CHAPTER II: THE HEAD START PROGRAM AND RELATED RESEARCH

Local agencies operate thousands of Head Start programs in communities across the country. Although all of the programs are required to meet the Head Start program performance standards, services are tailored to the needs of the local communities. To provide context for the evaluation design issues discussed in this report, we begin by describing the major characteristics and features of Head Start programs, highlighting the ones that are common to most programs, as well as the ones that generally vary among programs. We then describe the diverse demographic characteristics of the children and families served.

Research has been an integral part of Head Start over the years. Recent studies have examined several questions in an effort to understand better which Head Start practices work best for children. We discuss the goals and approaches of these in order to identify the unique contribution that the current study will make to the recent Head Start research efforts. We also discuss the Head Start Planned Variation study that was conducted from 1967 to 1977 and the lessons from that effort that inform our design today.

1. OVERVIEW OF THE HEAD START PROGRAM

Since 1965, the Head Start program has provided comprehensive child development services to approximately 21 million low-income children and their families. In an effort to break the cycle of poverty the Office of Economic Opportunity launched Head Start as an eight-week summer program for children about to enter kindergarten. Head Start soon expanded to offer comprehensive services to four-year-old children in half-day, center-based programs that operated during the school year.

Head Start now is the largest federally funded early childhood program. In fiscal year (FY) 2004, the Administration for Children and Families (ACF) provided $6.6 billion to local grantees across the country to offer comprehensive child development services, predominantly to three- and four-year-old children.1 Grantees operated approximately 18,269 centers and served 976,013 children and their families.2 The program provides half-day and full-day program options, two-year programs (for three- and four-year-old children), and other service variations for children in rural areas. Early Head Start, begun in 1995, extends Head Start services to younger children by providing child development and family development services to pregnant women and to children up to age three in home-based and/or center-based settings. In 2003, Early Head Start programs served approximately 71,000 children and 9,500 pregnant women.

Head Start is administered by the Head Start Bureau, which is part of the Administration on Children, Youth, and Families, ACF, U.S. Department of Health and Human Services (DHHS). Grants to operate Head Start programs at the community level are awarded directly from the federal government to community action agencies, public or private school systems, nonprofit organizations, local government agencies, private or public for-profit agencies, and Tribal or Alaska Native government organizations. Staff in the 12 regional ACF offices work closely with Head Start Bureau staff and serve as a direct link to local grantee programs. 3 The regional offices monitor the operation of local programs.

The Head Start program offers a comprehensive set of services to low-income children and their families that focus on preparing economically disadvantaged children for school. The services fall into two main domains:

1. Early childhood development and health services cover the areas of child health and development, education and early childhood development, child health and safety, child nutrition, and child mental health.

2. Family and community partnerships include family goal-setting, access to community services and resources, services for pregnant women, involvement of parents, formation of partnerships in the community, advisory committees, and transition services.

Head Start programs are charged with providing services that promote children’s development and school readiness in culturally appropriate ways that meet the needs of the children and their families. Grantees must conduct periodic community needs surveys to update their knowledge about the needs of families in the community. In the following section, we discuss the diversity of Head Start program models and characteristics and then describe the diversity of families served by Head Start programs. This discussion is based on analyses of data from the Head Start PIR for the 2002-2003 program year. PIR data are collected annually from Head Start grantees.

2. CHARACTERISTICS OF HEAD START PROGRAMS

A Head Start grantee is a public or private agency or organization that receives funds to provide Head Start services, and that is responsible for meeting the conditions of the grant. Grantees may select delegate agencies to help provide Head Start services. A delegate agency is a public or private nonprofit organization or agency to which a grantee has delegated, by written agreement, responsibility for carrying out all or part of its Head Start program. A Head Start program can be either a grantee or a delegate agency. Many grantees (65 percent of all programs operating during the 2002-2003 program year) directly operate Head Start programs without any delegate agencies. A few grantees (five percent of all programs) both directly operate programs and have delegate agencies that operate programs. A very small proportion of grantees (one percent of all programs) maintains only central office staff and delegates all program operations. In total, 28 percent of all Head Start programs in the 2002-2003 program year were delegate agencies.4 When working with Head Start programs, researchers will have to approach grantees and, if applicable, delegate agencies to obtain permission to conduct the evaluation, and to obtain contact information about center directors.

Private or public nonprofit agencies and community action agencies were the most common types of agencies running Head Start programs during the 2002-2003 program year, operating 35 percent and 32 percent of all Head Start programs, respectively. School systems ran a substantial 20 percent of all Head Start programs.5 Head Start programs operated in 18,269 centers, which ran 44,656 classes. In addition, Head Start programs operated with 3,096 child care partnerships, 1,374 family child care homes, and 3,824 home-based groups.

The program options and intensity of services offered to families by Head Start programs are driven largely by the programs’ community resource and needs assessments. In recent years, welfare-related work requirements have led to an increasing need for child care while parents work. ACF has recognized this need by increasing funding in FY 2000 for full-day, full-year services designed to help working families that are moving from welfare to work. One of the ways in which programs provide full-day, full-year services to their enrolled children is by partnering with other child care centers or family child care homes. Services provided by the child care partners must meet the Head Start performance standards, including requirements for maximum child-adult ratios and the quality of the educational program.

During the 2002-2003 program year, nearly all Head Start children (93 percent) were enrolled in a part- or full-day center-based program. Almost half the children (49 percent) were enrolled in the full-day option (four or five days per week); 45 percent were enrolled in the part-day option (also four or five days per week); 3 percent were in the home-based option; and the remaining 4 percent were in combination, family child care, and locally designed options. The proportion of children served in a full-time, center-based program has increased substantially since the 1996-1997 program year, when only 24 percent of Head Start children received Head Start services in full-day, full-week, center-based care settings (Schumacher and Rakpraja 2003). Part-day programs offer a center-based program for fewer than six hours per day. Some of these programs are “double-session” classes in which a teacher provides center-based services to two different groups of children by offering morning and afternoon sessions.

