Explaining Local Variability in Child Care Quality ...

Early Education and Development

ISSN: 1040-9289 (Print) 1556-6935 (Online) Journal homepage: http://www.tandfonline.com/loi/heed20

Explaining Local Variability in Child Care Quality: State Funding and Regulation in California Bruce Fuller , Susan D. Holloway , Laurie Bozzi , Elizabeth Burr , Nancy Cohen & Sawako Suzuki To cite this article: Bruce Fuller , Susan D. Holloway , Laurie Bozzi , Elizabeth Burr , Nancy Cohen & Sawako Suzuki (2003) Explaining Local Variability in Child Care Quality: State Funding and Regulation in California, Early Education and Development, 14:1, 47-66, DOI: 10.1207/ s15566935eed1401_4 To link to this article: http://dx.doi.org/10.1207/s15566935eed1401_4

Published online: 08 Jun 2010.

Submit your article to this journal

Article views: 64

View related articles

Citing articles: 5 View citing articles

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=heed20 Download by: [University of California, Berkeley]

Date: 30 July 2016, At: 20:26

Early Education & Development

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

Volume 14, Number 1, January 2003

Explaining Local Variability in Child Care Quality: State Funding and Regulation in California Bruce Fuller & Susan D. Holloway University of California, Berkeley Laurie Bozzi Harvard University Elizabeth Burr, Nancy Cohen & Sawako Suzuki University of California, Berkeley

The uneven availability of child-care centers for different kinds of families has been detailed in recent years. Much less is known about the distribution of center quality across communities. Nor do we understand the role that local contexts or state policies may play in shaping quality levels. This paper describes several quality indicators, based on reports of 170 center directors situated in three California counties. We found that most centers in lower-income and workingclass communities displayed at least moderate levels of quality along structural measures, such as class size, the ratio of children per staff member, and staff education levels. About one in six failed to meet recommended quality standards. Some quality indicators were lower for centers located in communities with lower supply of center enrollment slots per capita, possibly due to excess family demand for center-based care. Center quality was not consistently influenced by community conditions such as poverty levels, ethnic composition, or maternal employment rates. Quality was higher among centers receiving stronger flows of public subsidies. We discuss the success of state agencies in advancing quality among centers and remaining policy challenges.

This project was supported by the Packard Foundation. Special thanks to Marie Young for her steady support. Patricia Siegel and Fran Kipnis, California Child Care Resource and Referral Network, were wonderful partners in designing and implementing the study. Correspondence concerning this article should be addressed to Bruce Fuller, University of California, Graduate School of Education, Tolman Hall 3659, Berkeley, California 94720 (e-mail: [email protected]).

48

Fuller, Holloway, Bozzi, Burr, Cohen, & Suzuki

Explaining Local Variability in Child Care Quality: State Funding and Regulation in California Overview: How is center quality distributed among communities?

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

The quality of child-care centers available to different kinds of families has received considerable attention over the past decade. Two-thirds of all mothers with a child, age 0-5 years, now work outside the home, spurring rising demand for child care (Hofferth, 1999a). A total of 5.7 million children were attending center-based programs in 1995, equaling 30% of all youngsters, age 0-5, nationwide (Smith, 2000). Recent work has revealed that basic access to center-based programs varies among different kinds of communities, and the quality of centers also may vary systematically. The proportion of 3-4 year-olds from affluent families who attended a center was almost twice the rate reported for children from low-income families (80% versus 45%, respectively) in 1995 (West, Wright, & Hausken, 1993). Counts of local centers correspond to the economic and demographic attributes of zip codes nationwide, sometimes lowest in working-class communities, not poor neighborhoods (Fuller & Strath, 2001). National data also reveal that 5 year-olds’ emerging reading skills and rudimentary math skills at the start of kindergarten are unequally distributed across social-class groups (West, Denton, & Reaney, 2000). Fewer studies have focused on the distribution of the quality of center-based programs observed among differing communities. Some studies have shown that quality indicators among centers in poor neighborhoods fall significantly below levels observed in centers situated in affluent or middle-class areas. But other research reveals that quality – along certain organizational attributes such as child-to-staff ratios or staff salaries – can be higher in subsidized centers, compared to those supported by parental fees (Fuller, Raudenbush, Wei, & Holloway, 1993; Phillips, Voran, Kisker, Howes, & Whitebook, 1994). Just as relatively high levels of center supply have been observed in low-income communities – compared to working-class neighborhoods where families can not afford center fees and few qualify for subsidized slots – we also may observe stronger quality in lower-income communities. This may be more likely in states that progressively target funds on centers situated in low-income communities, then regulate along quality standards. California is a good place to explore this working hypothesis, since this state contracts with about 1,300 different local agencies, including school districts, churches, and community based organizations (CBOs), to operate centers that serve low-income families. The present study examines variability in the quality of centers situated in low-income and working-class communities, nested in three California counties: Los Angeles, San Francisco, and Santa Clara. Directors of 170 centers were interviewed about a variety of organizational characteristics and quality indicators. We also assembled data on the 20 zip codes in which these centers were located. In exploring variations in center quality among communities, we were guided by these empirical questions: • What levels of center quality are observed among centers situated in different lower income communities? • Does center quality covary with local supply conditions? Do we observe, for example, lower quality centers in neighborhoods with a higher supply of centers?

