The Legal Claim to a Place in Kindergarten. A Study of Methodological Approach in Public Child Care Provision

1. Introduction

The relevance of public child care provision has been an important issue in many countries for decades. Since many women with young children want to participate in the labor market, it is indispensable to have institutional provision for pre-school aged children. In order to in­crease public child care coverage in Germany, the state introduced a legal claim to a place in kindergarten, which became effective in 1996. The central goal of the legislation was the ex­tension of public child care for pre-school children, in order to increase female labor market participation. As reconciliation of family and working life has become a relevant political agenda in the past few years, there is a growing empirical interest to exploit the effect of subsidized public child care on maternal employment rates. However, the increasing empiri­cal literature shows mixed results for this topic. The large part of the studies identifies clearly positive effects of public child care provision on mothers’ employment (Berlinski et al., 2011 ; Lefebvre/Merrigan, 2008; Schlosser, 2007). For some countries, however, the economists identify no causal relationship between subsidized child care provision and maternal labor market outcomes (Havnes/Mogstad, 2011).

In their paper Stefan Bauernschuster and Martin Schlotter (2013) analyse the relationship between public child care and mothers’ labor supply in Germany after the introduction of the legal claim to a place in kindergarten in the time period from 1996 to 2001. The main econo­metric challenge of identifying the causal effect of kindergarten attendance of the child on mother’s employment lies in the non-random selection of pre-school education. To address this problem, they use the fact that (due to overdemand of child care) German children had to reach a given age before the kindergarten start to enroll in kindergarten. The authors apply the special feature of the cut-off rule for their quasi-experimental settings, which they investi­gate with an instrumental variable approach and difference-in-differences techniques. The empirical results of these quasi-experiments show that increasing public child care rises fe­male labor market participation. The fundamental question is, however, whether the results really give an empirically sound feedback in term of program effectiveness. Therefore, the following sections are going to review on methodological interpretation of the causality, in order to examine the validity of the study and thus, to ensure the effectiveness and appropri­ateness of the legal claim to a place in kindergarten.

This seminar paper is organized as follows. The second section begins with the background information about kindergarten system in Germany and describes the data used for the eval­uation. The third section is going to review on the methodology of the study. It focuses on the identification strategy, the estimation results and the validity of the results. The last section gives a brief conclusion to the discussed points.

2. Background Information and Data

Pre-primary education in Germany is not obligatory, but recommendable for pre-school chil­dren. It includes all institutions, which offer suitable child care for pre-school-age children until school entrance. Already since the time of industrialization child care has become an important issue as more and more children were left unsupervised by the parents. The first child care institutions in Germany were founded by Pauline zur Lippe-Ditmold in 1802 and Friedrich Fröbel in 1840. After 1900 the approach of “general German kindergarten” was ex­tended by Montessori’s and Waldorf’s kindergarten models. During the World War II a large number of kindergartens was destroyed and had to be reconstructed. However, due to his­torical differences between the West and the East Germany (GDR) the reconstruction of the pre-school institutions was of higher political priority in the GDR. In the West Germany child care was considered to be a private matter and the family as the best place for young chil­dren. Pre-school education was supposed to be supplementary for family education of the older children (5-6 years old). In contrast to it, in the East Germany the state offered full pro­vision of institutional care for all children between 1/2 and six years in order to free women for employment in the state industries and bureaucracy (Textor, 2005, p. 1).

After the reunification, the pre-school system of the West Germany was heavily criticised and the state made huge political efforts to provide more places in the West German states. A significant step to increase public child care coverage in the West Germany was the introduc­tion of a legal claim to a kindergarten place^[1] by the German federal government in 1996. The regulation was defined in § 24 of the German Social Code^[2] VIII. According to it, every child between three and six years old was eligible to a place in a child care facility.

The reform had to solve the problem of the poor child care system and to support young mothers, willing to participate in the labor market. However, since the introduction of the legal claim to a kindergarten place the growing demand could not be satisfied with the available child care institutions. Thus, it became necessary to provide additional child care facilities to comply with the new regulation. Since many municipalities were confronted with the problem of overdemand, the German Federal Parliament^[3] adopted an additional legislative initiative by the Federal Council of Germany^[4]. It allowed municipalities to use the so called month-of- birth cut-off rules to determine, whether a child has the right to public child care or not. Due to these rules, children were eligible to a place in a kindergarten only, if they turned three before the start of the kindergarten year; otherwise they had to wait until the next kindergar- ten start. The introduction of the reform affected public child care attendance for three and four year old children, which increased sharply from 1996 to 2001 (Bauernschuster/Schlotter, 2013, p. 2-5).