Head Start programs across the country differ in important ways, reflecting regional and state differences in demographics, economic well-being, the states’ early childhood program contexts, and other characteristics. To some degree, we can illustrate the diversity of Head Start programs by comparing the characteristics of programs from different regions. Information about Head Start programs, by region, also is useful because evaluations of quality enhancement options are likely to be conducted in a few sites with sufficient geographic concentration to enable technical assistance staff to reach them easily. The characteristics of Head Start programs in specific regions might indicate that the region is a good (or a more challenging) place to try a particular innovative idea. To simplify the tables, we have collapsed the 10 geographically based regions (see Figure II.1) to form five regional groups: (1) Northeast (Regions 1, 2, and 3); (2) the South (Regions 4 and 6); (3) North Central (Region 5); (4) Mountain/Plains (Regions 7 and 8); and (5) the West (Regions 9 and 10). We also show separately the Migrant/Seasonal and American Indian/Alaska Native programs, which may be located anywhere in the country.

[D]




Table II.1 indicates that the mix of program options varies substantially across regions. During the 2002-2003 program year, full-day, full-week, center-based programs predominated in the South (where center-based child care is more available than in other regions and tends to be chosen more often than other forms of care) and among Migrant/Seasonal programs (because parents typically work full-time when seasonal jobs are available). Full-day, center-based care programs constituted half the program options offered in the Northeast. Half-day, center-based programs operating four or five days per week were the most common program options in the North Central, Mountain/Plains, and West regions, and in the American Indian/Alaska Native programs. The least common program options across all regions were the full-day, center-based programs operating four days per week and the home-based, combination, and family child care programs.

Table II.1 Percentage of Children Enrolled in Each Program Option, 2002-2003 Program Year

Program Option	All	Northeast (Regions 1, 2, and 3)	South (Regions 4 and 6)	North Central (Region 5)	Mountain and Plains (Regions 7 and 8)	West (Regions 9 and 10)	Migrant and Seasonal Programs	American Indian and Alaska Native Programs
Center-based: Five days/week, full-day	47	51	72	24	22	19	96	22
Center-based: Five days/week, half-day	22	28	20	9	11	42	1	13
Center-based: Four days/week, full-day	2	2	2	4	4	1	0	10
Center-based: Four days/week, half-day	23	13	4	54	53	27	0	47
Home-based	3	3	1	4	5	4	0	5
Combination	1	2	0	2	2	3	0	2
Family child care	1	1	0	1	1	1	3	0
Locally designed options	2	1	1	2	3	3	0	1

Source: Calculations based on 2002-2003 Head Start PIR data submitted by 1,969 Head Start programs and processed by Xtria, LLC, on February 12, 2004. Early Head Start programs were excluded from these calculations.




The number of children per adult in Head Start programs was low relative to other early childhood settings. Head Start Program Performance Standards require a ratio of 1 adult caring for no more than 7.5 to 8.5 children aged three years, and no more than 10 children aged four years. This ratio is comparable to the National Association for the Education of Young Children accreditation standards of 8 or fewer three-year-olds or 10 or fewer four-year-olds for each adult. The PIR data indicate that, on average, 7.8 children were cared for by each adult. Table II.2 indicates that the numbers of children per adult in the South and West were somewhat higher than the national average, but still within the limits set by the performance standards. The number of children per adult was lower in Migrant/Seasonal Head Start programs, which serve children younger than age three as well as preschool-age children.

Table II.2 Number of Children per Adult in Head Start Classrooms, 2002-2003 Program Year

All	Northeast (Regions 1, 2, and 3)	South (Regions 4 and 6)	North Central (Region 5)	Mountain and Plains (Regions 7 and 8)	West (Regions 9 and 10)	Migrant and Seasonal Programs	American Indian and Alaska Native Programs
7.8	7.7	8.5	7.6	7.0	8.2	5.1	7.0

Source: Calculations based on 2002-03 Head Start PIR data submitted by 1,969 Head Start programs and processed by Xtria, LLC, on February 12, 2004. Early Head Start programs were excluded from these calculations. Note: The child-to-adult ratio is calculated as the total funded environment (excluding home-based and family day care enrollment) subtracting one-half of the children enrolled in double sessions (to ensure that children are counted only once) and divided by the total number of classroom staff (teachers and assistant teachers).




Teachers’ education levels have increased in recent years in response to a requirement of the 1998 Head Start reauthorization law that, by 2003, at least half of all Head Start teachers in center-based programs must have an Associate in Arts degree (A.A.), a Bachelor of Arts degree (B.A.), or advanced degree in early childhood education (ECE) or in a related field. Table II.3 shows that 59 percent of the teachers had an associate’s, bachelor’s, or graduate degree in the 2002-2003 program year, thereby exceeding the teacher qualification mandate. A larger proportion of teachers in the Northeast relative to other regions had a bachelor’s or graduate degree.

The availability of direct child development staff (teachers, assistant teachers, home visitors, and family child care providers) who speak languages other than English to meet the needs of the linguistically diverse Head Start population is particularly significant. Overall, 28 percent of direct child development staff across all Head Start programs spoke languages other than English (see Table II.3). Migrant and American Indian/Alaska Native programs and programs in the West and Northeast far exceeded the national percentage in this respect, consistent with the substantial proportion of families in those regions who speak languages other than English. (This issue is discussed in greater detail in the next section.)

Table II.3 Percentage of Teachers by Education Levels and Languages Spoken, 2002-2003 Program Year

Education Levels/ Languages Spoken	All	Northeast(Regions 1, 2, and 3)	South (Regions4 and 6)	North Central (Region 5)	Mountain and Plains (Regions7 and 8)	West (Regions 9 and 10)	Migrant and Seasonal Programs	American Indian and Alaska Native Programs
Child Development Associate credential/state equivalent	26	15	35	20	25	21	42	38
Associate's degree, ECE/related	27	20	28	33	22	40	15	21
Bachelor's degree, ECE/related	28	47	23	29	35	19	11	9
Graduate degree, ECE/related	4	10	3	4	4	2	1	0
Child development staff speaking languages other than English	28	32	17	10	12	49	65	38

Source: Calculations based on 2002-2003 Head Start PIR data submitted by 1,969 Head Start programs and processed by Xtria, LLC, on February 12, 2004. Early Head Start programs were excluded from these calculations. ECE = early childhood education.




3. CHARACTERISTICS OF HEAD START CHILDREN AND FAMILIES

The 869,570 Head Start slots funded during the 2002-2003 program year were distributed unevenly across the geographic regions (see Table II.4). The South contained the largest proportion of Head Start-funded slots (nearly one-third of all funded slots), followed by the Northeast and North Central regional groups, each of which had approximately one-fifth of the funded slots. The Migrant/Seasonal programs and the American Indian/Alaska Native programs had the smallest proportions of funded slots.

