1
Parental health effect of childcare utilization
for children ≤ 4 years
Empirical evidence from the Netherlands
Sanne van der Laan (S2734508)
Faculty of Economics and Business Economics, University of Groningen Master Thesis: Business Administration – Health
22-06-2020
Supervisor: prof. dr. G.J. van den Berg Co-assessor: prof. dr. M.J. Postma
Abstract: This thesis analyzes whether the use of childcare for children ≤ 4 years affects the health of the parents. The endogeneity of childcare utilization will be modelled using the childcare supplement and the childcare supplement reform in 2012. Using Longitudinal Internet Studies for the Social Sciences (LISS), a fixed effects method for panel data is used. Based on this analysis, no parental health effects were found that could be related to the use of childcare for children ≤ 4 years. Using the survey data, the policy reform was found to have no impact on the childcare utilization.
2
1 Introduction
Childcare policies are crucial in helping parents reconcile care and can impact their employment decisions. According to the report of Kok et al (2015), 60 percent of Dutch children go to childcare and the average hours of childcare used is 19 per week1. Parents who work, who follow a training course/degree program, or who are looking for a job, can obtain a childcare supplement from the tax authority that covers part of their childcare costs. Childcare is an important cost factor for parents, making this childcare supplement quite important for them. In the study of Barber and Immervol (2006), it is stated that (even after deducting types of governmental support), the expenses for 2 pre-school children can add up to 20% of the household budgets. In 2012, the Dutch government decided to reform their policy concerning childcare supplement.
Until 2011, the Dutch government paid the majority of childcare, while the employers and parents each paid less than a third. In 2012, the Dutch government cut 1.1 billion euros in the childcare supplement. Prior to 2012, the childcare supplement was income related, while from 2012 onwards the supplement became linked to the number of hours worked of the least working parent. For some parents, childcare became unaffordable and they decided to use less formal childcare than before. Due to this reform, the number of children in daycare decreased, with a larger decrease for children aged 0-4 compared to those aged 4-12 (Roeters and Bucx (2018)).
While it is clear that the policy reform affects childcare utilization, there might be more consequences linked to this policy reform. Besides the financial impact this policy reform might have had, there could also be some health effects linked to this reform. It has been stated by several researchers that children at childcare are more likely to become ill compared to children who are not attending childcare (Magnus and Jaakkola (1999); Jarman and Kohlenberg (1991)). When children are in childcare, they are surrounded by many other children. Their immune system is being exposed to bacteria and viruses they have not yet seen before, which makes them more susceptible to getting sick (especially from highly infectious cold and flu viruses).
There exists a large literature gap about the effects of using childcare for young children (0-4 years) on the health of the parents. Although there is a considerable amount of research about the (health) effects of childcare on the health of the child, the parents are often not considered in this type of research. However, it is expected that there will also be an effect on the parents since they are most likely to stay at
3
home to take care of their child when he/she is sick. The parents are susceptible to viruses brought home by the sick child(ren) since these viruses change all the time and our immune system is not prepared for those(Neuzil et al (2002)).
This thesis contributes to the existing literature by doing research about the (indirect) parental health effects of childcare utilization. These health effects might be relevant consequences of childcare policy, which are often overlooked by others, such as researchers or policy makers. The goal of this thesis is to discover if there are parental health effects related to the use of childcare for children ≤ 4 years. In order to investigate the parental health effects, Dutch survey data, is used to provide a picture of the health effects of people with children ≤ 4 years.
The focus of this research is on parents with children between the ages 0 and 4. This age-category is chosen for three main reasons: (1) young children have a higher risk of getting sick from viruses/infectious compared to older children and are therefore more likely to transmit diseases (Fox et al (1975)); (2) the childcare utilization is higher for parents with young children compared to parents with older children (Ribar (1992)); (3) the effect of the policy reform on childcare utilization is the largest for young children. The main thought process throughout this thesis is as follows: the policy reform concerning the childcare supplement affects the childcare utilization, which will eventually influence the health of the parent. This thesis will answer the following questions:
1) Does the use of childcare for children ≤ 4 years impact the health of the parent? 2) Does the policy reform of 2012 impact childcare utilization?
In this thesis, both the effects of childcare utilization on a parent’s overall health and on their mental health will be analyzed. This is done using a fixed effect panel analysis and an instrumental variable (IV) analysis. The IV analysis is used to incorporate the effect of the policy change on the childcare utilization. Besides these methods, other empirical methods will be applied to get a better understanding of the mechanism and impact of the policy reform2.
The remainder of this thesis is outlined as follows. Section 2 includes the theory and previous findings about this research topic. Section 3 provides background information about the policy reform. Section 4 discusses the data, including the created variables and descriptive statistics. Section 5 presents the model that is used for this research, of which the results are being discussed in section 6. Section 7 provides a sensitivity analysis of the results found. The final section states the conclusions of this thesis.
4
2 Literature review
The literature section of this thesis is divided in 3 sections: “Endogeneity of childcare utilization”, “Children’s health effects of childcare” and “Transmission of infectious disease to parents”. These are the relevant topics that are key for this research. Per topic, relevant findings and studies will be discussed.
Endogeneity of childcare utilization:
In the publication of the Netherlands Institute for Social Research3, it is shown that, at the time childcare became more expensive for parents (in 20124), the number of children enrolled in childcare decreased. The number of children that were taken out of childcare went up during this time. Besides this evidence from the policy reform, there is sufficient empirical evidence to verify that the price of childcare drives the utilization of childcare. The most important findings are discussed below and are based on causal studies.
In the study of Baker et al (2008), the introduction of universal childcare in Quebec was studied. This type of childcare was implemented in the late 1990’s and was heavily subsidized, which made it more affordable for parents. To measure the effect of this policy, Baker et al (2008) used the change in Quebec (relative to other provinces) in 2000 or relative to 1997, while controlling for fixed effect. The proportion of children ≤ 4 years that went to childcare increased by 14 percentage point due to this reform. Each 10% increase in the subsidy rate raised the utilization of childcare by 4.6%. The subsidy crowded out informal childcare and made parents move from informal to subsidized childcare5.
This same Quebec policy reform has also been studied by Kottenlenberg and Lehrer (2014), who focussed their research on the effect of this reform for different ages of the child. A large negative intent to treat for newborn children was found, which is cut in half for children aged 1-3 years and even becomes insignificant from the age of 4. This finding shows the relevance to consider the age of the child in this type of studies.
