Table of Contents | Previous | Next |
D.9 ANALYSES OF PARENTING OUTCOMES AT 24 MONTHS AS MEDIATORS OF CHILD OUTCOMES AT 36 MONTHS
Early Head Start programs seek to influence children’s well-being by providing center-based and/or home-based child development services directly to children and by providing services to support and inform parents. Programs thus intervene to enhance children’s development both through direct services to the child and indirectly, through changes in parenting practices and behavior. Therefore, we would expect that changes in parenting behavior brought about by Early Head Start would, in concert with direct services from the program, help influence children’s outcomes in the future.
We conducted analyses to explore the relationships between Early Head Start impacts on parenting outcomes at 24 months and program impacts on children a year later. In this appendix, we describe the models and summarize the results.
A. MODELS OF PARENTING INFLUENCES ON CHILD OUTCOMES
At 36 months, Early Head Start had favorable impacts on children’s cognitive and language development and on some aspects of social-emotional behavior. In particular, 3-year-old children enrolled in Early Head Start had higher Bayley MDI scores, higher PPVT-III scores, higher levels of engagement with the parent and sustained attention with objects during semistructured play; and lower levels of negativity toward the parent during semistructured play and lower levels of aggressive behavior.
Theories of child development suggest that these favorable outcomes for children may be partly attributable to the program’s impacts on parents at an early point. For example, previous research has shown that children’s language development is related to the amount and variety of language they are exposed to by caregivers, so we would expect that earlier impacts on support for the cognitive, language, and literacy environment of the home and regular reading to the child could contribute to children’s language gains later on. Similarly, previous research indicates that children’s aggressive behavior is related to experiences of punitive parenting practices, so we would expect that the program’s success at reducing the incidence of physical punishment at 24 months could contribute to reductions in aggressive behavior later on.
At 24 months, Early Head Start had favorable impacts on several important aspects of parenting, including emotionally supportive parenting, support for language and cognitive development, parenting knowledge, insensitivity, and punitive behavior. To explore whether the pattern of Early Head Start program impacts on children at 36 months is consistent with developmental theory and program theories of change that suggest a relationship between earlier impacts on parenting behavior and future impacts on children, we conducted analyses of the association between impacts on child outcomes at 36 months and impacts on related parenting behavior at 24 months. In choosing parenting mediators for each child outcome, we have tried to identify one parenting mediator to represent a distinct aspect of parenting behavior such as emotionally supportive parenting, rather than using several variables from a single domain that may provide overlapping information.
For child cognitive and language impacts, we estimated their association with parenting practices that theory suggests would promote cognitive and language development. Thus, we included in these models supportiveness during semistructured play at 24 months, which is based on observations of parent-child play and indicates the extent to which parents responded to the child’s cues, showed sensitivity and positive regard for the child, and attempted to extend the play by providing language stimulation and learning opportunities. We included the support for cognitive, language, and literacy environment subscale of the HOME at 24 months because it measures materials in the child’s environment and parenting behavior with the child that provide cognitive and language stimulation (for example, the availability of a variety of toys to simulate development and frequent reading to the child). We also included whether the parent reads to the child every day at 24 months. The three variables give us an observer’s rating of the parent’s responsiveness and cognitive stimulation of the child, a measure of the stimulating materials in the child’s environment, and a measure of the parent’s reading within the structure of the day.
For positive aspects of children’s social-emotional behavior during semistructured play, we estimated their association with parenting practices that theory suggests would strengthen the child’s engagement of the parent and curiosity and attentiveness to a task (sustained attention). Thus, the model for engagement of parent includes variables measuring warm and supportive behavior, cognitive stimulation, and insensitivity, which together are expected to influence the child’s positive relationship with the parent. We included warm sensitivity during parent-child semistructured play at 24 months, or the extent to which the parent responded to the child’s cues and showed sensitivity and positive regard for the child.14 It also includes the emotional responsivity subscale of the HOME at 24 months, which measures the parent’s responsiveness to the child based on observations by the home interviewer. We included the support for cognitive, language, and literacy environment subscale of the HOME at 24 months because it measures parent activities to stimulate learning in part through play and reading, which are expected to strengthen the parent-child relationship. We included detachment during semistructured play at 24 months, or the extent to which the parent is inattentive to the child, inconsistently attentive, or interacts with the child in an indifferent manner, because detached parenting may dampen the child’s interest in trying to engage the parent in play.