The number of children enrolled in Head Start (976,013) during the 2002-2003 program year exceeded the number of funded slots because of normal turnover during the program year. Children are counted as enrolled in Head Start if they have attended at least one class session or have received at least one home visit. Head Start staff try to fill the slots of children who leave the program, although they cannot always do so. During the 2002-2003 program year, 16 percent of children who had enrolled left during the program year, and approximately 5 percent of the slots could not be filled. The proportion of enrolled children who leave during the program year is important because it lies in the fact that children in an evaluation study who leave the program during the program year still must be located and assessed at the follow-up points, even though this might be more difficult to do so if the children have left the program. The proportion of children who left Head Start during the 2002-2003 program year was lowest (12 percent to 13 percent) among children in the Northeast and South regional groups and among children in the American Indian/Alaska Native programs. Turnover was higher (17 percent to 23 percent) in the other regional groups.

Table II.4. Head Start Program Funded Enrollment, 2002-2003 Program Year

Enrollment	All	Northeast (Regions 1 ,2, and 3)	South (Regions 4 and 6)	North Central (Region 5)	Mountain and Plains (Regions 7 and 8)	West (Regions 9 and 10)	Migrant and Seasonal Programs	American Indian and Alaska Native Programs
Total funded enrollment	869,570	187,296	271,282	159,800	57,420	138,599	33,721	21,452
% funded enrollment, by region	100	22	31	18	7	19	4	3
Total actual enrollment	976,013	205,354	303,534	185,145	65,169	161,393	33,188	22,230
% who left during the program year	16	13	13	18	20	17	23	12
Number of programs	1,969	513	498	337	175	232	65	149

Source: Calculations based on 2002-2003 Head Start PIR data submitted by 1,969 Head Start programs and processed by Xtria, LLC, on February 12, 2004. Early Head Start programs were excluded from these calculations. Note: Funded enrollment is the number of children that the Head Start program is funded to serve. The actual enrollment, or the number of children enrolled who attended at least one class or who had at least one home visit, is larger than the funded enrollment because of normal attrition and replacement of children during the program year.




To be eligible for Head Start, a child must meet the program’s age requirements, and the family’s income must be lower than the official poverty guidelines. Up to 10 percent of the children enrolled in a Head Start program may be from families that exceed the poverty guidelines. A child from a family that receives public assistance or a child in foster care is eligible even if the family income exceeds the income guidelines. As shown in Table II.5, the majority of Head Start children and families (73.9 percent) were eligible strictly on income criteria, 18 percent were eligible because they received public assistance, 7 percent were from families that exceeded the income guidelines, and 1 percent were foster children. The highest percentage of children eligible by the public assistance criterion were from programs in the Northeast and North Central regional groups and from American Indian/Alaska Native programs; in general, however, the regional groups differed little in enrollment by eligibility status. Migrant/Seasonal programs served the highest number of children who met the federal poverty income guidelines.

Table II.5. Percentage of Children Enrolled in Head Start, by Type of Eligibility, 2002-2003 Program Year

Eligibility for Head Start	All	Northeast (Regions 1, 2, and 3)	South (Regions 4 and 6)	North Central (Region 5)	Mountain and Plains (Regions 7 and 8)	West (Regions 9 and 10)	Migrant and Seasonal Programs	American Indian and Alaska Native Programs
Receipt of public assistance	18	23	13	22	16	18	2	22
Income-eligible (<100% of the federal poverty line)	74	67	81	69	74	73	93	57
Over-income (>100% of the federal poverty line) a	7	8	6	8	8	7	5	19
Foster child status	1	2	1	1	2	2	0b	2

Source: Calculations based on 2002-2003 Head Start PIR data submitted by 1,969 Head Start programs and processed by Xtria, LLC, on February 12, 2004. Early Head Start programs were excluded from these calculations. a Up to 10 percent of children enrolled in a Head Start program can be from families with income above 100 percent of the federal poverty line. b One-tenth of one percent of children in Migrant/Seasonal programs have foster child status.




The majority of children in Head Start were four-year-olds (56 percent during the 2002-2003 program year), and more than one-third (36 percent) were three-year-olds (see Table II.6). This age distribution was similar in all regions except Migrant/Seasonal programs, which serve a somewhat more even distribution of children across the age range from birth to six years.

Seventy percent of the children enrolled in Head Start programs were in their first year of Head Start, and 28 percent were enrolled for a second year (see Table II.6). The composition of children according to the number of years in Head Start was fairly consistent across the geographic regions, but the Migrant/Seasonal programs, which traditionally serve infants, toddlers and preschool-age children, served a substantial percentage of children (14 percent) for a third year or longer.

Table II.6. Percentage of Children Enrolled in Head Start, by Age and Years of Enrollment, 2002-2003 Program Year

Characteristic		All	Northeast (Regions 1, 2, and 3)	South (Regions 4 and 6)	North Central (Region 5)	Mountain and Plains (Regions 7 and 8)	West (Regions 9 and 10)	Migrant and Seasonal Programs	American Indian and Alaska Native Programs
Age	Younger than one year	0	0	0	0	0	0	12	0
	One year	1	0	0	0	0	0	15	0
	Two years	1	1	0	2	1	1	18	1
	Three years	36	36	39	38	38	32	23	37
	Four years	56	58	57	53	58	64	21	49
	Five years	5	5	4	7	3	4	11	14
Years in Head Start	One	70	70	70	68	72	75	58	62
	Two	28	28	30	30	27	24	28	34
	Three or more	2	2	1	2	2	2	14	4

Source: Calculations based on 2002-2003 Head Start PIR data submitted by 1,969 Head Start programs and processed by Xtria, LLC, on February 12, 2004. Early Head Start programs were excluded from these calculations.




Slightly more than 10 percent of children enrolled in Head Start had a diagnosed disability. The majority of these children had speech and language disorders. A large proportion of those with disabilities were classified as having non-categorical/developmental delays. The next most common disabilities reported were health impairments, emotional/behavioral disorders, and multiple disabilities.