The finding that the utilization of childcare is endogenous, is besides the study of the Quebec policy, also discussed in the study of Ribar (1992). This study analyzes the demand for market and nonmarket childcare services and the impact on the work effort of married women. Using survey data from the United states, it is stated that the utilization of paid childcare depends on the number of children/adults, the quality of care, the cost of care, hours of work, utilization of unpaid childcare, wages, family income and
3 Sociaal en Cultureel Planbureau (SCP) 4 The year of the childcare supplement reform
5
people’s preferences6. The number of adults that are present in a household, is found to have a negative effect on the utilization of paid childcare. This entails that one parent households are more likely to use paid childcare compare to larger households. This result is confirmed by Hotz and Kilburn (1994) who conclude that the childcare utilization is indeed higher for households with only a mother compared to households with two parents7.
Ribar (1992) found that an increase in the marginal cost of market childcare reduces the likelihood of using childcare. Having older or more children is also likely to decreases the use of childcare. Besides its effect on the utilization of childcare, changing the price of childcare also decreases employment. The employment of the mother affects both the market and nonmarket use of childcare. Increasing costs for childcare, both decrease the use of paid childcare for households with working mothers as well as the number of working mothers. In the study of Ribar (1995), survey data from the winter 1984-1985 is used to analyze two policy changes with respect to the Child and Dependent Care Tax Credit (CDCTC). The cost of paid childcare is found to only have a small negative effect on labor the supply of married women, but a strong negative impact on the utilization of paid childcare. Holtz and Kilburn (1994) conclude that the use of childcare is found to be more price sensitive for working mothers compared to non-working mothers. In places where there are high childcare tax credits, households with working mothers are more likely to use non-parental childcare services and also to use more hours of such care. An increase in the price of childcare will likely lead to less hours used of childcare for those still using childcare.
Van den Berg and Siflinger (2018) find that daycare attendance rates increase more in areas that experienced a substantial decrease in daycare fees, without any major rationing and queueing8. Holtz and Kilburn (1994) confirm that a higher price for childcare negatively influences the decision of parents to use childcare for both households with and without working mothers. They studied the effect of a childcare tax credit for working mothers and a subsidy for provider on the childcare utilization, work decision of households and on the price of childcare. Both subsidies increased the price of childcare and reduced the
6 This is confirmed by Robins and Spiegelman (1978), who state that the demand for childcare is affected by
demographic characteristics of the household, the price of care and the family income. Their result is based on a multinomial logit probability model using the childcare survey in Seattle and Denver Income Maintenance Experiments.
7 Their results are based on an examination of existing state-level childcare regulations on the cost, price and
demand of market childcare and the labor participation of the mother using U.S. survey data from 1986 of the high school class of 1972. Across-state differences in legislated minimum quality standards are being exploited in order to see how differences in the stringency of childcare regulation drive demand and the price of market childcare.
8 This finding is based on the exploitation of the maximum fee rule reform, which reduces the daycare fees for most
6
utilization of childcare. Holtz and Kilburn (1994) mention that an increase in the price per unit of non-parental childcare will decrease the likelihood that parents will use non-non-parental childcare (ceteris paribus). Based on the evidence discussed, childcare utilization probably depends on other characteristics.
Hypothesis 1: Childcare utilization is endogenous.
Children’s health effects of childcare:
There is ample research about the effect of childcare utilization on the health of the child (Jarman and Kohlenberg (1991), Enserink et al (2014), Van den Berg and Siflinger (2018)). Childcare utilization can have both positive (Felfe et al (2014)) and negative (Nafstad et al (1999)) effects for the child’s health, but for different health aspects. First, the positive effects that are linked to childcare will be addressed, after which the health effects (both mental and physical) will be discussed.
Attending childcare might offer the opportunity for parents and children to be educated about infectious disease risk and nutrition (Berlinski et al (2008)). When childcare is used for children at a young age, the risk of infections at a later stage in life might be reduced. Conti et al (2016) looked at two early childhood interventions and concluded that boys who were enrolled in the Perry Preschool Project (PPP)9 had a lower prevalence of behavioral risk factors when they became adults, compared to those in the control group. Those boys that were in the Carolina Abecedarian Project (ABC)10 were found to have a better physical health later on in life, compared to the control group11. The conclusion that childcare has beneficial effects on a person’s health later in life, is also stated in the research of Nores and Barnett (2010) who review several studies12 concerning a total of 30 early childhood interventions in different (non-U.S.) countries. Based on this comparison, it is stated that children have cognitive, behavioral, health and educational benefits from these childcare interventions. Van den Berg and Siflinger (2018) support this by stating that that children at daycare had less behavioral and social disorders. The finding that these childcare effects might be persistent over time, is confirmed by Felfe et al (2014), who mention that the introduction of subsidized full-time high-quality childcare for 3-year old’s improved children’s reading skills
9 Program provided preschool education at ages 3-5 and home-based parenting guidance. This deals more with
behavior.
10 Program includes similar things as PPP, but also included a healthcare and nutritional component for children
aged 0 – 8. This program also includes a survey for health at age 34.
11 Both the PPP and the ABC were center-based small-scale programs and targeted disadvantaged populations. 12 These studies apply quasi-experimental or random assignment, which makes these results reliable. Nores and
7
at age 1513. This finding reveals that besides the social and health effect later on in life, childcare can also have educational effects.
While it is worth noting the positive effects of childcare, it is important to focus on the negative health effects of the child, as this will likely impact the health of the parent14. Children in center-based childcare tend to use more health care services than those not using childcare, which implies higher healthcare costs (Silverstein et al (2003)15). Besides the effects on children’s physical health, which are discussed below, going to childcare can also impact a child’s mental health. The cortisol levels of children increase when a child goes to daycare, while it decreases when a child stays home (Geoffrey et al (2006)16). This effect was stated to be larger for preschoolers compared to infants or school-aged children. Van den Berg and Siflinger (2018) also studied the effect of going to childcare on the mental and physical health of children. They stated that children at daycare had a better mental health and non-cognitive abilities in general compared to other children and that there is no gender difference in these effects.