To quantify the effect of the increasing child care possibilities on maternal employment rates Bauernschuster and Schlotter applied two identification strategies: an instrumental variable approach and a difference-in-difference approach. For the first part of their empirical re­search the authors used the data of the German Socio-Economic Panel (SOEP), a represen­tative longitudinal study, which has collected data of private households, individuals and families in Germany since 1984. The dataset includes, among other data, information on pre­school education as kindergarten attendance and daily hours in preschool for all children in the household (Spiess, 2009, p. 5). The SOEP-sample consists of mothers with children born between 1992 and 2000 that are older than 36 months at the time of the interview, but at most 48 months old at the time of the last kindergarten start. There are 1.228 mothers and 1.936 mother-child observations (Bauernschuster/Schlotter, 2013, p. 14-15).

For the second identification strategy the data of the German Micro Census were used, in order to exploit changes in the employment rates of mothers with three and four year old children after the introduction of the regulation. The German Micro Census provides informa­tion on the population and the labor market in Germany. It accumulates detailed statistical information on the population structure, the economic and social situation of the population, families, consensual unions and households, on employment, job search, education and training, the housing situation and health status (Statistisches Bundesamt, 2014). The samle includes Micro Census waves of the years 1991, 1993, 1995 and 1996 until 2001. Each wave contains more than 5.000 mothers, whose youngest child is three or four years old and more than 4.000 mothers, whose youngest child is ten or eleven (Bauernschuster/Schlotter, 2013, p. 15-16).

Investigating the data of the SOEP and the German Micro Census, Bauernschuster and Schlotter found out that increasing public child care for pre-school children is positively corre­lated with the labor market participation of the German mothers. An important question is, however, whether this statistical relationship can be interpreted as causal effect of the bet­ter/increasing child care provision. As there could be other reasons for increasing employ­ment rates of (young) mothers, the main challenge is to indicate plausible causality between pre-school child care attendance and maternal labor market outcomes. For that purpose, the authors had to overcome several econometric complications. First, participation in the labor market and child’s kindergarten attendance are jointly determined. This simultaneity bias prevents causal identification by using simple OLS models. Second, the decisions about mother’s employment and child’s pre-school attendance are plausibly correlated with mother's cognitive abilities. It means that mothers, who send their children to kindergarten, could be systematically different in their demographic and/or socioeconomic characteristics (e. g. age, education, nationality or migration status) from those, who do not. Since these characteristics could affect mother’s choice about sending her child to a kindergarten, they might also influence employment rates and have to be taken into consideration, in order to eliminate unobserved heterogeneity (Bauernschuster/Schlotter, 2013, p. 6).

3. Methodology of the Study

3.1 Identification Strategy

The main complication of the causality identification is the possibility of inconsistent parame­ter estimation due to existence of unobserved variables. A standard OLS regression model is defined as:

Abbildung in dieser Leseprobe nicht enthalten

where x is an independent exogenous variable and y is a dependent endogenous variable. Special interest lies in the parameter ß, measuring the mean effect of x on y; ε is an error term, assumed to be independent and identically distributed. The main assumption for standard regression is that the regressor is not correlated with the error term. So, the varia­ble x can affect the variable y only in a direct way via term ßx and there is no association between y and ε. However, in many situations this assumption is violated because of the association between regressor(s) and error term. As a consequence, there is both a direct effect on the dependent variable via ßx and an indirect effect via ε. Since the main objective of regression is the estimation of the direct effect on x, the OLS estimator is biased and in­consistent for ß, when there is an association between xand ε (Cameron, 2013, p. 35).

To overcome the limitations of the standard OLS model, the authors use an instrumental var­iable approach in their first identification strategy, in order to measure causal effects of pre­school education on mother’s labor market participation. Caused by the date-of-birth cut-off rules, the special feature of the legal claim to a kindergarten place creates variation in pre­school attendance, which is exogenous to confounding factors determining both mothers’ decision about child care and their labor market participation. The mere existence of cut-of- rules provides the authors with a quasi-experimental setting and helps to get a good instru­ment for an instrumental variable approach. Instrumenting kindergarten attendance D,· of a particular child i with the eligibility to child care Z, they define child care attendance in a fol­lowing model:

Abbildung in dieser Leseprobe nicht enthalten

As some children may enter kindergarten irrespective of the cut-off-rule, δ on Zi should define the share of children, who enter pre-school institution according to the rule, in order to check the relevance of the instrument.