Children enrolled in Head Start programs were ethnically diverse, with fairly equal proportions of black or African American children (32 percent), Hispanic or Latino children (31 percent), and white children (27 percent) enrolled. Children’s ethnic backgrounds varied substantially by geographic region (see Table II.7). The Northeast’s programs served the three major ethnic groups in fairly even proportions. Black or African American children predominated in the South and North Central regional groups; white children predominated in the Mountain/Plains regional group; and Hispanic or Latino children predominated in the West.

Table II.7. Percentage of Children Enrolled in Head Start by Ethnicity and Primary Home Language, 2002-2003 Program Year

Characteristic		All	Northeast (Regions 1, 2, and 3)	South (Regions 4 and 6)	North Central (Region 5)	Mountain and Plains (Regions 7 and 8)	West (Regions 9 and 10)	Migrant and Seasonal Programs	American Indian and Alaska Native Programs
Ethnicity	Black or African American	32	30	49	39	17	11	1	1
	Hispanic or Latino origin	31	35	23	13	22	56	97	2
	White	27	28	23	40	52	17	1	8
	American Indian or Alaska Native	3	1	1	1	3	2	0	86
	Native Hawaiian/other Pacific Islander	1	1	0	0	0	5	0	1
	Asian	2	2	1	2	1	5	0	0
	Biracial/multiracial	3	3	2	4	5	4	0	2
	Other/unspecified	1	1	1	1	1	2	1	1
Primary Language Spoken at Home	English	73	63	86	87	84	49	9	92
	Spanish	23	31	12	9	13	42	86	1
	Native Central American, South American, and Mexican	1	1	1	0	0	0	4	0
	Caribbean	0	1	1	0	0	0	1	0
	Middle Eastern and South Asian	1	1	0	1	0	1	0	0
	East Asian	1	1	0	1	1	3	0	0
	Native North American/Alaska Native	0	0	0	0	0	0	0	7
	Pacific Island	1	0	0	0	0	4	0	0
	European and Slavic	1	2	0	0	0	1	0	0
	African	0	1	0	1	1	0	0	0
	Other	0	0	0	0	0	0	0	0

Source: Calculations based on 2002-2003 Head Start PIR data submitted by 1,969 Head Start programs and processed by Xtria, LLC, on February 12, 2004. Early Head Start programs were excluded from these calculations.




More than one-quarter of Head Start children spoke a language other than English at home. Spanish was the most common non-English language, spoken by 23 percent of the Head Start children at home (see Table II.7). Migrant/Seasonal programs and programs in the West and Northeast had substantial percentages of Spanish-speaking children relative to the national average. The number of children and families speaking non-English languages in Head Start has implications for the hiring of bilingual/multilingual program staff, the program materials printed for families, translation needs at policy council and other parent-staff meetings, and the child assessments conducted by programs as part of their self-monitoring.

Slightly more than half of all Head Start families (56 percent) were headed by single parents (see Table II.8). Single-parent families were the predominant type in the Northeast, South, and North Central regional groups, comprising 56 percent to 64 percent of all families. In the West and among the Migrant/Seasonal and American Indian/Alaska Native programs, two-parent families were more common than were single-parent families. In the Migrant/Seasonal programs in particular, 80 percent of the families were headed by two parents.

Table II.8 Percentage of Families Enrolled in Head Start, by Household Composition and Employment Status, 2002-2003 Program Year

Characteristic		All	Northeast (Regions 1, 2, and 3)	South (Regions 4 and 6)	North Central (Region 5)	Mountain and Plains (Regions 7 and 8)	West (Regions 9 and 10)	Migrant and Seasonal Programs	American Indian and Alaska Native Programs
Household Composition	Two-parent familiesa	44	44	37	38	49	58	80	59
	Single-parent families	56	56	64	62	52	42	20	41
Employment Status of Two-Parent Families	Both parents employed	34	31	34	32	35	28	76	42
	One parent employed	52	53	54	53	53	59	17	42
	Neither parent employed	14	16	13	14	12	14	7	16
	Unknown	0	0	0	1	0	0	0	0
Employment Status of Single-Parent Families	Parent employed	61	58	62	64	60	60	82	58
	Parent not employed	39	42	38	36	40	40	18	42
	Unknown	0	0	0	1	0	0	0	0

Source: Calculations based on 2002-2003 Head Start PIR data submitted by 1,969 Head Start programs and processed by Xtria, LLC, on February 12, 2004. Early Head Start programs were excluded from these calculations. a Two-parent families may be married or unmarried.




Approximately one-third of two-parent Head Start families were families in which both parents worked; slightly more than half were families in which one of the two parents worked. Neither parent worked in just over one-tenth of the families. Among one-parent Head Start families, 61 percent of the parents worked and 39 percent did not. These proportions were fairly consistent across geographic regions (see Table II.8), although employment rates were higher among parents of children enrolled in Migrant/Seasonal programs than in other programs.

The majority (68 percent) of Head Start parents had at least a high school diploma or GED (high school equivalence certificate), although a substantial proportion (one-third) did not (see Table II.9). The distribution of education levels was similar across most regional groups, with three exceptions. Parents from programs in the West and from the Migrant/Seasonal programs were less likely than parents in the other regional groups to have completed high school or a GED. Nearly 82 percent of parents in the Migrant/Seasonal programs had not completed high school or a GED. In contrast, parents in the American Indian and Alaska Native programs had higher levels of schooling compared with parents in other regions, as 77 percent of the parents had completed high school, a GED, or higher levels of education.

Table II.9. Percentage of Children Enrolled in Head Start by Highest Education Level of Parents, 2002-2003 Program Year

Education Level	All	Northeast (Regions 1, 2, and 3)	South (Regions 4 and 6)	North Central (Region 5)	Mountain and Plains (Regions 7 and 8)	West (Regions 9 and 10)	Migrant and Seasonal Programs	American Indian and Alaska Native Programs
Less than high school graduate	32	27	32	27	29	42	82	23
High school graduate or GED	45	50	46	47	41	36	15	50
Some college, vocational school, or associate's degree	19	18	19	21	25	19	3	22
Bachelor's or advanced degree	4	5	3	4	5	3	0	5

Source: Calculations based on 2002-2003 Head Start PIR data submitted by 1,969 Head Start programs and processed by Xtria, LLC, on February 12, 2004. Early Head Start programs were excluded from these calculations.




