Table of Contents | Previous | Next |
III. QUALITY OF CHILD CARE USED BY EARLY HEAD START FAMILIES AND PROGRAM IMPACTS ON THE QUALITY OF CHILD CARE FAMILIES USED (cont'd)
C. SUMMARY OF EARLY HEAD START QUALITY OF CHILD CARE
Considering several measures of quality, we find that Early Head Start children generally experienced good quality—quality that meets the Head Start Program Performance Standards—in the child care centers they were enrolled in. On standard, widely used global measures, the quality of care appeared somewhat lower in family child care than in center arrangements. However, when compared using a measure of specific caregiver-child interactions, we found evidence that family care arrangements may provide advantages for children in terms of the amount of caregiver verbal interactions with the Early Head Start child. However, given the small percentage of family care arrangements we were able to observe, it is difficult to draw conclusions about the quality of family care experienced by the full sample of Early Head Start children. Community centers used by Early Head Start families were generally of lower quality than the Early Head Start centers, but their quality ratings improved over time so that by the time Early Head Start families were placing their 3-year-olds in community centers, their quality was very close to that of Early Head Start centers.
D. EARLY HEAD START PROGRAMS’ IMPACTS ON CHILD CARE QUALITY
We now come to one of the central questions about the role of Early Head Start programs in providing child care opportunities for low-income families with infants and toddlers: was the program effective in ensuring that its children were in child care settings of higher quality than available to the families who had been randomly assigned to the control group. First, we examine the extent to which families’ participation in an Early Head Start program increased their likelihood of using higher-quality child care. This analysis is followed by a second set of analyses in which we show how Early Head Start programs made a difference in the levels of quality in the child care arrangements in which they placed their children. In short, this section reports findings related to whether Early Head Start programs accomplished what they set out to do—to improve the chances for children to experience good-quality child care.
1. Approach to the Analyses of Impacts on Child Care Quality
Sample Limitations. As described in Chapter I, observations of quality in child care settings were completed for a subsample of the families who reported using child care. Not all families used child care, nor were we able to observe all those who did. The sample used for analysis of the impact of Early Head Start programs on child care quality includes all four center-based sites, four mixed-approach sites in all three time periods, and one additional site for the mixed-approach analysis at 36 months.6 Since we were not able to observe quality in most home-based or family child care arrangements, we did not conduct analyses of the programs’ impact on the quality of family child care used by Early Head Start children. Nevertheless, we were able to address a significant part of the question the Advisory Committee posed regarding Early Head Start increasing the probability that children would receive quality child care. The analyses reported provide clear answers to the question: Do Early Head Start programs increase the probability that children will receive good-quality center care?7
Analytic Issues. Child care quality could be assessed only if families were using child care. It seems likely that the families who were using child care differed in important characteristics from those families who were not using child care. Furthermore, it is possible that the factors affecting which families placed their children in child care were different depending on whether the family was in the program or control group. Therefore, because of the potential for bias, it would not be appropriate simply to compare program and control-group quality for those children who were in child care. We already know, as presented in Chapter II, that program families were more likely to use center care, so it is possible that characteristics of families in the two groups differ, and are themselves correlated with the quality of the centers used.8 Additional biases could have been introduced because we were not able to complete observations in all eligible settings, and we do not know the extent to which the centers observed are representative of all centers that could have been observed. It is possible, for example, that the centers that allowed our observers in were of higher quality than those who refused to be observed. To minimize these selection-bias concerns, we conducted the impact analysis in two stages.