The model for children’s sustained attention toward objects includes a similar set of parenting variables, but with somewhat more emphasis on cognitive stimulation along with emotional support. Thus, in addition to the HOME support for cognitive, language, and literacy environment subscale at 24 months, we also include supportiveness during semistructured play at 24 months. We include knowledge of infant development because parents who are more knowledgeable are expected to provide the emotional support and cognitive stimulation that can enhance the child’s curiosity and attention to a task. We include parental distress because parents who are distressed in their parenting role may be less able to provide the emotional support and cognitive stimulation needed to enhance children’s attention to play tasks.
For negative aspects of children’s social-emotional development, we included emotionally supportive parenting behavior, punitive behavior, parental distress, and insensitivity (for child negativity) and structuring the day (for aggressive behavior). For negativity toward the parent in play, we included warm sensitivity during semistructured play at 24 months, or the parent’s responsiveness and warmth toward the child during semistructured play, because we expect children to show less negativity toward a parent who is more warm and sensitive during play. We included physical punishment in the past week at 24 months because we expect use of physical punishment to increase child negativity toward the parent. We included parental distress because stress and depression in the parenting role is likely to be detrimental to the parent-child relationship and thus increase child negativity toward the parent. We included intrusiveness during semistructured play at 24 months, or the extent to which the parent controlled the pace and direction of play, grabbed toys from the child, or did not take turns or consider the child’s perspective in play. Such parenting behavior could provoke child negativity toward the parent.
Our model for child aggressive behavior includes warm sensitivity, physical punishment in the past week, and parental distress, all measured when the child was 24 months old, because lower levels of emotional support, and higher levels of punitiveness or stress and depression in the parenting role are expected to increase the child’s level of aggressive behavior. In addition, we included whether the child had a regular bedtime at 24 months because parents who keep the child on a bedtime schedule may help ensure that the child feels rested and secure, which may tend to reduce aggressive behavior.
B. APPROACH TO ESTIMATION
The approach to the mediated analysis can be thought of as having three stages. In the first stage, the child outcome at 36 months is regressed on the 24-month parenting mediators and other explanatory variables that were not changed by the program, such as the parent’s age, ethnicity, and other characteristics at enrollment (moderators). In the second stage, the regression coefficient on each mediator is multiplied by the Early Head Start impact on that mediator. These products are what we would expect the program impacts on the child outcome to be on the basis of the estimated relationship between the parenting mediators and the child outcome (in other words, what the program impact on the child is expected to be if all of the program’s influence came through the earlier impacts on parenting). We label these products the “implied” impacts. Finally, the implied impacts are compared to the actual impact on the child outcome. These results indicate the extent to which impacts on the child outcome variable can be partitioned into impacts attributable to each parenting mediator.
Formally, we conducted the mediated analysis by first estimating the following regression model:
where IMi is the impact on the mediator.
In this formulation, the parameter, yi, represents the marginal effect of a particular mediator) on the outcome measure, holding constant the effects of the other mediators and moderators. For example, it represents the change in the longer-term outcome measure if the value of the mediator were increased by one unit, all else equal. Thus, the impact of Early Head Start on the longer-term outcome in equation (1) can be decomposed into two parts: one due to the mediators (the “implied” impacts) and the second due to residual factors (represented by the parameter ai). Our analysis focuses on the part due to the mediators and the extent to which these implied impacts account for the impact on the longer-term outcome.