4. PROGRAM MONITORING AND OVERSIGHT ACTIVITIES

As funding for Head Start grew throughout the 1990s, policymakers sought not only to expand enrollment, but also to enhance the quality of Head Start services and to increase program accountability (U.S. Department of Health and Human Services 1993). For example, during the 1990s, the Head Start Program Performance Standards, which establish minimum levels of service provision and quality for Head Start programs, were revised and strengthened. Some of the initiatives, for example, the National Reporting System assessments of all four-year-old children in Head Start, have involved collecting data that has the potential to inform ideas for quality enhancements, or that can be used as part of an evaluation. Other initiatives have demonstrated the capacity of the Head Start Bureau to implement quality enhancement ideas on a national scale. In 2002, the Head Start Bureau sponsored the Strategic Teacher Education Program (STEP), an initiative to impart to every Head Start teacher techniques and activities designed to promote children’s emergent literacy skills. Head Start’s training and technical assistance system has been redesigned; it could play a role in implementing and monitoring initiatives identified for field testing. This section discusses the initiatives and their potential applicability to evaluations of quality enhancement initiatives.

In response to the Government Performance and Results Act (GPRA; P.L. 103-62) and the Head Start Act reauthorization in 1994 (42 USC 9831 et seq.), Head Start also began developing specific performance indicators in 1995. In 1997, the Head Start Bureau launched the Family and Child Experiences Survey (FACES) to collect data on the indicators (described in detail in the next section). These indicators cover all the major Head Start program objectives, including program management and parents’ involvement; the provision of educational, health, and nutritional services for children; the ability to link children and families with community services; family economic well-being, health, and parenting; and children’s school readiness (U.S. Department of Health and Human Services 1995; 2001; 2003a). As we discuss in Section E, FACES has provided extensive data on a random sample of programs. Although they are purely descriptive, the data provide insights into program characteristics associated with higher classroom quality and more favorable child outcomes. These insights suggest potential ideas for quality enhancements to be developed and tested.

The Coats Human Services Reauthorization Act of 1998 (P.L. 105-285) amended the Head Start Act with new provisions designed to further strengthen the program’s quality and accountability. The new law required the Secretary of Health and Human Services to establish new educational performance measures related to school readiness, and to ensure that the measures were “results-based” and adaptable for use in peer review and program evaluation. Moreover, program monitoring was to be expanded to include determining whether programs are meeting results-based performance measures.

To guide the development of program assessments of children, the Head Start Bureau convened the Technical Work Group on the Assessment of Program Outcomes, which advised ACF on the development of the Head Start Child Outcomes Framework (U.S. Department of Health and Human Services 2003b). This framework consists of eight general domains of early learning and child development, 27 domain elements, and examples of specific indicators of children’s skills, abilities, knowledge, and behavior—including the four specific congressionally mandated domains and nine mandated indicators of early literacy, language, and mathematics skills (see Appendix A). Since 2003, each Head Start program has been required to assess children throughout the Head Start program year on the congressionally mandated domain elements and on some aspect of each of the eight domains of child learning and development (U.S. Department of Health and Human Services 2000a and 2000b). Programs select their own measures for conducting these assessments. Programs use this information on child outcomes in their self-assessment and continuous program improvement activities. Although the types of assessments that programs choose typically vary, some states (for example, California) have begun to require programs to use specific measures, which could help to make programs a source of outcome data for evaluations.

The 1998 Head Start reauthorization also required that, by September 2003, at least 50 percent of Head Start teachers in center-based programs have an associate’s, bachelor’s, or advanced degree in ECE, child development, or a related field, as well as experience teaching preschool children. A General Accounting Office (GAO) study (2003) found that, based on Head Start PIR data, at the end of the 2002 program year, 52 percent of Head Start teachers nationwide had at least an associate’s degree in ECE or a related field. GAO could not determine whether each classroom had a teacher with minimum credentials in 2002; on average, however, the percentage of teachers with minimum credentials across the 12 ACF regions increased by 14 percentage points during the period from 1999 to 2002.

In recent years, the Head Start Bureau has used several key tools to monitor programs’ compliance with the program performance standards, help programs to meet the standards, and help them to improve the quality of services. All Head Start programs participate in an on-site monitoring review at least every three years to assess compliance with the program performance standards. Monitoring teams use the Program Review Instrument for Systems Monitoring (PRISM), are led by staff from the ACF regional offices, and consist of peer and consultant reviewers with relevant expertise. The current PRISM form includes very few detailed, well-defined data items. Moreover, because of the three-year span between assessments, this data collection effort is less helpful for evaluation work than are other Head Start data sources.

In addition to monitoring, the Head Start Bureau provides training and technical assistance to help programs to meet standards and improve quality. Head Start’s training and technical assistance system has been redesigned, and implementation is under way. The new system, initiated on September 1, 2003, consists of 12 contracted centers that are managed by the ACF regional offices. In the 10 geographically based regions, two or three training and technical assistance managers and several content experts (in such areas as early literacy, disabilities, health, and administration) work out of the regional offices, with the goal of facilitating closer communication and coordination. In addition, training and technical assistance specialists work directly with a group of 12 to 15 Head Start grantees, under the supervision of the training and technical assistance managers. According to Head Start Bureau staff, the new training and technical assistance managers and other staff are a mix of experts who were part of the previous training and technical assistance system, former Head Start program staff, and other experts in priority areas. The training and technical assistance staff provide a potential source of assistance in implementing initiatives that are ready for a broad field test.

In addition to the new staffing structure for training and technical assistance, the Head Start Bureau plans to launch an extensive on-line network of resources for training and technical assistance staff and programs. The intent of this on-line network is to ensure that all staff members providing support to programs have access to the same set of high-quality resources.

The Head Start National Reporting System (NRS), implemented for the first time in fall 2003, is a child assessment system to collect uniform outcome data on all four- and five-year-olds in Head Start who are expected to enter kindergarten the following year. Under the NRS, all Head Start programs assess their four- and five-year-olds at the beginning and end of the program year on a selected set of language, literacy, and mathematics indicators, using a standard set of assessments. The Head Start Bureau and regional offices will use the findings to identify areas in which programs need additional training and technical assistance to more effectively support the performance measures in the 1998 reauthorization.

