Table of Contents | Previous | Next |
Technical Report #2
Methodology Employed in Creating Family Clusters
Head Start families become Head Start families based on a single criterion: the family’s income at time of application is below the federal poverty line for a family of a given size. There is one exception to this: if a child has a disability, the family does not need to meet the income eligibility criterion, although in fact the overwhelming majority of Head Start children with disabilities also come from families with poverty level incomes.
In general, Head Start children are considered at elevated risk for non-optimal school performance, based on many studies and reports that children from poverty families do less well than children from higher income families on almost all measures of academic success, and sometimes on measures of social adjustment. Follow-up studies of former Head Start children suggested there may be a “loss of gains” over the first four years of school. Reviews of research on the effects of Head Start suggest there may be benefits in social development and academic readiness. As stated in the overview chapter, the rationale for this national project was largely to prevent this possible decline. The goal was to enhance children’s school adjustment over the first four years.
What is often ignored in studies of how children adjust to school is the relationship of poverty to many other life conditions and resources available to children, families, their schools, and communities. Accordingly, in this study, it was recognized from the beginning, that different children and families may have different needs, in part related to their life situation when the children enrolled in kindergarten. Further, families are not static, and changes in the family’s life situation may contribute to other important changes in the child’s adjustment to school.
In this chapter, we present descriptive information about the natural diversity that occurs within a large sample of former Head Start children. It is important to note, however, that this is not necessarily a nationally representative sample of former Head Start children. In fact, there is an under representation of the Hispanic Latino families in this sample. The sample represents grantees who were skillful in presenting a strong proposal for how they would enhance the school-age outcomes for former Head Start children and their families, and these sites typically are somewhat above the national average on a number of sociological and community level markers, such as rates of unemployment, income, crime, and parent educational level. However, this sample does include a wide range of characteristics, which is typical of national data presented about Head Start (ACYF, 1998).
We also emphasize here that despite the risk conditions associated with poverty, there are a number of important strengths that characterize many former Head Start families. These may be just as important in helping to understand the different school age outcomes for children and families. The slanted view often presented on families that are marginalized because of income or other factors, such as ethnicity, has been well articulated (Huston, McLoyd, Garcia Coll, 1994; Huston, 1991).
Since family strengths historically have been neglected when studying children of poverty, we begin by presenting these characteristics (referred to as the family strengths index). Next, we describe the extent to which certain parent and family characteristics typically considered “challenges” or “risk” variables affect this sample (referred to as the family challenges index). Finally, we present several ways of looking at these characteristics collectively or holistically. The findings indicate that there is tremendous diversity in this 31 site study. This diversity is so strong that many of the analyses presented in later chapters on the outcomes for children, families, schools, and communities take this natural diversity -- present when the study began -- into account.
Family Strengths Index
A family strengths index was created based on six characteristics related to child outcomes. These characteristics included: primary caregiver has college degree or higher; mother and father active as parents (defined as either living in the home or active in helping with parenting duties); income greater than or equal to 150% of poverty; family rates the probability for success in the neighborhood as high to very high (e.g. graduating from high school, attending college); family member reads daily to child; and family routines are highly organized (see Table 1). These binary variables were summed to create a family strengths index, with a possible range of 0 - 6. The majority of the families had either one (39%) or two (22%) strengths. Only about 12% of the families reported having three or more strengths. Interestingly, more than a quarter of the families did not report strengths. The mean strength was 1.22 (1.06 SD).