Besides the mental health effects, childcare utilization is likely to also affect the physical health of the child. Researchers focusing on epidemiology often base their result on correlation instead of causation, which makes the results less reliable (Fox et al (1975); Jarman and Kohlenberg (1991); ); Byington et al (2015); Nafstad et al (1999)). Fox et al (1975) conclude that children under the age of 6 had more frequent rhinovirus (RV)-associated illnesses compared to older children. These RV illnesses were stated to be most severe for children younger than 2 years. Byington et al (2015) investigate the transmission and duration of virus detection17 and found that children younger than 5 years have more often symptoms, compared to older children. However, these health effects might occur at earlier ages. For instance, Jarman and Kohlenberg (1991)18 conclude that children enrolled in group daycare facilities have a higher frequency of respiratory infections, in their first 3 years, and are more prone to diarrhea and other conditions transmitted by the facial-oral route. Nafstad et al (1999) studied the health of children in daycare centers
13 Based on a study of the expansion of publicly subsided full-time high-quality childcare for 3-year-olds in Spain in
1990. A difference-in-difference approach is used to get a causal relationship.
14 Because of the assumption made in this thesis that childcare has negative health effects for the child, which will
be transmitted to the parents.
15 Based their study on sample data of a sample of children aged 0 – 5 in the U.S. Cross-sectional analysis is used
together with a weighted multivariate model.
16 Geoffrey et al (2006) summarized and compared studies about childcare experiences and cortisol levels from
1985 – 2006. They pooled the results of similar studies together.
17 The vital etiology of respiratory illness. This is done by studying nasal swabs from a substantial amount of people
in Utah.
8
and state that children in day care centers also have and increased risk of common cold and otitis media19. They further recognized that attendance to daycare centers increases the risk of upper respiratory symptoms and infections in 3 to 5 year old children.
While the previous results about the physical health of the child were based on correlations, the results of Kottenlenberg and Lehrer (2014)20 are based on causation. Kottenlenberg and Lehrer (2014) confirm the findings mentioned before by stating that subsidized childcare has significantly large negative impacts on the health of children who start childcare at young ages. They conclude that the age of the child has a significant influence on the impact of childcare on development outcomes. The younger the child is enrolled in day care, the more likely the child will have behavioral problems. Nafstad et al (1999) state that an early stating age in daycare increased the risk of developing recurrent otitis media. A similar finding was found by van den Berg and Siflinger (2018), who state that the effect of childcare starts to become positive for children at the age of 4 for low SES families. Children aged 2-3 have a higher probability of experiencing upper respiratory infections and have an increased likelihood of transmissible diseases and conditions. The effect of day care on a child’s health also depends on the SES of the family. Children in high-SES families, showed more beneficial effects when they entered childcare at 4-5 compared to when they enter childcare at age 2-3. If children started daycare in their first 2 years of life, the lifetime risk of asthma is also higher (Nafstad et al (1999)). This can be mediated via early respiratory infections that are more common in daycare centers compared to home care.
In case of most respiratory infections, keeping the child at home for 1 or 2 days does not reduce the risk to other children of getting sick (Jarman and Kohlenberg (1991)). If children in daycare are divided into small groups of 3 children, there is likely to be no increase in the risk of infections compared to children who stay at home. Côté et al (2010) use data from Quebec for 1998-2006 and conclude that children in early preschool had higher rates of respiratory tract infections and ear infections if they used large group-childcare facilities, compared to those children who stayed at home. This is based on a causal study and is true for the period the child goes to daycare. This implies that, in order to look at the health effects, it is important to focus on group childcare facilities21.
19 These findings are based on the relation between the occurrence of health outcomes and day care center
attendance using survey data of Norwegian children in Oslo in 1996.
20 A linear difference-in-difference design is used, which is also used by Baker et al (2008))
21 As there is no difference in the health effects for childcare options with small groups and letting the child stay at
9
Besides the risk of infections for children enrolled at childcare, these children might also transmit these infectious diseases on to their parents. Nafstad et al (1999) concluded that children aged 3- to 5-years are found to have a higher risk of upper respiratory symptoms and infections. As these are highly contagious sicknesses, it is likely that the parents of these children also get sick. Jarman and Kohlenberg (1991) state that children are contagious for several days during the first phase of their illness, which also puts their parents at risk. This is true for most of the infectious diseases. As mentioned before, these results22 are based on correlation rather than causation. Based on the literature discussed in this section, it can be assumed that children in childcare are more often sick compared to those not going to childcare. The transmission of diseases will be discussed in more detail in the next section, where some of the key findings are mentioned.
Transmission of infectious disease to parents:
Besides the evidence about the negative health effects of children in childcare, it is important to investigate how this influences the health of the parents. The parents are likely to be the ones that take care of the child and are therefore more likely to be exposed to the germs of the child. Below the most important findings are mentioned. It should be noted that these findings are based on correlations rather than causation, as this evidence is mostly from epidemiological studies.
Monto (2004) states that the continuing reinfection that occurs with common respiratory viruses, provides the route by which very young children become infected23. It is stated that children experience the highest frequency of respiratory illnesses and that major transmission of respiratory viruses occur in families and in pre-schools. The statement that the age of the child is relevant is confirmed by the finding that parents with young children experience a higher frequency of illness compared to parents with older children (Monto(2004)). The study of Neuzil et al (2002)focusses on the health of children in elementary school, but also pays attention to secondary illnesses among family members24. When the child gets sick, it is likely that people surrounding that child also get sick. Of all the people that became sick within 3 days after the child got sick, 46% were the parents. Fox et al (1975) found that when there were children in the family, adult illnesses were indeed found to be more frequent. By studying the correlation between health of parents and having children, Byington et al (2015) state that people living with children experienced 3 additional weeks of virus detection compared to single persons’ households. Children were found to be an
22 Nafstad et al (1999) & Jarman and Kohlenberg (1991) 23 This is based on the discussion of respiratory viruses.
24 They study the effect of the 2000 – 2001 influenza season on a group of children enrolled in a large elementary
10
important introducer of infection into families and in mediating community spread (Fox et al (1975); Monto (1994)).
Within the families, it is often the mother that is most affected by respiratory illness (Fox et al (1975); Monto (2004); Monto (1994)). This effect might imply that mothers have a closer contact with the child and are therefore more likely to get sick. Monto (1994)25 states that this health effect is heterogeneous among women. There is a difference in the illness records between women working outside their home and those staying at home. Women working outside their home were less often sick compared to women staying at home but were more often sick compared to men. However, this effect is based on correlation and not causation, which makes it less reliable.