The NRS potentially could provide baseline data for small- or large-scale evaluations, as well as follow-up data, when relevant, for large-scale field tests of quality enhancement initiatives. However, NRS data are not currently available for research purposes, nor are they available below the grantee level. Obtaining center- or classroom-level NRS data is likely to be critical for evaluating quality enhancement initiatives for a field test (Stage 3), and child-level data will be necessary to compute impact estimates in smaller-scale Stage 2 evaluations. A study of the implementation and system improvement of the NRS is ongoing, including a Technical Working Group that reviews current progress and recommends system improvement activities. This group is currently considering a field test of social-emotional measures, potential sampling options, and methods for assisting Head Start programs in interpreting and using the results of the NRS and other assessments. It is worth noting that a new DHHS Secretary’s Advisory Committee has recently been appointed, focusing particularly on assessing progress in the development and implementation of the NRS. This group is also charged with considering ways to integrate the NRS into broader assessments of early childhood learning found in FACES, the Head Start Impact Study, and the Child Outcomes Framework.

5. CURRENT HEAD START RESEARCH AND DATA

Rigorous evaluation of quality enhancement ideas and program variations in Head Start will build in important ways on the existing frameworks for Head Start research—the blueprint for Head Start research and the plan for evaluating Head Start impacts, set forth by two advisory committees (U.S. Department of Health and Human Services 1990 and 1999). In addition to emphasizing the importance of studying overall impacts of Head Start, the two advisory committees also stressed the importance of determining which program practices work best to support children’s development, and for which families and children the programs are most effective. In this section, we review recent and ongoing major research efforts on Head Start to illustrate the scope and purpose of existing research efforts, and to explain how the current design effort would help to fulfill the vision for Head Start research described by the two research advisory committees.

Our review is based on a discussion of four research efforts, each of which focuses on a subset of Head Start center-based programs. Given that many of these research efforts still are under way, the design for a study of Head Start quality enhancement and program variations will have to take into account the potential for “research fatigue” among programs selected for multiple studies. Ongoing research efforts could complicate the task of identifying programs and centers willing to participate in a study of Head Start quality enhancement. Nevertheless, in light of the large number of Head Start centers and the potential attractiveness of the offer of targeted assistance to implement a quality enhancement initiative, this challenge should not be an insurmountable one.

1. Family and Child Experiences Survey

FACES, launched in 1997, is a national longitudinal study of the characteristics, experiences, and outcomes of Head Start children and families. The survey was designed to provide information for the Head Start Program Performance Measures Initiative. The performance measures define the type and quality of services that programs must provide in order to successfully meet the goals and objectives of Head Start. Measures cover all of the major program objectives, including program management and parents’ involvement; the provision of educational, health, and nutritional services for children; the ability to link children and families with community services they need; families’ economic well-being, health, and parenting; and children’s school readiness (U.S. Department of Health and Human Services 2003a). Data collection includes assessments of the quality of the classroom environment and measures of children’s emergent literacy, mathematics skills, social skills, and problem behaviors.

The FACES study of Head Start programs and children is based on a nationally representative sample of center-based programs from the 10 ACF regions.6 The sample is drawn in two stages. For the fall 2000 cohort, a sample of programs stratified by Census region, percentage minority, and urban/rural status was drawn. This selection yielded 43 programs. Within those programs, a sample of classrooms was drawn, yielding 286 classrooms. All 2,790 children in the selected classrooms were eligible for the study. Sample selection for the earlier FACES cohort (1997) and for the most recent FACES cohort (2003) was conducted in a similar fashion. Forty programs were included in the sample for the 1997 cohort, and 60 programs were included in the 2003 cohort. The previous FACES study followed the 1997 cohort of three- and four-year-old Head Start children from program entry through first grade. Children from the 2000 cohort were followed through kindergarten (U.S. Department of Health and Human Services 2001 and 2003a), and children from the 2003 cohort are also being followed through kindergarten.

The FACES battery includes a child assessment, parent interview, teacher and staff interviews, and classroom observations. Children are directly assessed on vocabulary, emergent literacy and early mathematical skills. In addition, their social skills and problem behaviors are assessed through rating scales completed by parents and teachers. Parents are interviewed about parenting behaviors, the socioeconomic characteristics of the family, and parental health. Interviews with classroom teachers, center directors, program directors, and coordinators gather data on the staffs’ experience, education, and training, as well as on their attitudes about early childhood education practices and activities with children and parents. Classroom observations collect data on the structure of the classroom, the overall quality of the classroom environment, and teacher-child interactions (U.S. Department of Health and Human Services 2001 and 2003a).

The FACES studies provide a nationally representative, descriptive picture of center-based Head Start programs, and a longitudinal description of children’s development during the Head Start year and into kindergarten. The majority of Head Start programs were rated as good on a global measure of the quality of the classroom environment. Children achieved substantial gains (of one-third to more than one-half a standard deviation) in vocabulary, mathematics, and writing skills during the Head Start year. Moreover, scores at the end of the Head Start year and the size of the gains in scores during the year predicted achievement at the end of kindergarten. Although the study has not been designed to provide estimates of the contribution of staff qualifications or aspects of program quality to improvements in children’s achievement, it does suggest a relationship among staff qualifications, program quality, and children’s development.

2. The National Head Start Impact Study

In 1998, as part of Head Start’s reauthorization, Congress instructed DHHS to conduct a national study to measure the impact of Head Start on the children served by the program and took into account the recommendations of an advisory panel in its design (U.S. Department of Health and Human Services 1999). The National Head Start Impact Study includes nearly 5,000 three- and four-year-old preschool children who applied to 383 centers from 84 grantee/delegate agencies across the country. The children in the study have been randomly assigned to receive Head Start services (the treatment group) or not receive Head Start services (the control group).7

The impact study has two main objectives. The first is to estimate the impact of Head Start on the school readiness of children receiving Head Start services compared with children who are not enrolled in Head Start. The second objective is to investigate the conditions under which Head Start works best, and the children for whom the program has the greatest impacts. The impact study will examine various factors that may differentially contribute to Head Start’s impacts, including differences among children attending Head Start, differences in children’s home environments, variations in the types of Head Start programs available, and the availability and quality of other child care and preschool programs in a particular area that could influence outcomes for control-group children (Lopez et al. 2002).