Furthermore, it is important for mothers’ decision to work, whether the child, eligible to child care, is the youngest in the household. It is necessary to exclude observations, in which the child, eligible to child care, has younger siblings, since it may prevent mothers from entering the labor market. In the next step, therefore, the authors model a local average treatment effect (LATE) τ of pre-school attendance of the youngest child on maternal labor market out­comes, using the predicted values Ď^[5] from the first equation:

Abbildung in dieser Leseprobe nicht enthalten

As long as the exclusion restriction is not violated, τ estimates the causal effect of public child care on maternal employment Y. The exclusion restriction would be violated in case that the age of the child has a direct effect on mother’s decision to work (e. g. more informal child care possibilities for older children and/or changing attitude to career as the child grows older). That is why, it is reasonable to include the age of the mother’s youngest child as a covariate in the model. In addition to the child’s age in months, the authors include informa­tion about several demographic and socioeconomic characteristics of the mother and her partner; information about the household’s size, the youngest child and its siblings (Bauern­schuster/Schlotter, 2013, p. 9).

For their second identification strategy Bauernschuster and Schlotter apply difference-in­differences techniques to exploit the increase in child care attendance in the West Germany from 1996 to 2001 and its effect on mothers’ labor market participation. For this purpose they build three groups: the first group includes mothers, whose child is three or four years old, the second group (control group) combines mothers with school-age children (ten and eleven years old) and the third group (alternative control group) consists of mothers without children. The alternative control group includes two sub-samples: 1) all women aged between 18 and 60 and 2) women aged between 29 and 36 years. To receive information about the devel­opment of the groups, the authors run placebo treatments in the pre-treatment period (before 1996) and after-treatment period (after 2001). Comparing the employment rates from 1996 to 2001 of the treatment and control groups, the treatment effect Θ can be defined as:

Abbildung in dieser Leseprobe nicht enthalten

where D; = 1 stands for the treatment group of mothers with three to four years old children and D, = 0 for the control group of mothers with older children ; T¡ = 0 stands for the year 1996 and T = 1 for the year 2001. The outcome is modelled by the following equation:

Abbildung in dieser Leseprobe nicht enthalten

where & measures time trends common to treatment and control groups, 3 measures aver­age permanent differences between treatment and control groups and θ identifies the true treatment effect. In order to include further covariates into the model, the authors use an ex­tended version of the regression to exploit the development of maternal employment rates over the whole period from 1996 to 2001. Furthermore, they run a placebo treatment in the pre-treatment years to test the assumption that both treatment and control groups follow the same trend (Bauernschuster/Schlotter, 2013, p. 11-12).

3.2 Estimation Results

3.2.1 Instrumental Variable Approach

At first, the authors make a descriptive comparison of mothers, whose youngest child was 36 months or older at the time of the last kindergarten start (= eligible for child care) with moth­ers, whose youngest child was younger than 36 months at the time of the last kindergarten start und thus affected by the cut-off rule (= not eligible for child care). The comparison shows that there is no significant difference between both groups in terms of demographic and socioeconomic characteristics of mothers and their partners. It means that both groups systematically differ from each other in terms of child care eligibility, but not in other charac­teristics. The results of this comparison provide the first support for the validity of the instru­mental variable approach (Bauernschuster/Schlotter, 2013, p. 17).

Another relevant aspect, which checks the relevance of the instrument Z, is the child care attendance. The comparison of the both groups shows that 51.5 percent of children younger than 36 months at the last kindergarten start and 84.5 percent of children above the cut-off age at the last kindergarten start attend pre-school at the time of survey (about 1/2 year after the kindergarten start). At the same time, mothers with children, eligible to child care are 9.9 percentage points more likely to be employed and work on average 2.5 hours more per week than mothers, whose youngest child is younger than 36 months. The Wald estimate proves the significant impact of pre-school attendance on mothers’ labor market participation. It sug-

[...]

^[1] Rechtsanspruch auf einen Kindergartenplatz

^[2] Sozialgesetzbuch

^[3] Deutscher Bundestag

^[4] Bundesrat

^[5] Ď stands for D with л, since there is no corresponding letter in MS Word