akgunduz

issue in designing the child care market. This is especially relevant in the child care market as the claims of markets’ efficiency and competition’s ability to improve quality is based on well informed parents. If parents are unable to distinguish between high and low quality child care, a private market may instead lead to competition on prices and a race to the bottom in quality.

This chapter analyzes the impact of competition on child care quality in the Dutch market. The impact of competition in the more traditional education institutions such as secondary schools has been a long standing question. However, child care quality is difficult to observe for parents, since there are no obvious quality indicators such as graduation rates or grades. Consequently, the effects of competition may differ between secondary schools and child care centers. Whether market competition has the socially desirable effect of improving quality in child care markets is thus a separate and primarily empirical question.

The main data source used in the analysis is Pre-Cool, a survey of child care centers in the Netherlands. The Pre-Cool survey provides data on process quality of the participating child care centers. Process quality measures are designed to capture the experience a child has in the classroom by assessing child-caregiver interactions. Process quality data is collected by trained observers who visited each center to assess classrooms. The Pre-Cool also includes a survey of the center managers’ which includes questions on structural characteristics such as the centers’ age and size. Using data provided by Statistics Netherlands, we proxy the level of competition a center faces by the number of daycare centers around it. To overcome any potential endogeneity problems, the competition variable is instrumented using the density of primary schools within the same area. The lagged density of births in the neighborhood is introduced as a secondary instrument to test the robustness of the results.

Our results show that competition has a significantly positive, but modest effect on child care quality in daycare centers. These results are consistent across different instruments and do not change when price is taken into account. Some center characteristics also have significant effects on quality; older and smaller centers tend to have higher process quality scores. In line with the results of Blau (2000), classroom variables such as the number of staff and children have insignificant effects on process quality.

The main contribution of our study is to extend the analysis of the impact of competition from later stages of schooling to child care in early childhood. Perhaps

partly because of the difficulty of collecting reliable quality indicators, there are no previous studies on the effect of competition on child care quality that we are aware of. In contrast, the impact of competition on the quality of education in secondary schools has been empirically studied at various levels. The results indicate a positive effect both at the micro and cross-country levels, although the size of the effect tends to be modest (Belfield and Levin, 2002; Sandstrom and Bergstrom, 2005; West and Woessmann, 2010).

The remainder of the chapter is structured as follows. In the following section, the institutional characteristics of the Dutch child care sector are discussed. Section 3 provides a theoretical basis to interpret the role of competition within the Dutch child care sector. Sector 4 details the econometric issues encountered and describes the IV method applied. Section 5 summarizes and describes the data available. Section 6 presents the results. Section 7 concludes.

4.2 Child Care in the Netherlands

Prior to 2005, the Dutch child care sector consisted of locally subsidized, employer financed and privately financed centers, each with their own financing structure (OECD, 2002). The Child Care Act of 2005 privatized the entire daycare market and all parents now receive a subsidy from the government for their expenditure on formal daycare up to a set hourly price. By introducing a nationally organized demand driven financing system, the 2005 Child Care Act ensured that all parents have access to same subsidies and consequently pay similar net prices that differ only by their income and by the different gross prices charged by the centers. The underlying assumption in the shift towards private providers is that price-quality ratio will improve if providers efficiently respond to parental preferences. Nevertheless, the potential for a drop in quality as a result of competition is recognized. Therefore regulations are set up through negotiations between child care providers and parental organizations. Regulations are placed on structural quality indicators such as staff-to-child ratios and caregiver qualifications. Monitoring for compliance with the regulations are handled by the municipalities.

The price cap on the subsidies, which was e 6.36 in 2012, effectively places a soft cap on the prices in the market and limits variation. Since the portion of the hourly price above e 6.36 is paid in full by the parents, any increase above the government cap leads to a very large net rise in what parents pay. The financing system is

essentially similar to the voucher system that is often suggested for primary and secondary schools (Friedman, 1997), with the clear difference that parents only receive subsidies if they choose to use a daycare center, while schools are mandatory. Supply and availability has increased rapidly since 2005. Further increases in subsidies in 2007, despite the cutback in 2009, has led to the 2012 situation where nearly 60% of children under 4 in the Netherlands receive formal child care (Bettendorf et al., 2012).