Data collection for the study began in fall 2002 and continues through 2006. Data sources include parent interviews, child assessments, surveys of child care providers and teachers, direct observations of care settings, and teachers’ ratings of children. Children will be followed through the spring of their first grade year. The study will provide a nationally representative picture of the quality of Head Start classroom settings for three- and four-year-old children and a similar picture of the quality of early childhood settings (from other preschools to home-based child care) used by children who did not attend Head Start. The study will yield estimates of the impact of Head Start on children’s achievement into the early school years relative to the impact of other early childhood care experiences that children would have in the absence of Head Start. Subgroup impact estimates will indicate whether the impacts of Head Start are different for key subgroups and for Head Start programs with different characteristics.

The random assignment design can yield strong evidence about the impacts of Head Start on children’s achievement relative to other early childhood care options in the community in the absence of Head Start; however, evidence about the impacts of different program approaches and practices (based on subgroup analyses) is weaker. The subgroups are formed by grouping programs that have particular characteristics in common, and, even if differential impacts are found, they could stem from some other factor common to the group of programs that was not identified but has driven the results. Consequently, to rigorously examine the question of whether a particular promising program practice has stronger impacts on children’s development than do the basic Head Start program services, it is necessary to use a design such as we describe in this report, in which centers or classrooms implementing a new approach are compared with centers or classrooms offering regular Head Start Services.

3. Head Start Quality Research Center Consortium and Data Coordinating Center8

The Head Start Quality Research Center (QRC) Consortium originally was funded to explore research questions related to program quality and program outcomes in Head Start. The first QRC Consortium (funded from 1995 through 2000) included four teams of university-based researchers or consultants who worked with one or more Head Start program partners to study the relationships between aspects of program quality and child and family developmental outcomes. QRC Consortium members also served as technical advisors on the design, development, and implementation of program performance measures, including the first Head Start Family and Child Experiences Survey (FACES).

The current QRC Consortium was funded for five years beginning in March 2001 with the goal of developing interventions to promote the school readiness of preschool children in Head Start. QRC Consortium members include the same researchers and programs involved in the first QRC Consortium, plus four additional partnerships. Many of the researchers expanded their partnerships to include more Head Start programs. Interventions include the types of quality improvements and promising practices about which the Head Start community is eager to learn more, including a stronger implementation of a general classroom curriculum, approaches to enhancing children’s language and emergent literacy development, approaches to enhancing children’s social-emotional well-being and behavioral development, individualizing of instruction for children, and enhancement of parents’ ability to teach their children at home.

The plan for the QRC Consortium project has called for the researchers to implement the interventions on a pilot basis during the first year, and to collect outcome data on children during the fall and spring of that program year. Subsequently, using insights gained from the first year of implementation, the researchers are to replicate the interventions in several other program sites, and, using a random-assignment design, are to evaluate the effectiveness of the interventions. Grantees are funded to continue these replications for several additional years, in different programs. The pilot implementation year yielded many important lessons that were used to modify the design of the interventions and the strategies for implementing the interventions for the replication year. The QRC Consortium’s experience supports the need for pilot implementation of an intervention to learn practical lessons about the feasibility of implementing the intervention on a larger scale in the context of an evaluation.

ACF established the Data Coordinating Center (DCC) to systematically collect cross-site data for the QRC Consortium, in addition to the data collection carried out by the QRC researchers. The DCC provides pre- and post-intervention data collection and analysis on a core set of cross-site measures, such as program quality, child outcomes, and parents’ involvement. A consistent set of measures has been used over several replication years to maximize the analysis sample. The data collected by the DCC enable researchers to compare findings from the QRC studies with those from the national FACES sample. They also provide a mechanism for comparing site-specific instruments with the FACES instruments.

The QRC studies are an important forerunner to the current effort to design a framework for Head Start research on quality enhancements and program variations. Their fundamental approach—initially implementing an intervention to a high degree of fidelity on a pilot basis, and then implementing it with a rigorous experimental design and measuring children’s outcomes—is a strong one for studying the impacts of promising practices in Head Start. Because many of the QRC evaluations have included relatively small samples, their most important contribution may be to strengthen development of the enhancement idea and providing important lessons on the value of implementation partnerships. The QRC projects could thus offer a set of ideas and implementation strategies for Stage 2 evaluations.

4. Other Studies of Promising Practices that Include Head Start Programs

Several other studies funded by the U.S. Department of Education (ED) and a consortium of agencies including the National Institute for Child Health and Human Development (NICHD), ACF, the Office of the Assistant Secretary for Planning and Evaluation (DHHS), and the Office of Special Education and Rehabilitative Services (ED) are examining the impacts of promising practices in preschool programs on children’s school readiness. All of these projects include Head Start programs, although many other types of preschool programs also are included in the samples. The studies include program evaluations that are similar in design to the Head Start impact study, as well as evaluations of promising practices (primarily curricula) that are similar in design to the QRC Consortium projects.

The NICHD consortium study, called the Effectiveness of Early Childhood Programs, Curricula, and Interventions in Promoting School Readiness study, includes eight grantees, some focusing on or including Head Start programs. The grants are given to study the effectiveness of early childhood interventions and programs across a variety of early childhood settings in promoting school readiness for children, from birth through age five, who are at risk of later school difficulties. The interventions include several teacher training and curriculum studies (focusing on such areas as children’s language development, emergent literacy, mathematics skills, and social-emotional well-being) and a parent-focused intervention to enhance parent-child interactions. Researchers will use rigorous designs when implementing the interventions and will measure children’s development subsequent to the children’s exposure to the intervention (or control) early childhood environments. However, there is not a common, cross-site core of measures being administered by a data coordination center.

Head Start programs also are among those included in ED’s evaluations of alternative preschool curricula (the Preschool Curriculum Evaluation Research [PCER] project). This project includes 12 grantees selected during two rounds of grant competitions that are charged with implementing and rigorously testing alternative preschool curricula. Two national coordinating contractors, one for each PCER cohort, are collecting cross-site data on classroom quality and children’s outcomes and will compare the treatment curricula with prevailing practice, estimate overall impacts of the treatment curricula on a common set of child outcomes, and analyze subgroup impacts. Most PCER grantees have implemented the treatment curricula and are in their first year of data collection on children in preschools.