It is important to note that there are some important aspects of family life that are known to affect children about which we did not have information. Factors such as special mentor or special person who is actively engaged in a child’s life were not always known, and day-to-day family functioning was not measured objectively, but rather assessed through the perspective of the child’s primary caregiver.
| Variable | Family Challenges Index (12 variables) |
Family Strengths Index (6 variables) |
|---|---|---|
| Primary Caregiver |
||
| Caregiver education level | less than high school | college degree or higher |
| Caregiver has chronic health problem | yes | ----- |
| Caregiver has positive depression screen | yes | ----- |
| Caregiver's age at child's entry into kindergarten | less than 24 years | ----- |
| Mother and father active as parents (either live in home or help with parenting) | ----- | yes |
| Family Resources | ||
| Household income | < 50% federal poverty level | > 150% federal poverty level |
| Family receives AFDC | yes | ----- |
| Family has been homeless in past 12 months or lives in shelter | yes | ----- |
| Family moved 2 or more times in last 12 months | yes | ----- |
| Family has 4 or more children in home | yes | ----- |
| Neighborhood |
||
| Probability for success scale | low to very low | high to very high |
| Family Supports for Learning/School |
||
| Reading/storytelling to child | 1-2/wk or almost never | daily |
| Family routines | highly disorganized defined < 46 | highly organized defined as > as 72 |
Family Challenges Index
A family challenges index was created based on twelve characteristics. These characteristics included: primary caregiver has less than a high school diploma or GED; primary caregiver has a chronic health problem; primary caregiver was screened positive for depression; primary caregiver’s age at child’s kindergarten entry was less than 24 years; family income was less than 50% of poverty; family receives AFDC; family is homeless or in a shelter; family has moved two or more times in the past year; four or more children live in the home; family rates the probability for success in the neighborhood as low to very low (e.g. graduating from high school, attending college); family member reads infrequently to child (1-2 times per week or almost never); and family routines are highly disorganized (see Table 1 above). These binary variables were summed to create a family challenges index, with a possible range of 0 - 12. There were approximately equal numbers of families who reported one challenge (19%), two challenges (22%), and three challenges (18%). Somewhat fewer, but still a fair number of families, reported four challenges (14%) or five or more challenges (14%). More than one out of ten families (13%) reported that they did not have any challenges. The mean challenge was 2.52 (1.77 SD).
Again, there are some important factors about the family known to affect children about which we did not have information. These factors include substance abuse and alcoholism, domestic violence, child abuse and neglect, parental intellectual disability, and incarceration among others.
It is important to note that some of the same characteristics used in the family challenges index were also used in the family strengths index. The characteristics, however, are not the inverse of each other; rather they are either the upper or lower limit of the variable (e.g., less than high school education as a challenge and college degree or higher as a strength). The correlation between the family strengths and family challenges indices was -0.42 (p<.001).
Family Typology
A series of analyses were conducted to develop a typology of former Head Start families as the children entered school. The family typology reported in this chapter depicts the diversity within poverty families and illustrates how a set of standard descriptive variables are interrelated. Also, these analyses help to set the stage for decision making regarding whether number of challenge conditions, challenge-to-strength ratios, or clusters will be used as a standard option in looking for subgroup differences in response to the Transition Demonstration treatment.
The general strategy used for developing the family typology was as follows. First, fifteen variables known to describe relevant family characteristics were selected. These family variables were: percent receiving AFDC, percent receiving SSI, percent employed full-time, mean percent of poverty, percent finished high school, mean caregiver age (when child entered school), percent positive depression screen, percent with a chronic health problem that interferes with parenting duties, percent with father active in child’s life, percent with mother absent from child’s life, mean number of children, percent born outside of the United States, percent reporting a language other than English as the primary language spoken in the home, percent of families who have moved two or more times in the past year, and percent of families who were homeless in the past year (see Table 2 below). These variables are essentially the same as a set used in a previously reported paper (see Ramey, Ramey, & Lanzi, 1996). The set of analyses reported here, however, was conducted on the final analysis sample and included SSI, high mobility, and maternal depression, given our recognition of their importance and prevalence. In addition, other variables were modified to yield simpler categorical classification, in part based on the fact that detailed information may not be available in other administrative databases.