Local Variability in Center Quality

49

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

• Is center quality influenced by the director’s ability to draw public subsidies from government, or do quality levels necessarily covary with centers' surrounding economic and demographic contexts? The policy theory underlying targeted financing and state regulation of subsidized centers is that local conditions will not matter. We first review the literature on what’s known about center quality across different communities. Second, we describe our study design. Third, we report descriptive findings on center quality, then estimate levels of quality among participating centers. We discuss the extent to which state subsidy flows or community contexts account for variation in quality. Fourth, we discuss the policy implications of our findings.

The local distribution of center supply and quality Significant strides have been made in equalizing access to child-care centers since the early 1960s, including expansion of Head Start and state-funded preschools for families living near or below the poverty line. Disparities certainly persist, especially for working families who neither qualify for subsidized care nor have the economic means to pay fees charged by centers. One recent study in California found that the supply of enrollment slots in centers per 100 preschool-age children was three times greater in affluent zip codes (based on median household income) than in zip codes populated primarily by lower-income families (Fuller, Kipnis, Coonerty, & Choong, 1997). Yet new estimates of the supply of center teachers across the nation’s zip codes with preschool-age children confirm the presence of a curvilinear relationship between community wealth and the center workforce in some states. That is, the number of center teachers and classroom aides per capita is relatively high in poor zip codes, then falls for blue-collar and lower-middle areas, and then climbs upward again in affluent zip codes (Loeb, Fuller, & Strath, 2001). Evidence of this curvilinear pattern had surfaced earlier when looking at familylevel data and the propensity of parents to select center-based programs by income level (Phillips et al., 1994; Fuller, Holloway, & Liang, 1996).

How is center quality distributed across communities? This growing literature on the distribution of center supply leads to the parallel question of how the quality of centers may, or may not, be tied to the economic and demographic features of their local communities. Government’s ability to expand center-based programs in low-income communities over the past 40 years has moderated the linear association between family income and access to centers. But can we observe the same success when it comes to quality indicators? Has public action been able to detach quality levels from the immediate context in which centers are situated? Only a few studies have attempted to inform this question, and they have yielded mixed findings on the degree to which wealth or poverty levels are associated with quality. Drawing on two surveys of center quality, Phillips and colleagues (1994) found that centers serving lower-income families were of comparable or higher quality compared to centers serving middle-class families. These findings were based on structural indicators of quality, such as child-to-staff ratios and teacher salaries. On process indicators related to such things as the educational program and teacher-child interactions, however, the first group fell below the