Stage 1 Analysis. To avoid the selection bias issue for this research question we conducted the first stage analysis that included all Early Head Start and control-group families for whom we had complete 14-, 24-, or 36-month Parent Interviews at these sites. (Data presented in the evaluation’s final report demonstrated that attrition of parent interview respondents in the full sample did not produce any bias affecting the impact analyses [ACF 2002b].) To do this, we had to find a child care quality variable that would enable us to include every family in the sample. We created a dichotomous variable equal to 1 if the child was in a good-quality child care center, and 0 if in a lower-quality center or not in child care at all. Children whose classroom received a rating of 5.0 or above on the ITERS (at 14 and 24 months of age) or the ECERS-R (at 36 months) at the time of the observation received a score of 1; if not, they were given a zero.9 Similar cutoffs were set for child-adult ratios and the C-COS variables (which are described when the findings for each of those instruments are described). Results are presented as the difference between the percentage of children in the program and control groups who experienced good-quality care. We recognize that the impact estimates from Stage 1 represent the joint effects of impacts on the use of any center child care and impacts on the use of good-quality center care among those who used care (and for whom observations were completed). Additional analyses were needed to begin to address the question of differential levels of quality in the program and control groups.
Stage 2 Analysis. In the second stage we examined differences between the program and control groups in the levels of quality among those families who used center care. Although the differences in quality levels cannot be considered true program impacts (because they are based on potentially nonrandom subsets of the program and control groups), the results are indicative of differences in quality of Early Head Start centers and the centers that were available to control-group children in the community.
2. Early Head Start’s Impact on the Percentage of Families in Good-Quality Center Child Care—Global Measures
Our first analysis examined impacts across the eight sites at 14 and 24 months and nine sites at 36 months for which we had sufficiently large samples sizes and response rates; these include both center-based and mixed-approach program sites. At all three ages, Early Head Start had a large impact on the percentage of children who were in good quality center care at least 10 hours a week (Figure III.9). At 14 and 24 months of age, Early Head Start children were almost three times as likely to experience good quality (ITERS > 5.0) center child care as the control children (23 percent versus 8 percent at 14 months; 34 versus 12 percent at 24 months). The program-control difference narrowed slightly at 36 months, but more than 33 percent of Early Head Start children were in good-quality care when they were about 36 months old (> 5.0 on the ECERS-R), a percentage that was significantly greater than the 21 percent of control-group children.
[D] |
SOURCE: Based on responses to Parent Interviews and observations on children in “eligible” care arrangements, defined as care that occurs for at least 10 hours per week outside a child’s home, or by a nonrelative in the child’s home. Only one arrangement per child was observed. NOTE: High quality is defined as 5.0 or higher on the ITERS and ECERS-R. The sample includes all children at four center-based and four mixed-approach sites at all three time periods and an additional mixed-approach site at 36 months. Children in the same location at 14 months scheduled to be observed within three months of each other were assigned the same classroom characteristics. The probability of a high ITERS or ECERS-R score was imputed for children in care but not observed. ***Difference is statistically significant at the .01 level. |
Early Head Start impacts on the percentage of children in good-quality centers were greater within the sites at which the programs were center-based and somewhat smaller within the mixed-approach sites (Figures III.10 and III.11). At the four center-based sites, the percentage in good quality ranged from 26 to 37 percent of the sample, whereas only 9 to 16 percent of control-group children were in care that scored this high, a statistically significant difference at each age level. The large impact on the percentage in quality center care at mixed-approach sites was significant at 14 and 24 months, but only marginally significant at age 3 (31 versus 24 percent; Figure III.11).