| Total Sample | Number Missing | Percent Missing | |
|---|---|---|---|
| Receives AFDC | 37.80% | 21 | 0.30% |
| Receives SSI | 12.30% | 22 | 0.30% |
| Caregiver employed full-time | 32.10% | 7 | 0.10% |
| Percent of poverty (family income) | 79.47 -62.14 | 648 | 9.10% |
| Caregiver finished high school | 67.50% | 672 | 9.50% |
| Caregiver age (when child entered school) | 31.18 -7.54 | 218 | 3.10% |
| Caregiver has positive depression screen | 43.50% | 639 | 9.00% |
| Caregiver has a chronic health problem | 3.60% | 10 | 0.10% |
| Mother active in child's life | 94.90% | 0 | 0% |
| Father active in child's life | 52.40% | 0 | 0% |
| Mean number of children in family | 2.87 -1.43 | 18 | 0.20% |
| Caregiver born outside United States | 17.00% | 241 | 3.40% |
| English is primary language spoken in home | 85.70% | 9 | 0.10% |
| Family moved 2 or more times in last year | 7.50% | 181 | 2.60% |
| Homeless in past 12 months | 3.30% | 73 | 1.00% |
As a group, these 15 variables represent a combination of risk conditions, logistical challenges, and factors that historically have been associated with non-optimal school outcomes for children. For this analysis, data from the kindergarten year were used. Whenever possible, information gathered during the fall (baseline) was used, although for a subset of families, information was not available until the spring of the kindergarten year. The missing data for any given variable never exceeded 9 percent.
The correlation coefficient was used as the measure of similarity and Ward’s method as the clustering criterion. This approach has been used with previous studies (Ramey et al, 1984; Ramey, Ramey, & Lanzi, 1996). The validity of the cluster analysis solution is an important concern, since random data can give rise to seemingly appropriate cluster solutions (Dubes & Jain, 1979). Thus, consideration was given to whether groupings of families were an artifact of the cluster analysis. Because of the size of the current data set, replication and significance tests were completed for the independent variables. Hence, the data set was randomly divided into two equal groups. The cluster procedure was then applied to each data subset, using the fifteen variables listed above to determine if similar solutions were obtained.
Another critical issue in conducting a cluster analysis is determining the appropriate number of clusters or groups supported by the data (Milligan & Cooper, 1985). Milligan and Cooper (1985) list 30 techniques proposed for this problem. The clustering criterion available in SAS was selected. In applying the SAS criterion in the cluster analysis, seven clusters were identified in both data sets. There was a one-to-one correspondence between the two sets of clusters. The results obtained from this replication procedure clearly indicated the existence of seven groupings of families.
To determine a final cluster solution, the clustering procedure was applied to the original full data set. The findings from this analysis indicate that, among the families participating in this National Transition Demonstration Study, there are remarkably clear major distinctions. A clear identification of seven major family types emerged. Table 2 presents the distribution of each of the 15 variables for each family type. A brief characterization of these seven family types is found in Chapter 3. A central finding, as with the previous cluster analyses, is that all of these family types occur in all major ethnic/cultural groups studied: White/non-Hispanic, African American, Hispanic/Latino, Asian/Pacific Islander, and American Indian (see Table 3).