Factors that impacts the frequency of illness are the size of the family, the income of the family and the age of the household/children (Monto (1994); Monto (2004); Petrie et al (2013)). Individuals in households with more than 4 members and individuals in households with children aged below 5 years had significantly higher frequencies of acute respiratory illnesses compared to others (Monto et al (2014)26, Monte (2004)). However, the frequency of reported illness (and the infection rate) declined with family income (Monte(2004)). Petrie et al (2013) found evidence that adults had symptomatic illness for a longer duration compared to children. Younger households and households with children younger than 9 years were more likely to have influenza compared to other households. The risk of being infected was also higher for young children than for adults27.
Combining the empirical evidence of “Children’s health effects of childcare” with the evidence mentioned above, it is assumed that people with children at childcare are more often sick compared to those not using childcare.
Hypothesis 2: The health of parents with children (≤4 years) using childcare is worse compared to those not using childcare
25 Study of several studied studies about respiratory illnesses and families.
26 Based on a study of households with children above 3 years of which they collected respiratory tract specimens
from people with acute respiratory illnesses.
27 Based on the study of 328 households with children during the 2010-2011 influenza season in Michigan.
11
3 Background information
In 2012, the government introduced a reform which changed the way in which the childcare supplement is offered by the tax authority. This reform impacts parents with children who attend childcare. Rather than the childcare supplement to be solely linked to the overall income of the household, from 2012 onwards it is also linked to the hours worked of the least working parent. For childcare, parents are entitled to a supplement of 140% of the working-hours of the least working partner. For after-school childcare, the supplement is reduced to 70% of the hours worked of the least working partner28. The maximum number of hours for which the parents can receive childcare supplement is set to be 230 hours per child per month (Algemene Rekenkamer (2014)). For the financial consequences of this policy change, it is important to know what determines the amount of childcare supplement people receive. The difference between before and after 2012 is listed in table 1.
Table 1: Difference childcare supplement before and after 2012
Before 2012 After 2012
• the number of hours for which childcare was used per child per month
• the number of hours for which childcare was used per child per month
• the hourly price of childcare • the hourly price of childcare • the overall income of the family • the overall income of the family • the type of childcare • the type of childcare
• hours worked of the least working parent
For families with more than one child, one should take into consideration that there is a difference between the supplement they can obtain from each of these children. The child for which most childcare is used, is reported to be child 1 by the tax authorities. The childcare supplement percentage is lower for child 1 compared to the other child(ren). These childcare supplement percentages are linked to the joint income of both parents. A different supplement percentage is attained to different income categories29. These percentages are changed every year slightly. In 2012, the parental contribution for the second and subsequent child is increased, and the maximum hourly rates are not indexed.
28 This percentage is for school-going children (4-12 years)
12
Due to the budget cuts in the childcare supplement, parents will pay more for the childcare of their child(ren). For example, parents with minimum wage and 2 children who use childcare for two dayparts per week pay 37 euro’s more per month due to this budget cut. Parents with a high income pay 290 euro’s more. The policy reform has the most impact for those people of which their child uses more than 140% of the number of hours the least working parent works.
Based on the LISS panel, figure 1 is constructed to model the trend in the tax supplement and the childcare utilization over the years30. This figure present graphical evidence that there was a substantial drop in the tax supplement around 2012.
Figure 1: Childcare utilization and childcare supplement per month (over time)31
30 A similar graph over an extended time period (2008-2018) is shown in figure C.1 of appendix C
31 The use of childcare is modelled by a summation (per year) of the amount of people using a paid-childcare option
13
Based on the figure 1, the childcare utilization appears to be stable over time. However, national aggregated data shows that there in fact was a decrease in childcare utilization after 2012. While an increase in the use of childcare was expected for 2012 and 2013, there was a large drop measured. Possible reasons for this drop in childcare utilization include: an increased unemployment due to the slow economic growth, a lower birthrate, and the stricter policy of the tax authority (Algemene Rekenkamer (2014)). This drop around 2012 is depicted in figure 2 from Statistics Netherlands (in Dutch: CBS). The aggregated data concerns a wider population, which makes it is more trustworthy compared to the small subsample in the LISS data, which showed no drop in childcare utilization (see figure 1).
Besides the size of the sample, another possible explanation for the lack of a fall in the LISS panel around 2012 might be the representativeness of the subsample used in this thesis. This specific subsample might be less representative compared to the general Dutch population due to the fact that people need time to fill in these questionnaires. It might be that those who filled in these questions would use childcare independent of the price of childcare. In a critical study of Knoef and de Vos (2009), some issues with the representativeness of the LISS panel are discussed. A relevant issue for this thesis is the finding that single person households are underrepresented in the LISS panel (Knoef and de Vos). Couples younger than 65 without children and couples with 2 children are also found to be overrepresented. It is further mentioned that although people aged 16-20 are overrepresented in the panel, their responses to the “Health” survey are low. Females are found to be more responsive in the questionnaires compared to males. All these findings might have impacted the findings, making figure 1 less comparable to the figure 2.
14
For the remainder of this thesis, it is important to get a clear understanding of the financial consequences of this policy change. An example will be used to illustrate the financial impact of the changed policy. This is done by comparing the difference in the childcare supplement received between the period before 2012 and the period after 2012. First, the childcare supplement will be calculated for 2011 (before the policy change), after which the childcare supplement received will be calculated for 2013 (after the policy change).
Example of the childcare supplement:
In a specific household, both of the parents work and together they have one child (3 years old). The mother works 30 hours per week and earns a wage of €30.000 per year, the father works 40 hours per week and earns a wage of €40.000, which means that their joint income is €70.000. On average, the child spends 100 hours a month in a daycare center, which has an hourly price of €6. As there is only one child in this family, this child is indicated as child 1 by the tax authorities32.
First the amount of childcare supplement before the reform will be discussed. As the hourly price lies below the maximum amount for daycare (€6,36)33, these parents will receive the supplement over the entire amount. As the overall income lies between 69.440 and 72.216, the childcare supplement of the government and the employer-contribution for this child is 52,33%34. The amount of childcare supplement this household will receive per month is calculated by multiplying the total childcare costs (number of hours the child uses childcare time the hourly price35) by the childcare supplement percentage. This family is entitled to receive a childcare supplement of €313,98 per month36.
Next, the calculations of the childcare supplement in 2013 will be described. As the child goes to daycare, the number of hours worked of the least working parent are first multiplied by 140%. This means that these people are entitled to childcare supplement for 182 hours per month per child37. The costs of childcare for this child are calculated by multiplying the number of hours for which the child uses childcare
32 This will be used to determine the childcare supplement percentage (as this percentage is different for the first
child compared to child 2 (and subsequent children)).