50

Fuller, Holloway, Bozzi, Burr, Cohen, & Suzuki

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

second. These findings are quite similar to a subsequent study drawing on a national probability sample of almost 1,900 centers spread across 38 states (Fuller et al., 1993). More recent evidence from the national evaluation of Early Head Start programs shows that mothers at or below the poverty line selected centers for their infants that displayed at least moderate quality levels (Boller, Paulsell, & Raikes, 2001). Other research teams have found a significant linear association between center quality and the economic or social-class characteristics of the families being served. Authors of the Cost, Quality, and Child Outcomes study, for example, reported this linear relationship. Children of poor families were enrolled in centers displaying lower quality in terms of structural features, such as scores on the Early Childhood Environment Rating Scale (ECERS), as well as on process measures, compared to centers serving children from middle-class and affluent families (Helburn et al., 1985). Recent data from the Growing Up in Poverty Project also revealed that center quality (in terms of ECERS scores) was significantly lower for those serving welfare-poor families, compared to earlier quality assessments done in two of the three states studied: Connecticut and Florida (Fuller et al., 2000). But in the study’s two California sites – San Francisco and Santa Clara counties – the quality of centers serving welfare-poor families was actually higher on average than centers previously assessed and situated in middle-class communities.

Can state policies mediate the effects of local context? Center quality in California This apparent inconsistency in findings may be attributable to a pair of mediating forces. First, we now know that the per capita supply of center (or preschool) programs is not linearly associated with wealth or poverty among neighborhoods in some states (Fuller & Strath, 2001). This may be due to the effective targeting of subsidies on lower-income communities via Head Start, state preschool and related center-based programs, and the growth of child-care vouchers that provide revenues for centers serving children from poor families. Whether similar progress has occurred in equalizing the quality of centers is the question on which the present paper focuses. The intensity with which states regulate the quality of all centers, not just those situated in lower-income communities, may represent a second factor that mediates the relationship between community (or family) wealth and the quality of proximate center-based programs. The overall effect of regulation and inspection is uneven among states, although the intensity of inspection appears to help sustain higher center quality in certain states, according to one review (Hofferth, 1999b). Centers that contract directly with the state, or those operated by school districts, are held to specific quality standards. California is a useful case in point. State and federal spending on child care and preschool programs now equals $3.2 billion annually. This funding is distributed in part through about 1,300 contracted agencies that provide center-based programs situated in low-income and working-class neighborhoods. In addition, about half of this funding is distributed in the form of vouchers (or “alternative payments”) directly to parents, mediated by local agencies that also provide information to families about child-care options. Families earning up to 75% of the state median income, about $32,000, can qualify for a center slot or portable voucher which may be used in a center or home-based arrangement (Fuller, Kagan, Caspary, & Gauthier, 2002).

Local Variability in Center Quality

51

California also places all centers under one of two different regulatory regimes. Title 5 centers include those that receive public funding, and they must follow strict quality standards related to class size, children-to-staff ratios, and teacher qualifications. These quality standards are high, at least relative to centers that fall under Title 22 regulations which include private for-profit and nonprofit centers that do not participate in subsidy programs. Under this regulatory structure, those centers situated in lower-income communities face stronger quality standards, and perhaps more frequent on-site monitoring, than centers situated in middleclass communities.

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

Three empirical questions This earlier work leads to three intertwined research questions that framed our quality survey in California. We decided to examine variability in the quality of centers situated across lower-income and working-class communities. Our first empirical question: What levels of center quality are observed among centers situated in different lower-income communities? Given debate in the literature over the relative supply and quality of centers in poor, working-class, and middle-class communities, we decided to focus on lower-income communities. The present study focuses on structural indicators of quality that could be reliably assessed through our phone interview with 170 directors located in one of three California counties. In two of the counties, we have conducted observational assessments of center quality which include structural and process indicators, reported in another paper (Fuller, Chang, Suzuki, & Kagan, 2001). The second empirical question: Does center quality covary with local supply conditions? We may, for instance, observe lower quality in centers that are situated in neighborhoods with a higher supply of center-based enrollment slots (center capacity per capita). Policy makers may face a trade-off between expanding supply or improving quality. Head Start continues to confront the question of how to expand enrollment without thinning-out quality. On the other hand, if state government expands center-based programs under relatively demanding quality standards and targets new funding carefully on low-income communities, then we would not necessarily observe lower quality in communities with richer supply. Third, we ask a broader question related to the efficacy of state targeting and regulatory action: Is center quality influenced by the director’s ability to draw public subsidies from government, or do quality levels necessarily covary with cemters' surrounding economic and demographic contexts? The policy theory underlying targeted financing and state regulation is that local conditions should not be driving quality. These forms of state action are intended to advance quality standards independent of families' ability to pay for higher quality. We can empirically test this proposition by seeing how community-level variables are related to indicators of quality, given significant variation across the lower-income and working-class neighborhoods from which we sampled centers. Directors do vary in their capacity to attract federal and state subsidies, including state contracts, federal food program dollars, and child-care voucher support.