While the mediated analyses allow us to estimate relationships among variables that developmental theory predicts are related, these models are not structural models, and therefore cannot measure causal relations between parent and child measures. Structural analyses of parent behavior and child development are very difficult to conduct because of the complex relationships among various measures of the parent’s mental health and parenting behavior and simultaneity problems that lead to bias in the estimated relationships between parent and child outcomes. Therefore, the goal of these analyses is more modest than establishing a measured causal link between parenting impacts and child impacts. Instead, the goal is best viewed as establishing whether there are associations between the parenting and child impacts that are consistent with theories of change. We cannot measure the individual parameters reliably, but the patterns of association are likely to indicate that causal relations exists. In particular, these analyses are designed to provide some plausible support for or raise questions about programs’ theories of change that suggest programs have an impact on children through earlier impacts on parenting behavior.
C. RESULTS OF THE MEDIATED ANALYSES FOR THE FULL SAMPLE
Table D.9A presents the results of estimating the models of children’s cognitive and language development. The first column lists the 24-month parent variables entered into the model as mediators of the 36-month child impact listed in the column heading. The second column shows the estimated relationships between each of the parenting outcomes in the model and the child cognitive outcome; and the third column indicates whether this association is significantly different from zero. For the fourth column, we use the estimated relationships between the parenting outcomes and child outcomes and the impacts on parenting and child outcomes to compute the percentage of the impact on the child outcome that is associated with the impact on the parenting outcome.
| 24-Month Mediator | Estimated Effect of Parenting Outcomes
on Bayley MDI |
Significance Level | Percentage of Impact on Bayley MDI Associated with Mediator |
|---|---|---|---|
| Supportiveness: Semistructured Play | 2.32 | *** | 9.8 |
| HOME Support of Cognitive,
Language, and Literacy Environment |
1.23 | *** | 14 |
| Read Daily | 1.16 | * | 3.2 |
| Total | 27.1 |
| 24-Month Mediator | Estimated Effect of Parenting Outcomes on PPVT-III | Significance Level | Percentage of Impact on Bayley PPVT-III Associated with Mediator |
|---|---|---|---|
| Supportiveness: Semistructured Play | 2.37 | *** | 6.9 |
| HOME Support of Cognitive,
Language, and Literacy Environment |
1.45 | *** | 11.4 |
| Read Daily | 1.46 | n.s. | 2.8 |
| Total | 21.1 |
| *Significantly different from zero at the
.10 level, two-tailed test. **Significantly different from zero at the .05 level, two-tailed test. ***Significantly different from zero at the .01 level, two-tailed test. |
These analyses indicate that children’s scores on the Bayley MDI at 36 months are related to higher levels of parent supportiveness in semistructured play, greater support for cognitive and language development, and daily reading at 24 months. In total, the estimates suggest support for the idea that some of the Early Head Start impact on children’s cognitive development could have occurred because of the program’s impacts on parents’ sensitivity and cognitive stimulation in interactions with the child, and their support in the home for the child’s cognitive and language development. Estimates also suggest a positive relationship between 36-month PPVT III scores and parent supportiveness in play and support for cognitive and language development, but not daily reading. In total, these estimates suggest that part of the Early Head Start impact on children’s receptive language ability at 3 years of age could have emerged because of earlier impacts on the parent’s sensitivity, cognitive stimulation, and support for the child’s language development across a range of parenting situations (during play, through regular daily reading, and during everyday interactions in the home).
Table D.9B displays the results of estimating the models of parenting behavior and positive aspects of children’s social-emotional behavior at 36 months. The estimates indicate that children’s engagement of the parent during semistructured play is positively related to the parent’s warm sensitivity during observed semistructured play a year earlier; parent’s emotional responsivity observed a year earlier; and the level of language and cognitive stimulation available in the home environment a year earlier. The relationship between child engagement and parent detachment during play a year earlier was not significantly different from zero. In total, the estimates suggest that Early Head Start positive impacts on the child’s engagement of the parent in semistructured play at 36 months are consistent with earlier positive program impacts on the parent’s sensitivity during play, responsiveness to the child, and cognitive stimulation and support for language development in the home.