| VARIABLES USED IN CLUSTER ANALYSIS | FAMILY TYPE | ||||||
|---|---|---|---|---|---|---|---|
| Total Sample | A Resourceful |
B Single Parent Welfare |
C Foreign Language |
D Highly Mobile |
E Mother Absent |
F Chronic Health Problem |
|
| n=2584 | n=1840 | n=656 | n=336 | n=280 | n=198 | ||
| 42% | 30% | 11% | 6% | 5% | 3% | ||
| Receives AFDC | 38% | 4% | 85% | 21% | 49% | 40% | 50% |
| Receives SSI | 12% | 0.90% | 27% | 5% | 14% | 21% | 27% |
| Primary Caregiver Employed Full-Time | 32% | 48% | 12% | 37% | 31% | 40% | 15% |
| Percent of Poverty (Family Income) | M= 79.4 | M=105.1 | M=49.7 | M=76.6 | M=70.5 | M=93.1 | M= 67.7 |
| Percent of Poverty (Family Income) | SD=(62.1) | SD=(73.6) | SD=(32.9) | SD=(46.0) | SD=(49.2) | SD=(65.4) | SD=(52.9) |
| Primary Caregiver Finished High School | 67% | 79% | 64% | 42% | 68% | 63% | 63% |
| Primary Caregiver Age (When Child Entered School) | M=31.2 | M=31.05 | M=29.3 | M=32.0 | M=28.1 | M= 43.1 | M=34.3 |
| Primary Caregiver Age (When Child Entered School) | SD=(7.5) | SD=(6.96) | SD=(5.3) | SD=(6.8) | SD=(4.9) | SD=(12.1) | SD=(9.7) |
| Primary Caregiver with Depressive Signs | 43% | 37% | 48% | 40% | 53% | 36% | 62% |
| Primary Caregiver has a Chronic Health Problem that Interferes with Parenting Duties |
4% | 0% | 0% | 0% | 0% | 0% | 100% |
| Father Active in Child's Life | 52% | 64% | 32% | 77% | 43% | 38% | 51% |
| Mother Absent from Child's Life | 5% | 0% | 0% | 0% | 0% | 100% | 8% |
| Number of Children in Family | M=2.9 | M=2.6 | M=3.0 | M=2.9 | M=2.8 | M=3.0 | M=2.7 |
| Number of Children in Family | SD=(1.4) | SD=(1.1) | SD=(1.5) | SD=(1.5) | SD=(1.1) | SD=(1.8) | SD=(1.5) |
| Parent(s) Born Outside the United States | 17% | 6% | 0.90% | 91% | 11% | 6% | 14% |
| Language other than English Spoken in Home | 14% | 0% | 0% | 100% | 8% | 4% | 9% |
| Family Moved 2 or More Times in Last Year | 8% | 0% | 0% | 0% | 100% | 6% | 6% |
| Homeless in Past 12 Months | 3% | 0% | 0% | 0% | 0% | 0% | 0% |
| ETHNIC GROUPS | TOTAL SAMPLE | Resourceful | Single Parent Welfare | Foreign Language | Highly Mobile | Mother Absent | Chronic Health Problem | Recently Homeless |
|---|---|---|---|---|---|---|---|---|
| White/non-Hispanic | 2863 | 1496 | 860 | 21 | 177 | 120 | 97 | 92 |
| 47% | 58% | 47% | 3% | 53% | 43% | 49% | 47% | |
| African American | 1940 | 784 | 801 | 20 | 96 | 118 | 64 | 57 |
| 32% | 30% | 44% | 3% | 29% | 42% | 32% | 29% | |
| Hispanic/Latino | 797 | 149 | 75 | 480 | 34 | 17 | 17 | 25 |
| 13% | 6% | 4% | 73% | 10% | 6% | 9% | 13% | |
| Asian/Pacific | 114 | 16 | 2 | 80 | 5 | 1 | 7 | 3 |
| Islander | 2% | 0.60% | 0.10% | 12% | 1.50% | 0.40% | 4% | 2% |
| American Indian | 156 | 71 | 41 | 4 | 10 | 15 | 6 | 9 |
| 3% | 3% | 2% | 0.60% | 3% | 5% | 3% | 5% | |
| Other | 212 | 63 | 60 | 51 | 14 | 9 | 7 | 8 |
| 4% | 3% | 3% | 8% | 4% | 3% | 3% | 4% | |
| Total | 6082 | 2579 | 1839 | 656 | 336 | 280 | 198 | 194 |
| 100% | 42% | 30% | 11% | 6% | 5% | 3% | 3% |
| Table of Contents | Previous | Next |