Method: Gauging Center Quality in Lower-Income Communities Sampling communities and centers We first selected zip codes in each of the three study counties – Los Angeles, San Francisco, and Santa Clara – in which a large share of the families earned less than 200% of the poverty line, equaling about $24,000 in annual income in 1995. At the same time, we

Fuller, Holloway, Bozzi, Burr, Cohen, & Suzuki

52

wanted to maximize variation in center supply across selected zip codes. Working with representatives of local child-care agencies, we selected six or seven zip codes in each county that met these criteria. In Los Angeles County the selected zip codes were drawn exclusively from the south-central region to simplify data collection.

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

We then assembled a list of all operating centers in each sampled zip code. All of these centers were contacted by phone or letter, and we attempted to interview the center director. Note that this sampling method yields a set of centers that are largely representative of lower-income communities stratified by center supply levels. The resulting set of centers is not representative of each county overall. Characteristics of sampled communities are reported in Table 1. These descriptive statistics are for each participating county. Median household income and center enrollment capacity per 1,000 children under 6 years-old (a common indicator of center supply) are reported at the zip code level. Median income levels are low relative to each county’s overall income level in 1996. In addition, attributes of zip codes vary within a county. For example, income levels among the seven Santa Clara zip codes range from $26,122 to $49,245. Table 1. Characteristics of Sampled Neighborhoods in Three California Counties F-value and statistical (N = 48 centers) (N = 64 centers) (N = 58 centers) significance of mean differences

Los Angeles San Francisco

Santa Clara

Median family income in tracts of sampled centers1

$19,988

$26,724

$38,147

37.96*

Median family income countywide (1996 estimate)

$35,089

$36,162

$49,083

23.03*

Estimated maternal labor force participation (%)2

40

57

52

72.60*

Percentage of population, African-American

50

23

4

70.71*

Percentage of population, Latino

49

21

40

29.59*

Number of churches per 1,000 adults3

1.9

1.1

0.6

53.67*

Center capacity in slots per 1,000 children under 6

137

324

229

83.00*

FCCH capacity in slots per 1,000 children under 6

29

59

80

22.02*

Child care supply

* p < .0001. 1 Based on 1990 census data for the tracts in which sampled centers are located. 2 Percentage of mothers with preschool-age children employed outside the home. 3 Census data from 1990 for zip codes in which sampled centers are located.

Local Variability in Center Quality

53

Center enrollment capacity varies widely among and within counties. Table 1 shows that centers situated within selected San Francisco zip codes provided 324 slots per 1,000 children under 6, on average, compared to 229 and 137 slots in Santa Clara and Los Angeles in 1996, respectively. Among San Francisco’s six lower-income zip codes, capacity ranged from 177 to 453 slots per 1,000 children. Across the six Los Angeles zip codes, enrollment capacity ranged between 49 and 226 slots per 1,000.

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

Director interviews and center-level measures In spring 1998 we attempted to contact directors of all centers located in each selected zip code to conduct a phone interview. The overall response rate across the three counties equaled 84%.1 In total, 176 directors agreed to participate. Complete data on the major variables of interest were available for 170 centers. The interview questions were organized around four basic areas related to quality indicators, organizational features of the center, and attributes of the director. Quality indicators. Interviewers asked a series of questions that yielded structural indicators of center quality. These included the maximum number of children in class groups of 3 year-olds; the average ratio of children per adult across all classes comprised of 3 yearolds; the number of staff who left the center in the prior year; and the director’s own school attainment level. The first three indicators have been associated with higher levels of young children’s cognitive growth (for reviews, Burchinal, 1999; Shonkoff & Phillips, 2000). Organizational size and complexity. We looked at three indicators of complexity, each related to the center’s capacity to serve more, or a wider range of, young children: current enrollment of children, age 0-5, whether attending full- or part-time; whether the center served children under 2 years-old; and whether the center served children with special needs. Securing resources from external agencies. Quality may be influenced by the director’s ability to acquire resources from external agencies. Many centers located outside lowerincome communities rely solely on parental fees for operating revenues. But among the 170 participating centers, just 26% of the directors reported that they were not serving any subsidized children. We also asked directors about nine different forms of involvement or types of resources that could be gained from their local resource and referral (R&R) agency, including parent referrals, staff training, information about subsidy flows, and lending children’s books or other materials. Resource and referral agencies provide information to parents about child-care options and these services to providers, funded through federal and state funds and situated within geographic regions of counties. Director characteristics. These attributes included age, ethnicity, length of tenure at the center, school attainment, and training specific to early childhood development. California has a particular regulatory standard for directors who serve in publicly funded centers, requiring 12 credit hours of early childhood course work and two years of classroom experience.