The rise in child care use following the Child Care Act has been accompanied by a steep decline in observed quality of Dutch child care according to developmental psychologists working within the Dutch Consortium of Child Care Research (NCKO) (Vermeer et al., 2008). Using process quality instruments that measure child-caregiver interaction, NCKO researchers find that quality in Dutch child care centers has dropped from 5 to 3 on a scale from 1 to 7 between 1995 and 2008. The most recent study in 2012 show that the quality levels have stabilized, but did not rise back to its previous levels (NCKO, 2013). The results are interpreted as a decline from above average quality to below average. Lower process quality is observed despite continued regulations on structural characteristics of child care such as staff-to-child ratio and staff qualifications. The question remains whether the drop in quality can be attributed to the introduction of market forces.

4.3 Theoretical Framework

Most parents use child care that is nearby. As a result, the market for child care is not uniform across a country and is instead composed of many smaller local markets each serving an area with a small radius (Blau, 2000; Cleveland and Krashinsky, 2009). The geographical limitations of the child care market inevitably introduces differing degrees of competition for centers in different areas. The fundamental question in this chapter is thus not whether imperfect competition exists in child care but whether competition has an impact on process quality.

The standard reasoning on the potential impact of competition on efficiency or productivity is that firms are forced to become more efficient to survive against their competitors. With regards to efficiency, competition can be seen as unambiguously positive. In case of child care quality, the existence of any effects from competition is more ambiguous. The main issue is about the observation and processing of information about quality by the consumers, in this case the parents. Parents may be unable

to distinguish between high and low quality child care, especially in terms of process quality. Using data from the United States, Mocan (2007) finds that parents in the child care market are weakly rational and do not use all the available information in making their decisions. This leads to a market with both information asymmetry and adverse selection. Even if parents had information about process quality levels, they may remain insensitive to process quality and focus on other aspects of child care such as prices and flexibility (Plantenga, 2012). Blau and Hagy (1998)’s finding of a small elasticity of income for structural quality seems to be consistent with information asymmetry in the market. Without parents explicitly opting for higher quality, competition is not likely to have any impact. In the Dutch market, the impact of competition is further limited because of the regulations on structural quality indicators such as staff-to-child ratios and teacher qualifications. Using a panel dataset of child care centers, Hotz and Xiao (2011) report that such quality regulations in the child care market appear to improve quality in the United States, even though they result in a smaller sector size.

In a market in which structural characteristics are regulated, an additional mechanism through which competition can have an effect can be found in the literature on managerial slack (Nalebuff and Stiglitz, 1983; Schmidt, 1997). Partially based on an earlier study by Leibenstein (1966), the managerial slack hypothesis predicts that imperfectly monitored managers and staff who are employed in an uncompetitive market can slack, ensuring that the firm survives but not providing the effort that they would have, had they been employed in a more competitive market. This hypothesis does not exclusively refer to the manager’s effort. Caregivers may also have lower effort in markets where the managers have no incentive to monitor or replace employees with poor performance levels.

The managerial slack hypothesis in the Dutch daycare market with its limited price variation and structural regulation can be shown formally in a model with two players, the firm and the manager. Assuming that there is free entry into the market m for center j, profits P can be set to equal 0 in the equilibrium. We make a strong simplification and assume that costs for the firms are exogenously given and equal for all firms in the Dutch setting because of the structural quality regulations and the larger labor market which results in a common wage rate. To break even given the government set price p, the center needs to attract sufficient number of children by offering quality Q . Since for each market, the competition level em and the demand characteristics Dm are different, the equilibrium quality level is given by the

function Q (Dm,em), which is increasing in both Dm and em. Centers in areas where parents are less interested or informed about process quality will have a lower quality requirement for survival. Quality itself is produced through two inputs Q = s + e, the structural factors s and managerial effort e. To ensure an interior solution at the equilibrium Q , we assume Q > s . Managerial effort is determined by the manager’s utility function. Although there is no monitoring, the manager gets w only if the company survives by having profits equal to or greater than 0. Since P > 0 does not change the manager’s wage, the manager has no incentive to increase his effort beyond P = 0.