[D] |
SOURCE: Based on responses to Parent Interviews and observations on children in “eligible” care arrangements, defined as care that occurs for at least 10 hours per week outside a child’s home, or by a nonrelative in the child’s home. Only one arrangement per child was observed. NOTE: High quality is defined as 5.0 or higher on the ITERS and ECERS-R. The sample includes all children at four center-based sites at all three time periods. Children in the same location at 14 months scheduled to be observed within three months of each other were assigned the same classroom characteristics. The probability of a high ITERS or ECERS-R score was imputed for children in care but not observed. ***Difference is statistically significant at the .01 level. |
[D] |
SOURCE: Based on responses to Parent Interviews and observations on children in “eligible” care arrangements, defined as care that occurs for at least 10 hours per week outside a child’s home, or by a nonrelative in the child’s home. Only one arrangement per child was observed. NOTE: High quality is defined as 5.0 or higher on the ITERS and ECERS-R. The sample includes all children at the four mixed-approach sites at all three time periods and an additional site at 36 months. Children in the same location at 14 months scheduled to be observed within three months of each other were assigned the same classroom characteristics. The probability of a high ITERS or ECERS-R score was imputed for children in care but not observed. *Difference is statistically significant at the .10 percent
level. |
The Early Head Start program’s impact on the percentage of children in centers with child-adult ratios that met the performance standards was dramatic. At the four center-based sites at 14 months of age, Early Head Start children were more than twice as likely as control children (72 versus 29 percent) to be in classrooms with ratios of 4 to 1 or better (Figure III.12). At 24 and 36 months, four times as many Early Head Start as control children were in such classrooms. The impacts are somewhat less within the mixed-approach programs, but still demonstrate very substantial impacts that the program had on the percentage of families whose children are in good-quality center child care at all three ages (Figure III.13).
[D] |
SOURCE: Based on responses to Parent Interviews and observations on children in “eligible” care arrangements, defined as care that occurs for at least 10 hours per week outside a child’s home, or by a nonrelative in the child’s home. Only one arrangement per child was observed. NOTE: Child-adult ratios of 4.0 or lower meet Head Start Program Performance Standards for infants and toddlers. We use the same ratio for 36-months, although higher ratios meet the performance standard for children older than 36 months. The sample includes all children at the four center-based sites. Children in the same location at 14 months scheduled to be observed within three months of each other were assigned the same classroom characteristics. The probability of a child-adult ratio of 4.0 or lower was imputed for children in care but not observed. ***Difference is statistically significant at the .01 level. |
[D] |
SOURCE: Based on responses to Parent Interviews and observations on children in “eligible” care arrangements, defined as care that occurs for at least 10 hours per week outside a child’s home, or by a nonrelative in the child’s home. Only one arrangement per child was observed. NOTE: Child-adult ratios of 4.0 or lower meet Head Start Program Performance Standards for infants and toddlers. We use the same ratio for 36-months, although higher ratios meet the performance standard for children older than 36 months. The sample includes all children at the four center-based sites. Children in the same location at 14 months scheduled to be observed within three months of each other were assigned the same classroom characteristics. The probability of a child-adult ratio of 4.0 or lower was imputed for children in care but not observed. ***Difference is statistically significant at the .01 level. |
3. Early Head Start’s Impact on the Percentage of Families in Good-Quality Center Child Care—Child-Caregiver Interactions (C-COS Scores)
Because no established literature is available for setting “good quality” cutoff scores for the C-COS observations, we used scores that represent approximately the top quarter of the distribution of scores as a cutoff for the Stage 1 impact analyses. We set the cutoff for any talk at 34 incidents, for caregiver responding at 11 incidents, and for caregiver initiated talk at 28 incidents.
Early Head Start programs in center-based sites had a large and significant impact on total caregiver talk with child (Figure III.14a), caregiver responsiveness to child (Figure III.14b), and the caregiver’s initiation of talk (Figure III.14c) at both 24 and 36 months of age. For example, when children were 2 years old, 43 percent of Early Head Start children experienced caregiver talk above the cutoff (34 or more incidents), compared to only 19 percent of control-group children at the center-based sites. The impacts at 24 months were consistently larger and more robust than when children were 36 months old. The percentage of Early Head Start children experiencing a high level of any caregiver talk dropped to 26 percent at 36 months, while the percentage of control-group children stayed about the same (20 percent). Early Head Start had no impact on incidents of negative behavior; but, as we saw earlier (Figure III.8), very few negative behaviors were observed overall.