Community-level forces that may influence center quality We know that local economic and demographic forces contribute to the supply or enrollment capacity of centers, as reviewed above. Similarly, centers located in better-off 1

Nonresponses included 12 centers that had closed or did not serve any 3 year-old children.

Fuller, Holloway, Bozzi, Burr, Cohen, & Suzuki

54

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

zip codes may be able to sustain stronger levels of quality: they are less reliant on subsidy flows for revenue; working conditions may be more pleasant, reducing staff turnover; better educated teachers and staff may be attracted to these more desirable settings. The counter hypothesis is that subsidy flows and quality standards are equitably applied across a range of centers. Therefore, variability in community wealth, demographic features, or center supply will not significantly influence internal quality. In short, targeting subsidies and enforcing quality standards from the state capital may be insulating centers from external exigencies. We assembled economic and demographic data on the 20 zip codes and their 95 component census tracts in which all sampled centers were located, largely from 1990 census data. This allowed descriptive analysis of variability in neighborhoods and multiple regression analysis of whether quality levels are sensitive to their surrounding local contexts.2

Findings: Explaining Variability in Center Quality Descriptive findings: quality indicators by county Three structural indicators of center quality are summarized in Table 2: maximum group size, child-to-adult ratio (for 3-year-old groups), and the director’s school attainment (percentage with graduate-level training). These indicators differed among counties at marginal to strong levels of statistical significance. Across all participating centers, the mean maximum group size for three-year-olds equaled 16.0 children, with means ranging between 14.8 children in Santa Clara County centers and 18.4 in Los Angeles. Mean differences across the three counties are marginally significant (at p < .08). Table 2. Center Quality Indicators by County F-value and statistical (N = 48 centers) (N = 64 centers) (N = 58 centers) significance of mean differences

Los Angeles San Francisco

Santa Clara

Maximum child group size for 3-year-olds

18.4

15.9

14.8

2.64*

Ratio of children per adult in classrooms of 3-year-olds

5.9 to 1

5.8 to 1

7.6 to 1

6.91**

Percentage of center directors with some graduate school

38

41

26

2.39*

* p < 10; ** p < .001.

2

Our sampling design did not yield enough zip codes to study hierarchical linear models (HLM) to exhaustively assess the mediating influence of zip-level attributes, perhaps reducing the influence of center-level factors on quality indicators. Future work could formally explore these bi-level processes.

Local Variability in Center Quality

55

Figure 1 places these indicators in context by displaying average group size for earlier national samples of centers (maximum group sizes not consistently reported). Our sampled California centers compared favorably to the earlier investigations. The mean group sizes for 3-year-old classes were 14.8, 13.1, and 13.4 in Los Angeles, San Francisco, and Santa Clara counties, respectively. Even for the wider range of centers in the Cost, Quality, and Child Outcomes study (Helburn et al., 1995), that research team reported average class sizes at 13.7 for their subsample of California centers. Figure 1. Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

Comparison of mean group size in California centers and nationwide (pertaining to 3-4-year-old children).