33 The maximum hourly price is €6,36 for daycare, €5,93 for after-school childcare and €5,09 for a host parent
(https://wetten.overheid.nl/BWBR0017321/2011-01-01).
34 Since there was no data about the percentage for 2011, the percentages of 2012 are used in this study. https://www.kiekeboeputten.nl/docs/D-2011-02%20percentagetabel%20kinderopvangtoeslag%202012.pdf
35 In case the price is below the maximum hourly price, otherwise the price used will be the maximum price. 36 100 * €6 * 0,5233
15
by the hourly price38. This means that the costs for daycare in this case are €600 per month39. The childcare supplement percentage with whom this amount needs to be multiplied, changes a bit every year. In 2013, the childcare supplement percentage for an overall income of €70.000 was 45,4%40. The amount of childcare this household will receive is €272,40 per month41. This is a difference of €41,58 compared to the childcare supplement received in 2011.
As mentioned before, when the household consists of more than 1 child, the childcare supplement percentage differs between the 2 children. To see the effects for a family of 4, an additional example can be found is appendix B. However, for both the family with one child mentioned above and the family of 4, the difference between the amount of childcare supplement received before the reform and after the reform is only due to the lower childcare supplement percentage and not due to the link of the childcare supplement to the hours worked. If the hours of childcare used are increased and/or the hours worked of the least working partner decreased, the difference increased. This is illustrated in the second example in appendix B. The increased difference between the childcare supplement in 2011 and in 2013 is mainly due to the link of the supplement to the number of hours of the least working parent. This is due to the fact that the number of hours of childcare used exceeds the number of hours for which the family is entitled to a childcare supplement. Due to the reform, some hours of childcare used are not (partly) covered by the tax authority, while they would be covered by the old policy.
4 Data
For this research, longitudinal panel data is drawn from survey data of the Longitudinal Internet Studies for the Social Sciences (LISS) panel administrated by CentERdata (Tilburg University, The Netherlands). The LISS panel is a representative sample of Dutch individuals who participate in monthly Internet surveys and consists of approximately 7000 individuals. The panel is based on a true probability sample of households drawn from the population register. Because the recruitment was through address-based sampling, there is no self-selection bias. A longitudinal survey is filled in by the panel every year and covers a large number of domains such as education, income, health, and household composition. Using panel data, the unobserved individual heterogeneity in (unobserved) health can be taken into account.
38 If the hourly price of childcare is below the maximum price, the hourly price is used. When the hourly price of
childcare is more expensive compared to the maximum hourly price, the maximum price is used.
39 100 * €6
40
https://www.kiekeboeputten.nl/docs/D-2012-02%20percentagetabel%20van%20de%20kinderopvangtoeslag%202013.pdf
16
The LISS data has several distinctive surveys, of which the following are used for this research: “Family & Household”, “Health”, “Work & Schooling” and “Background variables”. From these questionnaires, the years 2009 until 2015 will be used42. The “Family and Household” and “Work & Schooling” survey were administrated in all the 7 waves of the LISS panel, whereas the “Health” survey was recorded in 6 waves (no data for 2014). While these questionnaires are recorded yearly, the questionnaire “Background variables” is recorded on a monthly basis. In order to merge this specific questionnaire with the other relevant questionnaires, the monthly files of the years 2009 – 2015 are appended into one file43. All the data from the different questionnaires and years are combined on the basis of a unique personal identifier. The total sample consists of 113,163 individual-year observations44. However, this sample not well balanced given the fact that there are no records of all respondents for all years. This dataset allows to identify the people using childcare, the amount of childcare supplement received and the health of the respondent.
While the LISS data has some weaknesses45, other data sources have more limitations, making them less applicable for this specific research. For instance, one could opt for CBS register data which consists of information about a large set of characteristics. However, due to the fact that the health records are mainly focussed on extreme cases, there is limited information about the (personal) health of people. The LISS data has personal data about the health of people and does not solely include extreme cases, which makes LISS data more suitable for this research. The Survey of Health, Ageing and Retirement in Europe (SHARE) is not used for this thesis as this survey focusses on individuals aged 50 or older while the maximum age of people who use childcare is also around 50. Because of this, the sample of the SHARE panel does not match the age category of the subsample used in this thesis. Besides this, the SHR measure is not comparable across countries as the reporting style of this measurement might differ. Other data sources, such as Lifelines, could not be used either, due to financial constraints.
42 By using these years, the policy change of 2012 is exactly in the middle of the study period. This implies that the
effect of the changing childcare supplement on the childcare utilization can be analyzed accurately.
43 This is done by merging the monthly data per year and drop the month indicator in the source-date identifier. As
the data per person is often similar for all the months in one specific years, only unique observations were recorded.
44 Observations per year: 2009 = 17632; 2010 = 17321; 2011 = 16004; 2012 = 15698; 2013 = 15214; 2014 = 15865;
2015 = 15429.
17
Measures:
Self-reported overall-health:
Data on the health of people were derived from the “Health” survey. From this survey, the following question will be used to measure the self-reported health (SRH) state of the respondent: “How would you describe your health generally speaking (between 1 and 5)?”. The responses will be used to construct a categorical variable with 5 health states: poor (1), moderate (2), good (3), very good (4) and excellent (5). However, due to possible inconsistencies in the individual response pattern of their health status over time, this variable needs to be corrected. This correction is done using the question “Can you indicate whether your health is poorer or better, compared to last year (from 1 to 5)46?” and follows the procedure of van Ooijen et al (2015). The responses to this question will be referred to as the self-reported change in health (SRCH) and are transferred into either “poorer”, “the same” or “better”47. First, the health in the first period (2009) is considered to portray a persons’ true health state. Next, the SRH of 2 consecutive years will be compared to assess whether the health of this person improved, stayed the same, or degenerated. This change will be compared to the reported SRCH. When comparing both, there are records of people filling in a higher/lower value of health compared to the previous year, while their SRHC shows no change. For these cases, the level of health is adjusted in the subsequent year, by reducing/increasing the level of the SHR by one unit (van Ooijen et al (2015)). If the SRHC shows a change, while a change is observed when comparing 2 SRHs, the health state is not corrected. The corrected health level for subsequent years will be constructed using the reported SRH and SRCH, rather than using the corrected health of the previous year48.
