0.002 (0.002) −0.008 (0.003) −0.007 (0.003) −0.007 (0.003) −0.009 (0.003) −0.010 (0.003) −0.008 (0.003) −0.007 (0.003) −0.005 (0.003) −0.003 (0.003)

Drop Copenhagen and Aarhus P (€10)

Notes: Bold indicates significance at the 5% level, italic indicates significance at the 10% level. No coefficients are statistically significantly different from those of the main specification.

No. of obs., 5 months

15

14

13

12

11

10

9

8

7

6

5

Months

Full sample P (€10)

More than one child P (€10)

Table A5. Marginal effects of price on female employment

Price of high-quality daycare and female employment 593

 C The editors of the Scandinavian Journal of Economics 2010.

594

M. Simonsen

References Averett, S. L., Peters, H. E., and Waldman, D. M. (1997), Tax Credits, Labor Supply, and Child Care, Review of Economics and Statistics 79, 125–135. Baker, M., Gruber, J., and Milligan, K. (2008), Universal Childcare, Maternal Labor Supply and Family-well-being, Journal of Political Economy 116, 709–745. Berger, M. C., and Black, D. A. (1992), Child Care Subsidies, Quality of Care, and the Labor Supply of Low-income, Single Mothers, Review of Economics and Statistics 74, 635–642. Connelly, R. (1992), The Effect of Child Care Costs on Married Women’s Labor Force Participation, Review of Economics and Statistics 74, 83–90. Dilling-Hansen, M., and Smith, V. (1996), Regional Mobility in Denmark, Danish Journal of Economics 134, 257–271. Gelbach, J. (2002), Public Schooling for Young Children and Maternal Labor Supply, American Economic Review 92, 307–322. Gustafsson, S., and Stafford, F. (1992), Child Care Subsidies and Labor Supply in Sweden, Journal of Human Resources 27, 204–230. Heckman, J. (1974), Effects of Child-care Programs on Women’s Work Effort, Journal of Political Economy 82, 136–169. Jaumotte, F. (2003), Female Labor Force Participation: Past Trends and Main Determinants in the OECD Countries, OECD Working Paper no. 376. Kimmel, J. (1998), Child Care Costs as a Barrier to Employment for Single and Married Mothers, Review of Economics and Statistics 80, 287–299. Larsson, L. (2003), Evaluation of Swedish Youth Programs, Journal of Human Resources 38, 891–927. Lechner, M. (2000), An Evaluation of Public-sector-sponsored Continuous Vocational Training Programs in East Germany, Journal of Human Resources 35, 347–375. Norstrand, R., and Andersen, A. K. (2002), Income, Settlement, and Education, AKF Forlaget, Copenhagen. OECD (2001), Early Childhood Education and Care Policy in Denmark, OECD country note. Powell, L. M. (2002), Joint Labor Supply and Childcare Choice Decisions of Married Mothers, Journal of Human Resources 37, 106–128. Ribar, D. C. (1992), Child Care and the Labor Supply of Married Women: Reduced Form Evidence, Journal of Human Resources 27, 134–165. Ruhm, C. J. (1998), The Economic Consequences of Parental Leave Mandates: Lessons from Europe, Quarterly Journal of Economics 113, 285–317. Ruhm, C. J. (2004), Parental Employment and Child Cognitive Development, Journal of Human Resources 39, 155–192. Simonsen, M. (2005), Availability and Price of High Quality Day Care and Female Employment, Working Paper no. 2005-8, School of Economics and Management, University of Aarhus. Society of Daycare Teachers (Pædagogisk Medhjælper Forbund) (2004), Efterspørgslen efter børnepasning (The Demand for Child Care). Tekin, E. (2004), Child Care Subsidy Receipt, Employment and Child Care Choices of Single Mothers, National Bureau of Statistics Working Paper no. 10459. First version submitted June 2007; final version received February 2010.

 C The editors of the Scandinavian Journal of Economics 2010.