Los Angeles County

14.8

San Francisco County

13.1

Santa Clara County

13.4

Cost Quality & Child Outcome Study

13.7

Growing Up in Poverty Project Centers

13.1

NAEYC Maximum Level

17.5

0

5

10

15

20

Average (Mean) Count of Children in Classrooms for 3-4-Year-Olds Data from the Cost, Quality, and Child Outcomes Study Team (1995) are from a sample of 400 centers spread across four states, including California. Centers observed by researchers from the Growing Up in Poverty Study are located in one of three states, including California (Fuller, Kagan et al., 2000). NAEYC professionally recommended maximum class group size for 3-year-olds are available from the National Association for the Education of Young Children, Washington, DC

One notable finding is that 28% of all centers in the present study reported a maximum group size of more than 20 children. While fairly strict quality regulations are holding class sizes at reasonable levels in most centers, some are unable to meet these recommended standards. These centers may not fall under the tighter Title 5 quality standards, since they do not have a contract with the state. While maximum group size was lowest in Santa Clara County, these centers reported the highest child-to-adult ratio, 7.6 children per adult staff member. This compares to 5.8 children per adult in San Francisco. Centers in Los Angeles are similarly able to maintain a low ratio, 5.9 children per adult, despite reporting the highest maximum group size.

Fuller, Holloway, Bozzi, Burr, Cohen, & Suzuki

56

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

California’s quality regulations for contracted centers under Title 5 require that the number of children per staff member not exceed eight. This standard is on the more demanding end of the range of acceptable ratios as recommended by one leading professional group, the National Association for the Education of Young Children (NAEYC). If centers are not linked to the state’s contracting system, they can legally raise their classroom ratio to 12 children per adult, lowering per child operating costs. Other states have a more stringent staffing benchmark. In our sample, 26 of the 159 directors (16%) who provided complete data reported that at least one class group (for 3 year-olds) exceeded the maximum allowable staff ratio under Title 5. Mean child-to-adult ratios for our sampled centers look very good, compared to the earlier national studies. Los Angeles and San Francisco centers, with ratios at 5.9:1 and 5.8:1, respectively, are two children below the nationally representative sample of non-profit centers surveyed in 1990 (Kisker et al., 1990). In this latter study, the ratios equaled 8.1 children per adult in Head Start classrooms, and 7.9 in non-profit center classroom (Figure 2.) The Cost, Quality, and Child Outcome study team, which included centers in a diverse range of communities, reported a mean staffing ratio of 7.0 per adult, above the staffing ratio our sampled centers within Los Angeles and San Francisco. Figure 2. Comparison of mean child:adult ratios in California centers and nationwide (pertaining to 3-4-year-old children).

Los Angeles County

5.9

San Francisco County

5.8

Santa Clara County National Center Sample

7.6

Head Start

8.1

Non-Profit

7.9

For Profit

8.9

Cost, Quality & Child Outcome Study

7.0

NAEYC Maximum Level

8.5

0

2.0

4.0

6.0

8.0

10.0

Adults per Child Observed or Recommended for 3-4-Year-Old Groups Staffing ratios for the national probability sample of centers pertain to 3-4-year-old groups as detailed in Fuller et al. (1993). Data from the Cost, Quality, and Child Outcomes Study Team (1995) are from a sample of 400 centers spread across four states, including California. NAEYC professionally recommended levels are available from the National Association for the Education of Young Children, Washington, DC.

Center directors in San Francisco reported higher levels of graduate training. The average director, across all three counties, had completed a four-year degree but no graduate training.

Local Variability in Center Quality

57

In San Francisco, 41% of all directors had completed some graduate work, compared to 26% of all directors in Santa Clara County. These school attainment levels are about one year higher than levels observed in the more middle-class sample of 100 California centers studied in the Cost, Quality, and Child Outcomes study (1995).

Organizational characteristics Descriptive statistics on organizational size and complexity are reported in Table 3. We should emphasize that sampled centers were of modest size, enrolling 68 children on average

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

Table 3. Characteristics of Center Organizations F-value and statistical (N = 48 centers) (N = 64 centers) (N = 58 centers) significance of mean differences

Los Angeles San Francisco

Organizational Size Enrollments, children of all ages1 Number of class groups for 3-year-olds1 Services and Organizational Complexity Provides infant care (%) Accommodates special needs children (%) Open nontraditional hours (%) Have changed regular hours (%) Resource acquisition and subsidies Contracts for subsidized child slots (%) Enrolls children with vouchers (%) Participates in child-care food program (%) No children enrolled with subsidies (%) Uses R&R for different services (%) Index of R&R linkages2

Santa Clara

58 (72) 1 (2.2)

42 (62) 1 (1.7)

45 (70) 1 (1.8)

27 49

13 50

36 50

4.88** 0.01

10

7

0

2.97*

21

11

32

4.03*

40

50

47

0.60

19

33

21

1.84

67

66

57

0.73

25

31

22

0.64

67

70

57

1.25

1.4

2.1

1.6

1.73

0.43 1.99

* p < 10; ** p < .001. 1 Medians are reported, along with mean values (in parentheses). The latter indicate several larger centers with higher enrollments and more class groups. 2 We asked directors about 9 different R&R services in which their center might be involved. This index is simply the count of these linkages, ranging from 0 to 9.