| 24-Month Mediator | Estimated Effect of Parenting Outcomes on Engagement | Significance Level | Percentage of Impact on Engagement Associated with Mediator |
|---|---|---|---|
| Warm Sensitivity: Semistructured Play | 0.21 | *** | 8.9 |
| HOME Emotional Responsivity | 0.05 | *** | 3.7 |
| HOME Support of Cognitive, Language, and Literacy Environment | 0.05 | *** | 6.1 |
| Detachment: Semistructured Play | -0.01 | n.s. | |
| Total | 19.3 |
| 24-Month Mediator | Estimated Effect of Parenting Outcomes on Sustained Attention | Significance Level | Percentage of Impact on Sustained Attention Associated with Mediator |
|---|---|---|---|
| Supportiveness: Semistructured Play | 0.15 | *** | 8.5 |
| HOME Support of Cognitive, Language, and Literacy Environment | 0.04 | ** | 5.9 |
| Knowledge of Infant Development | 0.16 | ** | 6.1 |
| PSI: Parental Distress | -0.006 | ** | 4.5 |
| Total | 25.0 |
| *Significantly different from zero at the
.10 level, two-tailed test. **Significantly different from zero at the .05 level, two-tailed test. ***Significantly different from zero at the .01 level, two-tailed test. |
The results of estimating the model of child sustained attention to objects during semistructured play at 36 months indicate that the child’s attention and focus on play is positively related to parent’s sensitivity and cognitive stimulation during semistructured play a year earlier; support for cognitive development and language stimulation in the home environment in the previous year; and the parent’s knowledge of child development measured at 24 months. Sustained attention toward objects during play at 36 months is negatively related to parental distress measured in the previous year. In total, the estimates suggest that part of the positive impact on children’s sustained attention to objects during semistructured play at 36 months could have come about because of earlier favorable program impacts on parent supportiveness in semistructured play; cognitive stimulation and language support in the home environment, and knowledge of child development; and through reductions in parental distress.
Table D.9C shows the results of estimating the models of parenting behavior and negative aspects of children’s social-emotional behavior at 36 months. The estimates indicate that children’s negativity toward the parent in semistructured play at 36 months is inversely related to parents’ warm sensitivity during semistructured play observed in the previous year; and positively related to levels of parental distress and intrusive behavior during semistructured play measured in the previous year. The relationship between child negativity at 36 months and the parent’s use of physical punishment a year earlier is not significantly different from zero. In total, the estimates suggest that part of the reduction in levels of child negativity toward the parent during semistructured play that came about through Early Head Start participation might be associated with Early Head Start-induced increases in parent warmth and sensitivity during play and reductions in parental distress and intrusiveness during play measured one year earlier.
| 24-Month Mediator | Estimated Effect of Parenting Outcomes on Child Negativity | Significance Level | Percentage of Impact on Negativity Associated with Mediator |
|---|---|---|---|
| Warm Sensitivity | -0.05 | *** | 6.0 |
| Physical Punishment Last Week | 0.02 | n.s. | 1.7 |
| PSI: Parental Distress | 0.004 | *** | 7.0 |
|
Intrusiveness: Semistructured Play |
0.06 | *** | 3.6 |
| Total | 18.2 |
| 24-Month Mediator |
Estimated Effect of Parenting Outcomes on Aggressive Behavior |
Significance Level |
Percentage of Impact on Aggressive Behavior Associated with Mediator |
|---|---|---|---|
| Warm Sensitivity: Semistructured Play | -0.46 | *** | 7.5 |
| Physical Punishment Last Week | 1.52 | *** | 17.3 |
| PSI: Parental Distress | 0.19 | *** | 44.6 |
| Regular Bedtime | -0.32 | n.s. | 2.7 |
| Total | 72.0 |
| *Significantly different from zero at the
.10 level, two-tailed test. **Significantly different from zero at the .05 level, two-tailed test. ***Significantly different from zero at the .01 level, two-tailed test. |
The estimates of the model of children’s aggression at 3 years of age and parenting behavior in the previous year indicate that children’s aggression is inversely related to the parent’s warm sensitivity during semistructured play and positively related to the use of physical punishment and levels of parental distress measured in the previous year. The relationship between aggression and the parent’s setting a regular bedtime for the child is not significantly different from zero. In total, the estimates indicate that part of the Early Head Start impact reducing levels of aggression in 3-year-old children is partly attributable to the program’s positive impact on parents’ warm sensitivity toward the child during play and to the program’s impact reducing the incidence of physical punishment in the previous year. The relationship between children’s aggressive behavior and earlier levels of parental distress appears fairly large, but the relationship may be overstated because of measurement error. Part of the correlation may occur because distressed parents may view their children’s behavior more negatively than an outside observer would.