Assuming that the function Q (Dm,em) is additively separable and linear for em, managerial effort e needed to reach the break even point can be easily determined.

The presented formalization is simple but clarifies the two main issues in the child care market. First, equation (4.3) shows that effort and quality rises with competition. Any surplus from a local monopoly is absorbed by the manager since the firm cannot monitor effort and adjust wages accordingly. At high levels of competition where e < w, the manager has no incentive to put in any effort and would prefer the 0 pay-off rather than a negative pay-off. Hypothetically, high levels of competition can even have a negative effect on quality. Second, the demand characteristics matter. In markets where parents do not demand higher process quality, there is less incentive to put in the effort required to supply it. Of course, many complications are left out of the model. While price variation is low, it does exist in the Dutch child care market. Similarly, centers can opt to have structural quality above that required by regulation or manage lower costs while complying with the regulations. Although our main hypothesis is that competition and process quality are positively related, the impact of competition on process quality remains a fundamentally empirical question.

4.4 Empirical Methodology

The method of estimation for the effect of competition on quality in class i of center j located in market m would ideally be a linear OLS regression such as equation 4.4, where xi jm are class specific, zjm are center specific and Mm are market or area level characteristics and vi jm is the error term. Considering that observations from the same center are likely to have correlated unobservable characteristics, standard errors clustered at the center level need to be estimated for equation (4.4). The main interest is on variable em, which is assumed to capture the effect of competition. Previous studies on competition and quality or firm performance have used variables such as market power, market share or concentration (Nickell, 1996). In the case of service sectors such as health care or schooling, the competition a firm faces is usually measured by the number of firms operating nearby. In health care, Propper et al. (2004) and Bloom et al. (2010) analyze the impact of competition on hospital quality using the density of hospitals in the area. Agasisti (2011) finds positive effects from competition on schooling outcomes to be driven by the number of schools in the area. We follow the same line of reasoning as the literature on schooling and health care and measure competition in child care by employing the average number of daycare centers within three kilometers in the area to measure competition em.

The main econometric concern is a possible endogeneity problem with regard to the competition variable. In terms of the theoretical framework presented, the potential endogeneity arises from a plausible correlation between the demand characteristics Dm not only with quality as assumed, but also with competition em. While we later control for average income and the degree of urbanization in the area, not all demand characteristics can be included in the regression analysis. Dual income families in urban areas may have a strong preference to use daycare regardless of quality, leading to a downward bias in the OLS estimate of b4. Additionally, more centers may be started in areas where care quality is low in order to take over low quality centers’ pupils, which would also lead to a negative relationship between the number of centers in an area and quality. Either way, there is an argument to be made for potential endogeneity issues in the ordinary least square (OLS) estimates which would place a downward bias on the estimated coefficients.

A reasonable instrument needs to be both valid, thus correlated with the indepen-

dent variable, and exogenous from the error term. Competition has previously been instrumented using a proxy for the level of demand in the electricity sector (Fabrizio et al., 2007). To instrument the density of primary and secondary schools, a similarly demand side instrument is popular, namely the proportion of Catholics in the local area who historically tend to prefer private schools (Cohen-Zada, 2009). In the case of Dutch child care, number of children in the neighborhood would be the obvious choice as the demand side instrument. However, child density in the area needs to be included as an independent variable since positive shocks in the number of children can cause waiting lists in daycare centers which would hamper competition until supply can adjust. Instead, we use the density of primary schools in the area as an instrument. The density of primary schools acts as a lag of the potential demand in the area, allowing us to circumvent short-term shocks in fertility. More crucially, primary school attendance is mandatory unlike child care and omitted demand characteristics which may have an impact on quality can not have an impact on the number of primary schools needed in an area. Furthermore, as of 2009, primary schools are directed to help parents find out-of-school care. The legislation implies that there are economies of scale to having daycare centers and primary schools at the same location, which is already a common occurrence in the Netherlands. Thus, we expect that primary school density would both be related to the number of child care centers in an area and exogenous from error term in equation (4.4). The first and second stages of the resulting estimation can be written as in equations (4.5) and (4.6). In all 2SLS estimates, standard errors clustered at the center level are specified.