58

Fuller, Holloway, Bozzi, Burr, Cohen, & Suzuki

and operating just one class group for 3 year-olds during a typical day. One-fourth of all centers provided infant care, and fully 74% served children who were subsidized in some way.

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

Table 3 also details how organizational features of centers varied across the three counties. For example, sampled centers in Los Angeles tended to be somewhat larger in terms of enrollments, serving on average 71.8 children, age 0-5, compared to 61.9 children among San Francisco centers. These small-scale organizations operate just one classroom for 3year-olds on average (medians). Mean values appearing in parentheses show that several larger centers do operate 2–4 class groups for this particular age group. In addition, 36% of all centers in Santa Clara County offered infant care, compared to just 13% in San Francisco. Very few centers were open during nontraditional hours — before 7:00 a.m. or after 6:00 p.m. But up to one-third have changed their hours of operation to accommodate parents’ schedules. Most centers in these lower-income communities, not surprisingly, draw public subsidies or serve families that do. Up to two-thirds of the sampled centers in Los Angeles and San Francisco participate in the federal child-care food program. About half across the counties have contracted enrollment slots with the state education department to serve children from lower-income families. This also means that they fall under the more stringent Title 5 state quality regulations. A sizeable share of centers have linkages to local R&R agencies, comprising 70% of the San Francisco centers. The number of services utilized by centers is modest, averaging just two of nine possible services offered by their local R&R. Overall, the structural indicators of quality look reasonably strong for the majority of participating centers. But staff turnover rates represent a notable exception. The developmental effects on children – resulting from daily interactions inside classrooms – can be significant and negative when children cannot form stable relationships with center teachers or classroom aides (Shonkoff & Phillips, 2000). We were able to estimate staff turnover rates in 128 centers, based on complete data provided by center directors.3 Of these, 48% reported that no teaching staff (lead teachers or classroom aides) had left in the prior year. Remember that most centers are small, operating two to three class groups total. Of those centers that did lose a staff member in the prior year, the median center had lost 18% of its staff, just under one in five teaching staff.4 Variability was wide, however, with one-fifth of all centers losing more than a quarter of their teaching staff in the prior year.

Variation in quality We observed wide variability in child-to-adult ratios across centers with complete data on this measure (n = 168). Despite low staffing ratios overall, many centers do have ratios that exceed the recommended maximum for class groups serving 3 year-olds (under Title 5 3

Staff turnover rates could be reliably estimated only for centers with no infant rooms, given the information that we collected during the director interview.

4

This average turnover rate is comparatively low, relative to higher rates of staff exit recently reported for centers in northern California (Whitebook, Sakai, Gerber, & Howes, 2001).

Local Variability in Center Quality

59

quality standards). Figure 3 plots the distribution of child-to-staff ratios for each county. The solid horizontal line within each box identifies the median ratio. The lower edge of the box marks the ratio value at the 25th percentile; the upper edge of the box indicates the value at the 75th percentile. While there is concentration around the median value which lies close to the regulated maximum of eight children per adult, 16% of all centers do exceed this value, with a slightly larger share over this regulated limit in Santa Clara County. Figure 3.

15 – Child-Staff Ratio

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

Distribution of child:staff ratios by county.