[D] |
SOURCE: Based on responses to Parent Interviews and observations on children in “eligible” care arrangements, defined as care that occurs for at least 10 hours per week outside a child’s home, or by a nonrelative in the child’s home. Only one arrangement per child was observed. NOTES: The sample includes all families at four center-based sites. Possible range for number of incidents of caregiver talk is 0-60 over a 2-hour observation period. The probability of high levels of caregiver talk was imputed for children in care, but not observed. |
In the centers operated by the mixed-approach sites included in this analysis, Early Head Start programs’ impact on child-caregiver interactions were smaller but followed a similar pattern with two exceptions (Figure III.15). At 36 months of age, the percentage of Early Head Start children experiencing high levels of any caregiver talk (12 versus 14 percent) and high levels of caregiver initiating talk with the child (8 versus 12 percent) was lower than that for the control group. Still, at age 2, the percentage of Early Head Start children who were in arrangements with caregivers who displayed high levels of any talking was more than twice as great as that for control children (28 versus 11 percent). Similarly, 2-year-old Early Head Start children experienced three times as much caregiver responsiveness in their classrooms as their control-group counterparts; and the program-control difference at 36 months, though smaller, was still significant.
[D] |
SOURCE: Based on responses to Parent Interviews and observations on children in “eligible” care arrangements, defined as care that occurs for at least 10 hours per week outside a child’s home, or by a nonrelative in the child’s home. Only one arrangement per child was observed. NOTES: The sample includes all families at four mixed-approach sites in both time periods and an additional site at 36 months. Possible range for number of incidents of caregiver talk is 0-60 over a 2-hour observation period. The probability of high levels of caregiver talk was imputed for children in care, but not observed. |
4. Summary of Program Impacts on Percentages of Children Receiving Good-Quality Center Child Care
Because of their participation in a center-based or mixed-approach Early Head Start program, infants and toddlers in low-income families that we studied experienced significantly higher-quality center child care. These analyses of the impacts of Early Head Start on the percentages of children experiencing good-quality care show strong effects of the program. This is true for all measures of quality used—structural and process quality, and both global quality and specific caregiver-child interaction measures. Impacts were particularly strong for families enrolled in the four center-based programs.
5. Differences in the Average Quality of Care for Children Observed in Care
Because of the strong impacts on the percentage of children receiving good-quality care, the quality data can also be used to understand the impacts by analyzing the relationship between program participation and the levels of quality. This second stage in our analysis is not a pure impact analysis due to the selection factors described earlier. However, because we know that Early Head Start has strong impacts for the full sample, if we also find differences in levels of quality when including only those we observed in the analysis, we are in a stronger position to argue that it was participation in the Early Head Start program that is responsible for the program-control differences in observed quality.
Tables III.5 and III.6 show the average classroom quality scores (ITERS, ECERS-R, Arnett, and child-adult ratios) experienced by Early Head Start program and control-group children at sites included in the impact analyses. Classroom quality scores, measured by the ITERS and ECERS-R, were consistently higher for Early Head Start than the control group at center-based sites. Average ITERS/ECERS-R ratings ranged from 4.7 to 4.9 for Early Head Start children. Scores for control-group children rose slightly over time, from an average of 3.9 to 4.1, but were always substantially lower than those experienced by Early Head Start children at all three time periods. The program-control differences represent effect sizes of about .7, .9, and .5 at the three ages. Classroom caregiver quality, as measured by the Arnett scale, was also higher for Early Head Start program children, with ratings of 3.4 and 3.3, which are consistently higher than the 3.0 to 3.2 ratings experienced by control-group children. Child-adult average ratios were also consistently lower (more favorable) for Early Head Start children than control children at the four center-based sites. Early Head Start ratios averaged a low of 2.8 to 1 for 14-month-old children and rose to 5.6 to 1 by 36 months. The ratio for control-group children was 3.9 to 1 at 14 months and rose to 6.8 to 1 by 36 months of age.