The relevance of competition in the child care sector with regard to quality can be estimated using equations (4.4) and (4.6). However, to identify the complete impact of competition on quality, we need to take into account the small variation in prices in the Dutch child care sector. Although price variation is limited in the Netherlands due to the government cap on subsidies, there is some variation which remains uncontrolled for in equation (4.6). Competition may drive down prices first and only then improve quality. For example, even disregarding small differences in prices might lead to an underestimation of the casual impact of competition on the overall quality-

price level. Theoretically, price itself may be affected by the level of competition and including it as a control variable in equation 4.6 would lead to what Angrist and Pischke (2008) refer to as the ’bad control’ problem. In the intuitively plausible case of a negative effect on prices from competition and positive effect from prices on quality, there would be an overestimation of the effects on quality if price is added as a control variable. Rather than including price as an independent variable, we make price a part of the dependent variable by estimating equation 4.7, where quality is divided by price. Equation 4.7 thus estimates the effect of competition on the quality-price ratio that the center offers rather than the effect of competition on the quality level itself.

4.5 Data

Throughout this study, we make use of two data sources. Data on child care centers’ quality and characteristics are obtained from the first wave (2010-2011) of the Pre-Cool survey that is being conducted in the Netherlands. In addition, information at the neighborhood or municipality level for income, population, child care center and school density are retrieved from the Dutch Statistics (CBS). These two data sources are supplemented by information obtained from the child care centers’ official websites and municipalities’ inspection reports on the centers.

Unlike schooling where quality related variables such as graduation rates or grades are easily observable, child care quality is intrinsically more difficult to judge. The Pre-Cool survey includes observations by trained personnel who rate the process quality in a classroom according to the Classroom Assessment Scoring System (CLASS) used by developmental psychologists (Mashburn et al., 2008; Howes et al., 2008). Similar process quality instruments have been utilized by economists as well (Blau, 1997). A classroom is graded based on its performances in factors belonging to two large domains: emotional support and instructional support. The emotional support domain is constructed using four dimensions that classroom observers give grades on: positive climate, teacher sensitivity, behavior guidance and regard for child perspectives. The instructional support domain is made up of three dimensions: facilitation of learning and development, quality of feedback and language modeling. All dimensions within the domains are graded on a scale from 1 to 7, with 1 as the lowest and

7 the highest. 1

There are process quality observations from 65 daycare centers in the Pre-Cool sample. Multiple observations are made per center, allowing observers to rate both different groups and activities. Due to missing data issues, we make use of 294 observations from 40 daycare centers in the regressions. Table 4.1 presents the averages and standard deviations of quality measurements in all daycare centers within the sample, showing the state of child care quality in the Netherlands as a whole. Overall, process quality is above the average level of 4 for emotional support around 5. On the other hand, the instructional support scores are below average, around 3. Compared to previous studies using CLASS, child care quality in Dutch centers appears to be slightly below Finnish and above or equal to American centers. Howes et al. (2008) find an average emotional support score of 5.29 and instructional support score of 2.20 in the United States while the average of the scores reported by Pakarinen et al. (2010) in Finland are 5.19 for emotional support and 3.97 for instructional support.

Directly taking the average of the various dimensions of instructional and emotional support domains ignores potential consistency problems between the measures and diminishes the variation between centers’ quality levels. We make use of factor analysis to generate scores for emotional support and instructional support domains. In addition to the two domain variables, we construct an overall quality measure using all quality dimensions. The constructed summary variables from principal component analysis show a correlation well above 60% with all but one of the dimensions listed under both emotional support and instructional support. The exact factor loadings are presented in the Appendix. After normalization, the constructed summary variables have a mean of 0 and a standard deviation of 1.

Dutch Statistics provides data on the number of daycare centers within three kilo-

1More details on Pre-Cool observations and dimension scores can be found in Leseman and Slot (2013). Constructed variables and basic results of the Pre-Cool survey are to be made publicly available through the Dutch Data Archiving and Network Services (DANS). More information on DANS can be found at http://www.dans.knaw.nl/en, details of the Pre-Cool project is available in Dutch at http://www.pre-cool.nl/.