10 –

5–

0– County

San Francisco County (n = 60)

Santa Clara County (n = 55)

Los Angeles County (n = 44)

Multivariate estimation of center quality Do the resources acquired by center directors or the particular contexts in which centers are located help to explain quality levels? The short answer is, yes, but only to a limited extent. Center quality – at least along these structural indicators – is not dependent upon the economic strength or poverty levels of local neighborhoods. Put another way: the state’s targeting of subsidies and accompanying quality standards have largely met their intended aims, ensuring fairly high quality levels on average and a fairly equal distribution of quality among the lower-income and working-class communities in which our sample of centers is located. In addition, higher levels of public subsidies are associated with modestly higher levels of quality. This provides further evidence that state-targeted financing and standards yield significant effects. At the same time, our multivariate analyses show that larger centers which face pressure to enroll more children – especially those in Los Angeles where overall center supply is low – display lower quality in terms of worse staffing ratios and higher group size maximums. In this way, family demand for center-based programs that exceed available supply appears to erode these structural indicators of quality. Let’s begin by looking at the ability of center directors to acquire public resources, subsidies that may be related to sustainable staffing ratios, groups sizes, and staff turnover. Table 4 reports basic regression results for the subset of predictors that are significantly

Fuller, Holloway, Bozzi, Burr, Cohen, & Suzuki

60

Table 4.

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

How Center Size and Resources Help to Explain Quality and Organizational Complexity (unstandardized beta coefficients and t-statistics reported1) (1) Estimating child: staff ratio

(2) Estimating maximum group

.006 (1.79)*

.035 (3.25)**

Enrollment size Draw child-care food program resources

(4) (3) Estimating Estimating likelihood of likelihood of serving infants serving specialneeds children .008 (.003)***

.003 (-.003)

-1.22 (-2.44)**

.46 (0.35)

-.45 (0.39)

.66 (0.34)*

Serves subsidized children

.59 (1.12)

.33 (0.23)

-.04 (0.42)

-.32 (0.37)

R&R linkages index

.17 (0.26)

-2.64 (-1.54)

.87 (0.52)*

.86 (0.46)*

7.01

13.78

-1.47

-.42

Total Equation Intercept 2

F-value (or x )

2.90*

3.29**

11.31*

9.87*

n of cases

152

115

163

162

.06

.11

.10

.08

r

2

* p < 10; p < .01; *** p < .001. 1

The first two equations are estimated via ordinary least-squares (OLS). Models in columns 3 and 4 are logistic regressions, estimating the odds that a center offers infant care or accommodates special-needs children, respectively. For the latter two models, standard errors appear in parentheses.

associated with a reduced set of quality indicators (our dependent variables). The other predictors that appeared in the measures section above held no significant association with quality indicators. Centers that enroll more children appear to feel upward pressure on their staffing ratio (column 1) and their maximum group size (column 2). Center participation in public subsidy programs, when indicated by participation in the child-care food program, is associated with a lower child-to-adult ratio but shows no relationship with maximum group size. We are not arguing that the resources flowing from the food program directly lowers the staffing ratio. Instead, this predictor may proxy a center’s ability to attract different lines of public support. Interestingly, centers that report more activities with the local R&R agency display lower group sizes. The causal direction is not clear: it may be that higher quality centers, benefiting from more resources, possess a stronger capacity to engage their local R&R agency.

Local Variability in Center Quality

61

In column 3 of Table 4 we include a logistic regression estimating the probability that centers offer infant care. This explores whether resource flows are related to a key organizational feature, rather than to a specific indicator of quality. Once again, enrollment size is predictive: larger centers are more likely to be serving infants. A stronger association is observed for the degree of involvement with the local R&R. For each major activity with the R&R, the center is about two times more likely to be serving infants. In short, the capacity necessary for providing infant care may be related to engagement with an R&R agency.5

Downloaded by [University of California, Berkeley] at 20:26 30 July 2016

Does community context influence center quality? Next we introduce economic and demographic features of each center’s local community, drawing from the zip-code level data. Several of the demographic variables were intercorrelated. After conducting a principal components analysis, we constructed an index of the community’s economic status from three individual variables: the maternal employment rate, median household income, and percentage of population who were non-Latino white. Table 5 reports regression results for contextual predictors, focusing on the same three structural indicators of quality, plus the provision of infant care. For each regression estimation we first enter the three contextual predictors, plus a dummy variable for centers situated in Santa Clara County where we observed site effects (reported in columns 1A, 2A, and 3A). Then, we enter the earlier reported center-level predictors of quality, and retain in the regression model those community-level predictors that were significant (at p