| Quality Measure | 14 Months | 24 Months | 36 Months | |||
|---|---|---|---|---|---|---|
| Program | Control | Program | Control | Program | Control | |
| ITERS/ECERS-R | ||||||
| Average (S.D.) | 4.8 (1.0) | 3.9 (1.3)*** | 4.9 (1.0) | 3.8 (1.2)*** | 4.7 (1.0) | 4.1 (1.3)*** |
| Range | 1.9 - 6.8 | 1.8 - 6.5 | 1.7 - 6.6 | 1.9 - 6.3 | 1.2 - 6.8 | 1.1 - 6.9 |
| N | 168 | 52 | 162 | 48 | 153 | 72 |
| Arnett | ||||||
| Average (S.D.) | 3.4 (0.3) | 3.1 (0.5)*** | 3.3 (0.4) | 3.0 (0.6)*** | 3.3 (0.5) | 3.2 (0.6)** |
| Range | 2.4 - 4.0 | 1.5 - 3.9 | 2.0 - 4.0 | 1.8 - 3.8 | 1.3 - 3.9 | 1.8 - 3.9 |
| N | 171 | 50 | 161 | 47 | 150 | 72 |
| Child-Adult Ratios | ||||||
| Average (S.D.) | 2.8 (1.0) | 3.9 (1.7)*** | 3.2 (1.1) | 5.5 (2.6)*** | 5.6 (3.0) | 6.8 (2.7)*** |
| Range | 1.0 - 6.8 | 1.0 - 7.4 | 1.0 - 8.8 | 1.0 - 14.0 | 1.7 - 14.8 | 1.0 - 14.5 |
| N | 162 | 45 | 159 | 47 | 152 | 72 |
| Source: Based on observations of “eligible” care
arrangements, defined as care that occurs for at least 10 hours
per week outside a child’s home, or by a nonrelative in
the child’s home. Only one arrangement per child was observed.
Note: Based on observations at the four center-based sites at all three time periods. Individual observations were not conducted for all children at 14 months. Children in the same locations who were scheduled to be observed within three months of each other were assigned the same classroom characteristics. ** Program-control difference is statistically significant
at the .05 level. |
| Quality Measure | 14 Months | 24 Months | 36 Months | |||
|---|---|---|---|---|---|---|
| Program | Control | Program | Control | Program | Control | |
| ITERS/ECERS-R | ||||||
| Average (S.D.) | 4.7 (1.2) | 3.7 (1.2)*** | 4.9 (1.3) | 4.3 (1.3)** | 5.0 (1.1) | 4.7 (1.2) |
| Range | 1.5 - 6.6 | 1.9 - 6.4 | 1.6 - 6.7 | 2.3 - 6.4 | 2.3 - 6.7 | 1.6 - 6.9 |
| N | 63 | 26 | 67 | 34 | 93 | 72 |
| Arnett | ||||||
| Average (S.D.) | 3.4 (0.5) | 3.0 (0.7)*** | 3.4 (0.7) | 3.4 (0.5) | 3.5 (0.5) | 3.4 (0.4) |
| Range | 1.5 - 4.0 | 1.4 - 3.9 | 1.3 - 4.0 | 2.0 - 4.0 | 1.7 - 4.0 | 2.0 - 4.0 |
| N | 63 | 26 | 67 | 34 | 90 | 72 |
| Child-Adult Ratios | ||||||
| Average (S.D.) | 2.8 (1.6) | 4.4 (1.7)*** | 3.8 (2.0) | 5.7 (2.0)*** | 5.1 (2.3) | 7.3 (2.8)*** |
| Range | 0.8 - 7.7 | 1.0 - 8.7 | 1.1 - 10.3 | 1.7 - 11.2 | 0.8 - 11.4 | 2.6 - 13.8 |
| N | 63 | 26 | 67 | 34 | 89 | 72 |
| Source: Based on observations of “eligible” care
arrangements, defined as care that occurs for at least 10 hours
per week outside a child’s home, or by a nonrelative in
the child’s home. Only one arrangement per child was observed.