While the SRH measure is proven to be useful to measure health, this measure has some shortcomings. These shortcomings include measurement error and biased results (van Ooijen et al (2015)). A person’s mood or health perception might affect the answer people give for this SRH question (Lindeboom and van Doorslaer (2004)). Due to the fact that this question is subjective, the judgements might not be comparable across respondents. The measurement error in SRH will likely reduce the estimated persistence in health. By means of correcting the SRH measure, some of the inconsistencies are reduced.
46 1 = considerably poorer; 2 = somewhat poorer; 3 = the same; 4 = somewhat better; 5 = considerably better 47 Responses of 1 and 2 are combined into one value indicating “poorer compared to previous year”; responses of 3
indicate “the same health state compared to previous year”; responses of 4 and 5 are combined into one value indicating “better compared to previous year”.
18
Mental health:
Besides considering the SRH, a measurement for the mental health of the person is used to differentiate between the effects on a person’s overall health (SRH) and their mental health. This measurement needs to be constructed, since the “Health” questionnaire does not include a specific mental health question. To measure the mental health state of the respondents, the following statements will be used: “I felt very anxious”; “I felt down that nothing could cheer me up”; “I felt depressed and gloomy”; “I felt calm and peaceful” and “I felt happy”. People need to indicate the occurrence of these statements in the past month from 1 (never felt this) to 6 (continuously felt this). These statements seem to be concerned with a person’s mental wellbeing and might therefore be related to the same factor. This factor is an unobserved or latent variable which explains the correlation between the statements. A higher value for the SHR question indicates a better health, while this cannot be stated directly about these mental health statements. For some of these statements, a higher value indicates a better mental health49, while for the other statements a higher value indicates a worse mental health50. To generate a consistent variable for which a higher value indicates a better mental health state, the “negative” statements need to be rewritten and recoded. The manner in which these statements are rewritten is stated in table C.1 of appendix C. After rewriting the statements, their scores are reversed in such a way that a high value of each of these statements coincides with a better mental health51. A respondent who filled in a low value in the original statements, would now have a high value since the statements switched side (from negatively to positively related to mental health)52.
After rewriting and recoding the statements, their answers are combined into one variable indicating a person’s mental health. Taking the average of the 5 answers of these statements might not measure the respondents’ mental health correctly because one statement might be more prominent in determining a person’s mental health compared to the other statements. To get a more precise measurement for the mental health, a factor analysis is performed for the statements: “I felt very anxious”; “I felt down that nothing could cheer me up”; “I felt calm and peaceful”; “I felt depressed and gloomy”; “I felt happy”. A factor analysis is a statistical method to describe the variability among observed and
49 This is related to the following statements that are positively related to mental health: “I felt calm and peaceful”
and “I felt happy”.
50 This is related to the following statements that are negatively related to mental health: “I felt very anxious”, “I
felt down that nothing could cheer me up” and “I felt depressed and gloomy”. These will be referred to as the
“negative” statements.
51 Originally a higher value of these “negative” statements would indicate a worse health. 52 6 becomes 1; 5 becomes 2; 4 becomes 3; 3 becomes 4; 2 becomes 5 and 1 becomes 6.
19
correlated statements in the dataset. A more detailed description of the factor analysis can be found in appendix A. Because “mental health” will be the dependent variable on which the model will be based, one should be careful in selecting the relevant statements. The statements are formulated quite clearly which decreases the chance of different interpretation among respondents. However, the answers remain subjective, as people interpret “mostly” different, or people are generally more positively or negatively minded, which impacts their responses. The answers of the selected statements are weighted according to their factor loadings from the factor analysis to create a single factor. It uses the appropriate weights in order to distinct important statements from less important ones. Based on the factor analysis, the statements “I felt calm and peaceful” and “I felt happy” were found to have a lower factor loading compared to the other 3 statement for the first factor, while having a higher factor loading for a second factor53. This might be explained by the fact that these 2 statements are more positive in nature compared to manner in which the other 3 were originally formulated. These 2 statements are likely to measure the positive attitude of people, rather than their mental health. Since this could still be considered as part of a person’s mental health state, all 5 statements will be used to measure the mental health state of the respondent in the main analysis of this thesis. In the sensitivity analysis, only the “negative” statements will be used to model mental health.
Childcare utilization:
The questionnaire “Family & Household” includes those questions concerning childcare utilization and the childcare supplement. A crucial factor of this research is the use of childcare for parents with a child ≤ 4 years. In this questionnaire, it has been asked (to people who have a living-at-home child that is ≤ 4 years) whether they make regular use (at least once a week) of the following childcare options: toddler playgroup, nursery school; child daycare center (‘kindergarten’- half-day childcare); pre-school childcare; after-school childcare; host parent where the child goes to (arranged through host parent agency); paid child-sitter, where the child goes to; paid child-sitter, that comes to your home; unpaid child-sitter; other childcare; no childcare. For each of these options, the respondents had to indicate whether or not they use it for their child. For the reason that the childcare supplement from the tax authority is linked to paid childcare options, only these options are pooled to measure the use of childcare54. The dummy variable “childcare” is created, for which a 1 indicates that this person uses at least one type of paid childcare for their child ≤ 4 years. A value of 0 indicates that this person has at least a child ≤ 4 years but does not use
53 While still having a factor loading above the 0.4 threshold used by researchers for both
54 This includes: toddler playgroup, nursery school; child daycare center (‘kindergarten’- half-day childcare);
20
childcare for this child (these children). Parents that only use free childcare (unpaid child-sitter) are not included in the subsample used for this analysis. This questionnaire is limited in how childcares utilization can be measured, due to the fact that the childcare utilization questions are not specified per child but are asked per age-category (0-4, 5-13 and 14-18 year).
Childcare supplement:
Besides the use of childcare, the price of childcare is also important as explained in the literature review. Because there is (1) no exact data in the LISS panel about the price of childcare used per child; (2) the price of childcare can differ per facility, the true price cannot be used. The LISS survey does ask about how much the person pays (in total) for childcare of the children, but this amount is not specified per child. Therefore, instead of using the exact price of childcare, the childcare supplement of the tax authority will be discussed. This childcare supplement indirectly influences the price people pay for childcare, which makes it useful for this research.