To check the robustness of these findings, we also substituted an alternative measure of parent reading: reading at bedtime. The alternative variable, reading at bedtime, indicates that the parent followed a bedtime routine most days in the past week and volunteered that it included reading. We found that the proportion of the impact on the Bayley MDI and PPVT-III at 36 months that is associated with bedtime reading is very similar to the proportion associated with daily reading, and the overall proportion of the impact associated with all of the parenting mediators in each of the models changes by only about 3 percentage points.
In summary, the estimates of models relating children’s behavior at 36 months to parenting behavior measured a year earlier in the full sample suggest some support for the theory that part of the Early Head Start program impact on children could have come about because of earlier favorable changes in parenting behavior. The estimates of the relationships between parenting behavior and children’s outcomes and the Early Head Start program impacts on these outcomes are consistent with the theory, although the models we have estimated are not structural and therefore cannot establish a causal link between the parenting impacts and impacts on children.
D. MODELS BY PROGRAM APPROACH
Early Head Start programs that chose different approaches to service delivery typically also had different theories of change regarding how the program would intervene in children’s lives. Center-based programs, which offered center-based child development services as well as parent education, expected changes to occur mainly through the direct services, with only a small impact of the program coming through changes in parenting. Home-based programs focused child development services directly on the child and on the parent, because these programs expected the parent to enhance the effects of the program on the child. Mixed programs, which blended center-based and home-based services in different patterns, would likely fall in the middle in terms of the expected program effects on the child that would be mediated by the parent.
To explore whether the impacts we have found for parenting measures at 24 months and child outcome measures at 36 months are consistent with the program-specific theories of change, we estimated mediated models by program approach that were similar to those estimated for the full sample. When a particular child outcome was not very different for program and control groups within a program type, we did not run a model predicting parenting effects on that impact. Although parenting variables likely do affect the child outcome in that case, it did not make sense to estimate the model because Early Head Start had no impact on that outcome. We also did not estimate a model if the impacts on parenting outcomes were not very different from zero at 24 months, because once again, while parenting behavior likely has an influence on particular child outcomes, parenting could not have been an important mediator if the program impacts on parenting were very small or zero. In some cases, when a particular parenting outcome was not changed by Early Head Start at 24 months, we substituted a similar parenting outcome from the same domain for which the program did have an impact so that we could estimate whether there was a relationship between parenting impacts and later child impacts. These substitutions were possible because parenting variables were selected for the main model so that a single variable represented a domain of parenting, and often, alternative variables measuring similar aspects of parenting were available.