Note: Based on observations at four mixed-approach sites at all three time periods and one additional site at 36 months. Individual observations were not conducted for all children at 14 months. Children in the same locations who were scheduled to be observed within three months of each other were assigned the same classroom characteristics. **Program-control difference is statistically significant
at the .05 level. |
Quality was also higher in the classrooms attended by Early Head Start children than control children at the mixed-approach sites (Table III.6). The program-control differences in the ITERS/ECERS-R ratings were not quite as dramatic as in the center-based sites, however. The largest difference occurred at 14 months of age, when Early Head Start children experienced an average of 4.7, while control children experienced classrooms rated at 3.7. The difference at 36 months was no longer statistically significant. Program-control differences in classroom caregiver Arnett ratings were also less pronounced than at the center-based sites. The only statistically significant difference occurred at 14 months, when Early Head Start children experienced higher average quality. Differences in child-adult ratios were somewhat larger at the mixed-approach sites. Early Head Start children experienced ratios of 2.8 to 1, on average, when they were 14 months old, compared to 4.4 to 1 for control-group children. A large difference persisted through to 36 months, when Early Head Start children experienced an average ratio of 5.1 to 1 compared to 7.3 to 1 for control-group children.
Tables III.7 and III.8 show C-COS scores for incidents of caregiver talk and negative child behavior. Although fewer program-control differences were significant than we saw with the global quality measures, almost all the differences are in the expected direction. Program children at the four center-based sites experienced more incidents of any caregiver talk (33.4) and caregiver responding to the child (8.1) than control-group children (30.6 and 5.9) when they were 24 months old. The same pattern was found at the four mixed-approach sites at 24 months, but the program-control differences were a bit larger, and more were statistically significant (Table III.8). Early Head Start program children experienced an average of 33.8 incidents of any caregiver talk, while control-group children experienced 27.6 incidents. Similarly, Early Head Start caregivers in the mixed-approach sites responded to the focus child 11.9 times, compared with 6.4 incidents of caregiver responding in the control group. None of the program-control differences in the child-caregiver interaction variables was significant when the children were 3 years old in either center-based or mixed-approach sites.
| 24 Months | 36 Months | |||
|---|---|---|---|---|
| Program | Control | Program | Control | |
| Incidents of Any Caregiver Talk |
34.4 (12.4) | 30.6 (13.3)** | 30.2 (12.4) | 27.8 (14.4) |
| Range | 10 - 60 | 9 - 60 | 5 - 59 | 0 - 60 |
| N | 161 | 58 | 154 | 79 |
| Incidents
of Caregiver Responding to Child |
8.1 (8.8) | 5.9 (7.2)* | 6.9 (5.9) | 6.7 (5.7) |
| Range | 0 - 48 | 0 - 30 | 0 - 28 | 0 - 23 |
| N | 161 | 58 | 154 | 79 |
| Incidents
of Caregiver Initiating Talk with Child |
27.0 (11.9) | 25.3 (13.6) | 23.8 (11.4) | 21.4 (13.9) |
| Range | 0 - 55 | 0 - 59 | 3 - 57 | 0 - 55 |
| N | 161 | 58 | 154 | 79 |
| Incidents
of Negative Behavior |
5.7 (5.7) | 5.1 (6.0) | 4.4 (4.5) | 4.5 (4.3) |
| Range | 0 - 32 | 0 - 35 | 0 - 28 | 0 - 23 |
| N | 161 | 58 | 154 | 79 |
| Source: Based on observations of “eligible” care
arrangements, defined as care that occurs for at least 10 hours
per week outside a child’s home, or by a nonrelative in
the child’s home. Only one arrangement per child was observed.