The Early Reading First evaluation will estimate the impacts of a new funding stream from ED designed to enhance the language and literacy environment of preschool classrooms on children’s language development and emergent literacy. Grantees can use the funds for a variety of purposes, including teacher education and training and the purchase of books and materials. Sixty grantees (each with approximately five preschool programs) were selected during two rounds of grant competition. All of them were expected to implement their Early Reading First programs by January 2004. Head Start programs comprise about 20 percent of the preschool programs receiving Early Reading First funds. A national evaluation contractor will select programs receiving Early Reading First funds and will compare the language development and emergent literacy skills of children enrolled in Early Reading First with those of children who are not enrolled in Early Reading First programs. Preschool programs receiving Early Reading First funds and those without such funds will be compared on the basis of observational assessments of the language and emergent literacy environments of the preschool classrooms.

ED’s Even Start program is designed to enhance the literacy skills of two generations—the parent and child—through an intervention focusing on adult literacy instruction; the enhancement of parent-child interactions; and center-based, child-focused literacy activities. Families are eligible if they have a child between three and eight years of age. Three previous evaluations did not find any impacts of the Even Start program on child outcomes (St. Pierre et al. 1995; Tao et al. 1998; St. Pierre et al. 2003). The current evaluation has been designed to implement and test alternative program enhancements. Two classroom curricula and two parent education curricula (focusing on parent-child interactions around language and literacy activities) are being implemented and evaluated using a rigorous design. The Classroom Literacy Interventions and Outcomes evaluation will estimate the impacts of the alternative classroom curricula (relative to current Even Start program practices) and of the additional impact of each of the parent education curricula on children’s language development and emergent literacy. Several Head Start programs offer Even Start services and are included in this study. Programs that agree to participate in the study will be randomly assigned to implement one of the classroom curricula or a classroom curriculum and a parent education curriculum, or to implement none of these curricula, but continue operating under their current designs.

6. LESSONS FROM PREVIOUS HEAD START PLANNED VARIATION STUDIES

One historic set of Head Start research initiatives that can inform the present design of research on Head Start quality enhancements is the Head Start Follow Through and Planned Variation studies of the 1960s. Like the current efforts to learn what works best in Head Start programs, these initiatives sought to address the question of how best to enhance the educational attainment of disadvantaged children participating in Head Start. Follow Through was conceived as an ongoing program that would provide additional education from kindergarten through grade three to children coming out of Head Start programs in an effort to preserve the gains children made during the Head Start year. When funding for Follow Through as an ongoing program was not forthcoming, policymakers decided to establish the program on a smaller scale, and to use it to test promising educational practices. A group of communities was invited to participate in the Follow Through study and was offered a choice of program models to implement. A year after Follow Through began, some communities already participating in that program were invited to participate in the Head Start Planned Variation study. If they agreed to participate, they were asked to implement in their Head Start program the Follow Through program models used in their communities.

More than 20 program models were tested in one or both of these studies. The models were extremely diverse. Some focused on promoting skills through drills (such as the Engelmann-Becker model), whereas others focused on fostering children’s development through experimentation, exploration, and verbalization. Still others focused on bilingual educational approaches. Some tested models that focused on parents’ involvement in the schools.

Policymakers and researchers alike concluded that these planned variation studies provided little new information; methodological weaknesses in the evaluations shed doubt on the findings (Rivlin and Timpane 1975). Yet, the lessons that emerged about the best approaches to designing research to accompany the development of new ideas for serving Head Start children remain useful today. These lessons provide critical insights for current efforts to design a research approach that will enable the Head Start community to identify and create better approaches to supporting the development of economically disadvantaged children participating in Head Start. Those lessons include:

• Clarify the Objectives of the Study for All Stakeholders. Perhaps because Follow Through originally was viewed as an ongoing program, many stakeholders were unclear about the differences between a service program and an evaluation. For example, many communities instituted large-scale adaptations to the program model, so that it was difficult to interpret the findings. Use an Experimental Design. Communities were not randomly assigned to models in either study. Many of the schools and Head Start centers used as comparison groups were chosen from different communities and frequently served a less disadvantaged population. Consequently, it was unclear whether the findings from the studies were due to the intervention or to the differences in the characteristics of families in the treatment and control groups.

• Use an Experimental Design. Communities were not randomly assigned to models in either study. Many of the schools and Head Start centers used as comparison groups were chosen from different communities and frequently served a less disadvantaged population. Consequently, it was unclear whether the findings from the studies were due to the intervention or to the differences in the characteristics of families in the treatment and control groups.

• Fully Develop and Define the Interventions Before Testing. Time constraints precluded full development of some of the models tested in Follow Through and Head Start Planned Variation at the time they were implemented. Much of the translation from theory into practice occurred during the first year of the evaluation. Because the key elements of the models were not well defined, interpretation of the findings was difficult.

• Provide Considerable Training and Technical Assistance. Teachers participated in a short workshop and then were asked to implement the model in the classroom. Many became overwhelmed and frustrated. In addition to the more theoretical underpinnings of the model, teachers need practical advice about the arrangement of the physical class space and the structure of the day. Curriculum assistants who observed the classes and provided feedback to the teachers were viewed as an essential component of implementation.

• Assess the Success of Implementation. The Head Start Planned Variation studies were criticized as having inadequate measures of how well the program model was implemented. As a result, it was not clear whether the absence of any effects was due to the program model or to problems with the implementation of the model.

• Use Outcome Measures that Can Discriminate Among Models. The outcome measures used in the early Head Start Planned Variation studies were criticized for being insufficiently tied to the objectives of the program models. The outcomes focused excessively on cognitive measures and ignored measures of other aspects of development that were the focus of some of the models. The meeting participants criticized the standardized measures used, such as the Metropolitan Achievement Test, because they were designed to give comparable measures of children's performance irrespective of the curriculum, and therefore were not good discriminators among the effects of different curricula. Outcome measures should represent key outcomes for all interventions, so that the relative success of different interventions can be compared.

This report builds on the lessons of both the early Head Start Planned Variations study and the more recent research efforts. The three stages of research are intended in part to support the development of quality enhancement ideas so that implementation is well-understood and fully established before any evaluation begins. Rigorous, experimental designs are proposed for the evaluations. Measurement plans include outcome measures that are relevant to the quality enhancements being tested. Partnerships with Head Start programs, staff, and families are critical elements for the overall success of the quality enhancement development and research process.