Table D.9D presents the results of estimating models of the 36-month child outcomes by program approach. For center-based programs, we estimated models of cognitive and language development and aggressive behavior. Models of the other three social-emotional outcomes could not be estimated because, within the center-based group, Early Head Start had no impact on nearly all key parenting mediators that might predict these outcomes. In each model that we did estimate for families in center-based programs, one or two of the parenting mediators was not changed by Early Head Start at 24 months, so the models did not include all of the variables used for the full sample. The results of the estimation suggest that parenting behavior at 24 months is related to the later child outcomes in the expected directions, but the implied pathway for program impacts through parenting behavior to children in the later period appears to be fairly small, in part because few of the parenting influences were affected by the program in the earlier period.
| Parenting Mediators |
Center-Based Programs | Mixed-Approach Programs | Home-Based Programs | |||
|---|---|---|---|---|---|---|
| Estimated Effect of Parenting Variables on Child Outcome | Percentage of Child Impact Associated with Mediator | Estimated Effect of Parenting Variables on Child Outcome | Percentage of Child Impact Associated with Mediator | Estimated Effect of Parenting Variables on Child Outcome | Percentage of Child Impact Associated with Mediator | |
| Bayley MDI | ||||||
| Supportiveness: Semistructured Play | n.a. | n.a. | 2.24*** | -149.3 | 1.92*** | 5 |
| HOME Support of Cognitive, Language, and Literacy Environment | n.a. | n.a. | 1.04** | -114 | 1.41*** | 16 |
| Parent-Child Play | 1.85* | 2.9 | n.a. | n.a. | n.a. | n.a. |
| Read Daily | 0.14 | 0.3 | 1.75 | -65.3 | 1.31 | 1.7 |
| Percentage of Child Outcome Attributed to Parenting | -- | 3.2 | -- | -328.6 | -- | 22.6 |
| PPVT Score Supportiveness: Semistructured Play | n.a. | n.a. | 3.00*** | 13.6 | 0.79 | 2.7 |
| HOME Support of Cognitive, Language, and Literacy Environment | n.a. | n.a. | 1.16* | 8.4 | 1.30** | 19.2 |
| Parent-Child Play | 1.31 | -2 | n.a. | n.a. | n.a. | n.a. |
| Read Daily | 5.86*** | 14.5 | 0.01 | 0 | 0.55 | 0.9 |
| Percentage of Child Outcome Attributed to Parenting | -- | 12.5 | -- | 22 | -- | 22.8 |
| Sustained Attention with Objects: Semistructured Play | Not Estimated | |||||
| Supportiveness: Semistructured Play | 0.09* | 7.3 | 0.12*** | 5.3 | ||
| HOME Support of Cognitive, Language, and Literacy Environment | 0.03 | 4.4 | 0.04 | 6.9 | ||
| Knowledge of Infant Development Inventory | 0.19 | 4.9 | 0.28*** | 15.1 | ||
| PSI: Parental Distress | -0.01 | 5.8 | 0 | 1.9 | ||
| Percentage of Child Outcome Attributed to Parenting | -- | 22.4 | -- | 29.3 | ||
| Engagement of Parent: Semistructured Play | Not Estimated | |||||
| Warm Sensitivity: Semistructured Play | 0.18*** | 15.2 | 0.22*** | 4.7 | ||
| HOME Emotional Responsivity | -0.003 | -0.3 | 0.11*** | 6.8 | ||
| HOME Support of Cognitive, Language, and Literacy Environment | 0.06* | 9 | 0.05* | 5 | ||
| Parent-Child Play | n.a. | n.a. | n.a. | n.a. | ||
| Detachment: Semistructured Play | -0.14** | 11.4 | n.a. | n.a. | ||
| Percentage of Child Outcome Attributed to Parenting | -- | 35.3 | -- | 16.5 | ||
| Negativity toward Parent: Semistructured Play | Not Estimated |
Not Estimated |
||||
| Warm Sensitivity: Semistructured Play | -0.01 | 0.8 | ||||
| Physical Punishment in Past Week | -0.03 | -2.4 | ||||
| PSI: Parental Distress | 0.01 | 6.8 | ||||
| Intrusiveness: Semistructured Play | 0.07** | 4.9 | ||||
| Percentage of Child Outcome Attributed to Parenting | -- | 10.1 | ||||
| Aggressive Behavior | ||||||
| Warm Sensitivity: Semistructured Play | n.a. | n.a. | -0.54* | 12.1 | -0.33 | 33.6 |
| 2.13*** | 10.2 | 1.34** | 17.6 | 1.44*** | 81.2 | |
| PSI: Parental Distress | n.a. | n.a. | 0.14*** | 35.9 | 0.20*** | 393.7 |
| Regular Bedtime | -1.45** | 7.5 | 0.14 | -0.5 | -0.3 | 20.7 |
| Percentage of Child Outcome Attributed to Parenting | -- | 17.8 | -- | 65.1 | -- | 529.2 |
| SOURCE: Parent interviews, child assessments, and assessments of parent-child interactions when children were approximately 24 and 36 months old. |
For home-based programs, we estimated all of the models except for negativity toward the parent during semistructured play at three years. At 24 months, Early Head Start had a favorable impact on nearly all of the aspects of parenting used in these models, so only one variable was omitted from one model (engagement of parent). The estimated relationships between parenting behavior variables at 24 months and children’s outcomes at 36 months were consistently in the expected directions. Overall, the estimates suggest that part of the Early Head Start impact on the cognitive, language, and socio-emotional development of children in home-based programs could have emerged because of earlier impacts on related parenting behavior.