Note: Based on observations at the four center-based sites at 24 and 36 months. Possible range of the number of incidents is 0 to 60 over a 2-hour observation period. *Program-control difference is statistically significant
at the .10 level. |
| 24 Months | 36 Months | |||
|---|---|---|---|---|
| Program | Control | Program | Control | |
| Incidents of Any Caregiver Talk |
33.8 (12.4) | 27.6 (13.2)*** | 26.7 (12.2) | 28.8 (13.3) |
| Range | 6 - 59 | 6 - 59 | 5 - 60 | 5 - 59 |
| N | 90 | 47 | 93 | 73 |
| Incidents of
Caregiver Responding to Child |
11.9 (13.0) | 6.4 (6.3)*** | 8.7 (8.2) | 8.4 (7.7) |
| Range | 0 - 58 | 0 - 31 | 0 - 38 | 0 - 35 |
| N | 90 | 47 | 93 | 73 |
| Incidents
of Caregiver Initiating Talk with Child |
22.8 (11.3) | 21.9 (9.4) | 19.1 (10.3) | 21.4 (11.3) |
| Range | 0 - 58 | 5 - 42 | 0 - 55 | 0 - 56 |
| N | 90 | 47 | 93 | 73 |
| Incidents of Negative Behavior |
4.5 (5.8) | 3.4 (4.5) | 3.0 (4.2) | 3.0 (4.4) |
| Range | 0 - 36 | 0 - 20 | 0 - 17 | 0 - 21 |
| N | 90 | 47 | 93 | 73 |
Source: Based on observations of “eligible” care arrangements, defined as care that occurs for at least 10 hours per week outside a child’s home, or by a nonrelative in the child’s home. Only one arrangement per child was observed. Note: Based on observations at four mixed-approach sites at 24 months and five sites at 36 months. Possible range of number of incidents is 0 to 60 over a 2-hour observation period. ***Program-control difference is statistically significant at the .01 level. |
6. Summary of Program-Control Differences in Quality of Center Care Received
Because of the study’s experimental design and the analytic approach taken, we can conclude that it is highly likely that the center-based and mixed-approach Early Head Start programs included in this analysis succeeded in ensuring that their children received significantly higher levels of quality center care than control children received. In the four center-based sites, average classroom global quality was consistently higher for the centers attended by Early Head Start children than for the centers control children attended. Early Head Start children experienced better child-adult ratios than their control counterparts, and their caregivers were rated more favorably. Early Head Start children in sites with mixed-approach programs also benefited from being in the program, but the program-control differences in levels of classroom quality were smaller than those in the center-based sites.
E. RELATIONSHIPS BETWEEN CHILD CARE QUALITY AND INTENSITY AND CHILD OUTCOMES AMONG EARLY HEAD START CHILDREN
We examined whether associations between child care quality and the intensity of child care use among Early Head Start children were related to three key child outcomes at 24 and 36 months of age. These analyses included a different sample than the impact analyses just described, as we included available observational data obtained from the settings of all children in center care across all types of Early Head Start programs (center-based, home-based, and mixed) and included children who had been observed at least once. Child care quality was measured by the ITERS or ECERS-R and by child-adult ratio. Intensity was measured by average hours in center child care.
Child outcomes were assessed when children were 24 months old, using the Bayley Scales of Infant Development Mental Development Index (BSID-MDI), the MacArthur Communicative Development Inventory (CDI) language production scale, and the Child Behavior Checklist (CBCL) aggressive behavior scale (see ACYF 2001, Chapter V, for details). At 36 months, children were assessed on the BSID-MDI, the Peabody Picture-Vocabulary Test-Third Edition (PPVT-III), and the CBCL aggressive behavior scale (see ACF 2002b, Chapter V, for details).
Using Ordinary Least Squares (OLS) regression analyses, we examined how indices of child care quality (ITERS or ECERS-R and child-adult ratio), and intensity of child care (average hours in care) were related to child outcomes at 24 and 36 months. Mean quality and intensity scores at 14 and 24 months were used to predict 24-month outcomes; mean quality and intensity scores at 14, 24, and 36 months were used to predict 36-month outcomes. All regression analyses controlled for child gender, child age at time of assessment, maternal race/ethnicity, mother’s education and marital status, whether mother was a teenager (under 19 years of age) at the time of the child’s birth, and whether the site was urban. Appendix Table A.6 presents descriptive statistics for the child outcome measures and child care quality and use measures included in these analyses. Results of the regression analyses are shown in Appendix Table A.7.