Control data:
Besides the two main topics (health state and childcare), several control variables are included in the model which are based on the questionnaires “Work” and “Background variables”. The control variables used for this study include: income (based on net income), working hours, education (based on the highest education level for which they received a diploma/certificate), age, living situation (including living with a partner), number of people in the household, number of children in the household, degree of urbanization where the person lives (from 1 to 5: 1 indicating extremely urban (2500 or ore addresses per km2); 5 indicating not urban (less than 500 addresses per km2)), primary occupation (employed; works/ assists in family business; self-employed; job seeker; student; takes care of housekeeping; retired; unemployment benefit; volunteer; something else). Most of these control variables are also used by other researchers (Baker et al (2008), Kottelenberg and Lehrer (2014)55, Felfe et al (2014)56, Ribar (1992; 1995), Hotz and Killburn (1994), van den Berg and Siflinger (2018)), which justifies their use. There is controlled for the number of children in the family, such that large family size will not confound the estimates found (Baker et al (2008)).
55 Kottelenberg and Lehrer (2014) use the same control variables as Baker et al (2008)
56 The state-specific control variables used by Felfe et al (2014) (educational level, GDP, employment rate) are
21
Descriptive statistics:
To get an insight in the relationship between having a child at day care and the health of the parents, several descriptive statistics will be shown. As this study incorporates the effect of the childcare supplement of the tax authority on the use of childcare, it is noteworthy that most of the people using childcare do receive the supplement57, as can be seen in figure 3 and figure C.2 in appendix C. This indicates that the majority of the parents will be affected in case there is an effect of the supplement on the use of childcare. The average number of hours the parents use childcare for their child ≤ 4 years is 57,5 hours per month with a minimum of 4,3 hours and a maximum of 398,7 hours. The distribution of the hours of childcare used by the respondents is shown in figure C.3 in appendix C.
The health effect within the household might be different for the father compared to the mother. This might depend on the degree in which they are responsible for the health-related tasks with the child58. Being responsible for these tasks implies being in closer contact with the child59 and therefore having a higher risk of receiving infectious diseases from their child. This parent might also experience more stress from taking care of the child. Based on the sample of this study, it can be stated and seen in figure 4, that the woman is more often responsible for these specific tasks compared to men. This is even more apparent when looking solely at the task “staying home with the child when unwell”, which can be seen in figure C.4 of appendix C.
57 Around 68% of the people using paid childcare for children ≤ 4 years receive a childcare supplement. 5% of these
people stated that they applied for the supplement but that they did not receive it yet (these 2 groups are combined in the graph) and 27% of these people do not receive a childcare supplement.
58 This is based on the following tasks: “changing diapers”; “getting up from bed during the night”; “washing the
child”; “taking the child to the doctor when unwell”; “staying at home with the child when unwell”.
22
Figure 3: Pie chart use childcare (child ≤ 4 years) Figure 4: Division of care-related tasks
As mentioned before, the SRH of the respondents is being corrected to corrected for inconsistencies over the years. In table 2, the percentage of the people per health state are listed, both for uncorrected as well as the corrected health measure. Most people are found to have a good health status (60%). As shown in figure C.5 of appendix C, there are almost no differences between the health distribution of those using childcare compared to those not using childcare However, this is the overall health effect, in which there is not differentiated for characteristics, which will be done in the analysis part of this thesis.
Table 2: % respondents per health state:
Uncorrected (entire sample)
Uncorrected (those having at least one
child ≤ 4 years)
Corrected (those having at least
one child ≤ 4 years)
Poor 1,36 % 0,45 % 0,55%
Moderate 14,48 % 8,78 % 9,33%
Good 59,45 % 61,40 % 60,10%
Very good 18,74 % 23,54% 24,74%
excellent 4,97 % 5,84% 5,29%
23
in table 3. When comparing table 3 with the descriptive statistics of the entire sample in table C.2 of appendix C, the mean age, and the number of children of the treatment and control group is higher compared to the overall sample. This is due to the fact that both the control and treatment group are parents, while the entire sample also contains non-parents.
Table 3: Descriptive statistics of people using childcare (and those not using childcare)
Variable Number of
observations:
Average Min Max Age (years) • No childcare • Use childcare 495 2495 34,95 35,27 20 19 66 67 Gender • No childcare • Use childcare 495 2495 0,53 0,56 1 1 2 2 Number of people in the household
• No childcare • Use childcare 495 2495 3,87 3,69 1 1 8 10 Number of children • No childcare • Use childcare 495 2495 2,26 1,99 1 1 6 9 Having a partner • No childcare • Use childcare 495 2495 0,92 0,94 0 0 1 1 Net-income • No childcare • Use childcare 460 2347 1121,72 1612,48 0 0 4500 30000 Educational level • No childcare • Use childcare 493 2491 3,77 4,28 1 1 6 6 Having a paid job
• No childcare • Use childcare 422 2189 0,62 0,89 0 0 1 1 Health • No childcare • Use childcare 314 1691 3,23 3,26 1 1 5 5
24
do not use this type of childcare is significant. This is done by performing two-sample t-tests60 for the variables mentioned in table 3. It can be concluded that the means of “age”, “gender”, “having a partner” and “health” are not statistically different between the two groups. The mean of the “number of people in the household” and the mean of the “number of children in the household” are significantly larger for parents who use no childcare compared to those using no childcare (for their child ≤ 4 years). The mean of the “monthly net income”, the mean “educational level of the respondent” and the mean of “having a paid job” were found to be significantly smaller for the group of parents who use no childcare (for their child(ren) ≤ 4 years) compared to using this type of childcare. These conclusions are confirmed using Welch’s formula which assumes unequal variances.
This finding shows that, while the means of the relevant variables look similar, they might be statistically different. Although there are some (statistical) differences, the treatment group is still relatively similar to the control group. Since these 2 groups are relatively comparable, any result that may be found after the analysis cannot be driven by any of these characteristics.
The hours use of childcare are graphically shown in figure C.6 in appendix C. When comparing the characteristics of women who use childcare more than 30 hours of childcare per week (indicating those using much childcare) with those using less than 10 hours of childcare (indicating those using little childcare), several small changes can be noted. The average income is slightly lower for those using more childcare (> 30 hours) compared to those using little childcare (1230 vs 1345 respectively)61. This this could be explained as the average education level is slightly higher for those using little childcare (3,90 vs 3,12)62. The average age of those using more childcare is higher compared to the average age of those using less childcare (>30 hours) (39,42 vs 34,75 respectively)63. The amount of people in a household is also slightly lower if the respondent uses more childcare (> 30 hours) (3,67 vs 3,09 respectively) 64. A more striking difference is the % people having a partner. While this is percentage 92 when using less childcare, this is only 79 when the respondent uses more childcare (> 30 hours) 65. However, this is not surprising since respondents with a partner have an
60 This test assumes equal variances. The null-hypothesis in this type of t-test states that the difference in the
variable between those using no childcare and those using childcare equals zero, meaning there is no statistical difference between the 2 groups.