The model for aggressive behavior among children in home-based programs has the striking result that more than 500 percent of children’s aggressive behavior at 36 months is potentially associated with the earlier changes in parenting. Most of the association between parenting and children’s aggression stems from a large estimated relationship between parenting behavior at 24 months and children’s aggressive behavior at 36 months, which again could be partly attributable to measurement error leading to some degree of spurious correlation between these two measures.
For mixed-approach programs, we estimated all of the models, and since Early Head Start influenced all key parenting outcomes at 24 months, none had to be omitted from any model. For the most part, the estimated relationships between parenting behavior at 24 months and child outcomes a year later are usually in the expected directions. Supportiveness, cognitive stimulation, and language support are all positively related to cognitive and language development and positive aspects of social-emotional development and inversely related to negative aspects of social-emotional development. Intrusiveness, detachment, and parental distress are all inversely related to positive aspects of social-emotional development and positively related to negative aspects of social-emotional development. Within the mixed-program group, there are a few exceptions to these rules, but in these cases the estimates are usually small (not different from zero) and the percentage of the child impact associated with the parenting mediator is small. Overall, the estimates are consistent with the theory that, for families in mixed-approach programs, part of the Early Head Start impact on children’s outcomes may be mediated by earlier impacts on parenting behavior.
In the model relating the Bayley MDI scores to parenting behavior a year earlier for families in mixed-approach programs, the estimated relationships appear to be particularly strong, which makes the proportion of the Bayley MDI impact that is associated with earlier parenting impacts unreasonably high. Unfortunately, such a result is possible with the two-stage estimation procedure, which cannot force the results to fall between 0 and 100 percent. Instead, the procedure takes the estimated association between the parenting outcomes and child outcomes and checks the consistency of the earlier parenting impacts and that association with the ultimate child impacts a year later. An unreasonable result such as this can suggest either that the theory of which parenting behaviors affect the child outcome is incorrect, or (more likely) that this model is incorrect because it does not correctly capture all of the structural relationships among parenting behavior, genetics, other home influences, and children’s outcomes.
Nevertheless, while the specific parameter estimates from the models are likely to be biased, the overall pattern of association between parenting impacts at 24 months and children’s impacts at 36 months can provide an indication of whether the impacts are consistent with the programs’ theory of change. Estimates for mixed and home-based programs do lend some support to Early Head Start program theories of change that suggest a role for parenting as a mediator of program impacts on children. For center-based programs, parenting appears not to have had much of a role in mediating program impacts on children, in large part because few impacts on parenting were found at the 24-month assessment.
14Warm sensitivity is a composite of two out of three variables that comprised the measure of Supportiveness. Warm sensitivity includes Positive Regard and Sensitivity, but omits Cognitive Stimulation; all three of the variables are averaged to create the Supportiveness measure. (back)
| Table of Contents | Previous | Next |