Among the Early Head Start children who attended child care centers, those in higher-quality center care showed enhanced developmental outcomes. Mean child care quality over time predicted higher scores on the 24-month Bayley MDI and 36-month PPVT-III. Mean child-adult ratio over time did not significantly predict child outcomes. Mean hours in center care over time predicted higher scores on the 24- and 36-month Bayley and the 36-month PPVT-III. Neither the quality nor the intensity of child care predicted child aggressive behavior at 24 or 36 months.
Three interactions were tested separately for each outcome at each age level: (1) quality by hours in care, (2) quality by child-adult ratio, and (3) hours in care by child-adult ratio. Of the interactions tested, only one was significant: hours in care by ratio predicted 24-month aggressive behavior problems and the 36-month Bayley. For children in centers with higher child-adult ratios (that is, with less-favorable ratios), more hours in care was related to more behavior problems at 24 months. For children in centers with more-favorable child-adult ratios, more hours in care was not significantly related to behavior problems at 24 months.
Consistent with previous research, these findings demonstrate that among this sample of Early Head Start children, the quality of the child care centers they attend was positively associated with children’s cognitive and language development. Further, (1) spending more time in center-based child care was associated with higher cognitive scores at 24 and 36 months and higher language scores at 36 months; and (2) more time in child care was related to increased behavior problems only if children were in settings with worse child-adult ratios (and only at 24 months).
As with all studies of such relationships, we must interpret the associations with some caution, in that selection factors could at least partially account for the relationships between quality and child outcomes. In this analysis, however, it is reasonable to expect selection bias to be less an issue than in most child care studies. All children in the sample included here were in families who applied for, and were enrolled in, Early Head Start programs. A substantial portion of their child care settings either were provided by the Early Head Start program or were arrangements to which the program referred families. Thus, it is less likely that selection factors affected which classrooms children attended.10 Follow-up analyses will consider such issues as child care mediating the impact of Early Head Start on child outcomes in the full sample of Early Head Start and control-group children, corrections for potential selection factors, and quality of child care in home and family-based settings.
6 As reported in the Early Head Start implementation study (ACF 2002c), designation of programs as mixed-approach indicated that they provided some combination of center- and home-based services. Based on the 1997 site visits, seven programs were designated as mixed-approach; four of these had sufficient samples of center observations to be included in the impact analyses. One additional mixed-approach program subsequently added a center and had a large enough observation sample to allow us to add a fifth site to our analyses during the period in which the 36-month observations were conducted. (back)
7 Even though the analyses we conducted to address this question were based on a subset of the research sites, the sites included in the impact analyses represent between 75 and 85 percent of all center care that children in our sample experienced. (back)
8 Appendix Tables A.4 and A.5 show the baseline characteristics of program and control families at the sites where we conducted the impact analyses. The groups differ on a small number of demographic characteristics but are highly similar. Nevertheless, unmeasured differences due to selection factors are still possible (Duncan, Magnuson, and Ludwig in press). These factors relate to at least three different circumstances: (1) family characteristics, such as the need for and motivation to seek child care for their child; (2) child care setting characteristics, such as the provider's willingness to allow observers into the center or home, and their stability, as some settings were no longer operating when observers arrived; and (3) the researchers' ability to complete the observations within a reasonable time following the birthday-related interview. (back)
9 All families in the four center-based sites and four mixed-approach sites who completed a 14- or 24-month Parent Interview-and all families in the four center-based and five mixed-approach sites who completed the 36-month Parent Interview-are included in the analysis. Thus, these analyses provide estimates of the impacts for eligible applicants. For those children who were in an eligible child care setting but not observed, we imputed the mean value of receiving good-quality care among the children in that group (by program approach and program-control status) who were observed. In essence, this procedure assumes that the percentage of good-quality care settings was the same in the eligible settings we did not observe as in the settings we did observe. (back)
10 In addition, the 24- and 36-month child care quality observations were conducted at approximately the same time as the 24- and 36-month child outcomes were assessed, making it difficult to draw a casual inference between the quality measures and child outcomes. (back)
| Table of Contents | Previous | Next |