25
extra person that can take care of the child, which means that less market-provided childcare is required.
Before analyzing the data, it is important to inspect and clean the data. Inconsistent values of several questions are being excluded from the analysis66. When using the number of children in the household from the “Background” questionnaire, there is an odd finding. There are respondents who stated to have no living-at-home children in the “background variable” questionnaire, while they filled out to use childcare for their child in the “family” questionnaire. This was solved using the count of the number of children for which age was filled in67. Besides this inconsistent finding, the data also contains some outliers for income. However, these “outliers” are found to be consistent over time, which means that this systematical measurement error cancels out when using fixed effects analysis . By retaining these “outliers” in the analysis, the subsample used will be larger. Whether or not keeping these “outliers” impacted the results found, will be tested in the “sensitivity” part of this research.
5 Analysis
A multivariate analysis is used to model the effect of all relevant characteristics on the health of the parent68. This study analyzes the parental health effect of a subset of the LISS panel respondents, namely, people who have at least one child below the age of 5 year in the timefram e 2009-2015. A total of 2990 observations over 7 years met this criterion. The health state of parents with children ≤ 4 years who use childcare for this/these child(ren) will be compared to the health state of parents with children in the same age category who do not use childcare for these children. The analyzes were conducted using Stata 15.1.
66 Inconsistent observations are those for which people filled in 9999 or 99999 (meaning “I don’t know”) for some
of the questions. It also includes those observations for which people filled in -13 (meaning “I don’t know”); -14 (meaning “prefer not to say”) or -15 (meaning “unknown (missing)”) for their net income. These observations were changes into missing values. This is also been done for the answers “not completed any education” and “not yet started education” for the variable “education”.
67 If a person only filled in the age of child 1, this will be recorded as having 1 child. Filling in the age of child
1 and 2 will be recorded as having 2 children, etcetera
26
Throughout this analysis, a fixed effects approach will be applied69. This approach is chosen to control for the different effect the policy reform has for different people70. Individual characteristics are likely to impact childcare utilization. The fixed-effects approach is used to control for possible endogeneity. Fixed-effects estimation is used to deal with endogeneity when respondent-specific (time-invariant) characteristics are correlated with the explanatory variable (Wintoki et al, 2012). To the extent that the endogeneity is caused by time-invariant respondent endogeneity, the fixed effects estimation will correct for endogeneity. An advantage of the fixed effects approach is that is allows the individual specific effects to be correlated with explanatory variables. The fixed effects approach further controls for other factors that may influence the health of the person, such that the net effect of childcare utilization on the SRH state can be assessed Shortcoming to the use of the fixed effects approach is that it uses a “strict exogeneity” assumption and that it removes the effect of time-invariant characteristics (such as gender and ethnicity). These time-invariant variables provide no within-person variance to compare. While the use of fixed effects is not supported by the Hausman test, a fixed effects model will be applied for the reason that this thesis looks at within-panel effects. The fixed effects analysis covers the possible endogeneity issue of childcare utilization, without depending on the amount of the childcare supplement and/or the policy reform. Throughout the analysis, the standard error will be clustered at the individual level.
Hours worked:
Before going to the main analysis of this thesis, the hours a person works are analyzed. This is done to investigate the possibility of a causal link between the allowance and the hours worked. The policy reform might be an inventive for people to alter their number of hours worked on order to obtain a higher childcare supplement. This way, the policy reform might impact the health of the parent indirectly through the effect on the person’s work decision. Prior to the policy reform, the childcare supplement was linked solely to the overall income of the household, which is less easily adjustable. To see if this policy reform was an incentive of people to start working more or less, the hours worked of the least working parent will be modelled.
The least working parent is identified by using both the respondent identifier and the household identifier. First, the number of respondents per household that filled in their hours worked are identified. The number of respondents per household can differ over time due to drop-out of certain members of the
69 Individual fixed effects are included into the model.
70 Due to the fact that the childcare supplement is linked to the income and hours worked (of the least working
27
household71. Based on the identification of the number of people per households, the parent who works the least can be identified. Zero values for “hours worked” are converted into missing values. When taking the hours of the least working parent, one needs to pay attention to the age of the respondent, since children also filled in their hours worked. These hours cannot be used for this analysis due to fact that they are not linked to the childcare supplement. The person with the least number of hours work is identified by taking the minimum of the hours worked per household-year for those filling in that they have children or for those older than 2572. Of this subsample, 48,7% is male and 51,3% is female. However, this subsample might contain the person that works the least within the household in reality. If only 1 person of the household has filled in the work question for a particular year (while both him/her and his/her partner filled in these questions for other years), this person is considered both the least and most working parent. The following fixed-effects equation is used and is specified for the least working parent:
Hours worked = αi + β1Policy reform + β2controlit+ εit (1) In this equation, αi corresponds to the unobserved individual-level component (fixed effect) and εit represent unobserved variables influencing SHR which cannot be measured. The subscripts i and t index individual and year, respectively. The variable dummy variable “policy reform” is used as an instrument and indicates whether the observation is before or after the policy reform. A value of 1 indicates those observations that take place from 2012 onwards, while a value of 0 indicates those before 2012. Because the reform was enforced at the 1st of January 2012 , the observations in 2012 will also be considered to happen after the policy reform. The questions in the survey are all asked in such a manner that the information provided is about the year the survey is constructed73. This analysis includes the following control variables: age, occupation, education, income, having a partner, number of people in the household and the number of children. These factors are likely to influence the hours a person decides to work. The coefficient β1 is the focus of this analysis, since this coefficient show the effect of the policy reform on the number of hours worked. The results of this analysis will be discussed in the “results” section of this thesis.
71 This should not be confused by the amount of people in the household which is used as a control variable in the
regression, which is based on the survey question “Number of household members”. The number of respondents in the household in this case means the amount of people who filled in all the surveys themselves.
72 The questionnaires had a specific question that asked whether or not the person has children. From the age of 25
onwards, the distribution of the childcare utilization for those ≤4 years started to rise.
73 The data of the surveys are collected at the end of the year. The questions are mainly asked about the
