Long-Term Effects of Early Childhood Programs on Cognitive and ...

25

Long-Term Effects of EarlyChildhood Programs onCognitive and SchoolOutcomesW. Steven Barnett

Abstract

The extent to which early childhood programs produce long-term benefits in chil-dren’s cognitive development, socialization, and school success is a matter of somecontroversy. This article reviews 36 studies of both model demonstration projects andlarge-scale public programs to examine the long-term effects of these programs onchildren from low-income families. The review carefully considers issues related toresearch design. It includes studies of preschool education, Head Start, child care, andhome visiting programs, and focuses primarily on the effects of program participationon children’s cognitive development. Results indicate that early childhood programscan produce large short-term benefits for children on intelligence quotient (IQ) andsizable long-term effects on school achievement, grade retention, placement in specialeducation, and social adjustment. Not all programs produce these benefits, perhapsbecause of differences in quality and funding across programs. The article concludeswith recommendations for future action.

The contribution of early childhood care and education (ECCE) tothe healthy development and future well-being of children who areeconomically and socially disadvantaged has become a vital public

issue with important implications for families, business, private philan-thropy, and government. It will be shown through a detailed, critical reviewof research that public investments in quality early childhood care and edu-cation can produce important long-term improvements in the intellectualand social development of disadvantaged children. Unfortunately, becausethe United States underinvests in both the quantity and quality of early careand education, the nation forgoes many of the potential benefits at an annu-al cost estimated in billions of dollars.

The above conclusions could easily come as a surprise to those who arefamiliar with the debates that have swirled around child care and other

W. Steven Barnett, Ph.D.,is a professor at theGraduate School ofEducation at RutgersUniversity.

The Future of Children LONG-TERM OUTCOMES OF EARLY CHILDHOOD PROGRAMS Vol. 5 • No. 3 – Winter 1995

ECCE programs such as state-sponsored preschool education and the fed-eral Head Start program in the past several years. There is widespread agree-ment that ECCE programs can produce short-term gains in disadvantagedchildren’s performance on standardized tests of intelligence and academicability and that some preschool programs have reduced later grade reten-tion and special education placement. However, there has been consider-able disagreement about how these gains are produced, what they mean,whether they persist, and what other long-term consequences might beexpected.1–4

The disagreements regarding long-term effects are so extreme that it isimpossible to make public policy recommendations without resolving someof them. The key questions to be addressed for ECCE policy are as follows:

1. What are the effects of ECCE programs on the cognitive develop-ment, socialization, and school success of disadvantaged children? Howlong do they persist?

2. Are some types of ECCE programs more successful than others (forexample, home visits versus center-based programs; model interventionsversus large-scale programs such as Head Start or traditional child care; pro-grams that start at birth versus those that begin at age three or four years)?Do some children benefit more than others?

3. To what extent can the effects of model programs be generalized toexisting public and private programs?

These policy questions are addressed by conducting a critical review ofresearch on the effects of ECCE on disadvantaged children with an empha-sis on cognitive development. This review focuses on long-term effectsbecause they are the subject of greatest disagreement, and it includes a vari-ety of ECCE programs—child care, early intervention, preschool education,and Head Start. The article concludes with recommendations for futureaction.

26 THE FUTURE OF CHILDREN – WINTER 199526

Research on Immediateand Short-Term ProgramEffectsHundreds of studies have examined theimmediate and short-term (that is, within ayear or two after children exit a program)effects of ECCE programs of various types.These studies are found in two largely sepa-rate streams of research, one on the effectsof ordinary child care on children from allbackgrounds and the other on the effectsof ECCE programs specially designed toimprove the cognitive development of eco-nomically and otherwise disadvantagedchildren.

Initially, research on child care focusedon potential negative effects on the mother-child relationship and the child’s socializa-

tion, with less attention to cognitive develop-ment. More recently, child care research hasbegun to examine the effects of variations inboth the quality of nonparental care and thechild’s home environment and family cir-cumstances. Research on programs thatserved disadvantaged children first empha-sized their effects on cognitive development,particularly IQ, but has since expanded toexamine the effects of such interventions onother aspects of cognitive development andon socialization.

Short-Term Effects of Child CareThe child care research literature presentsno consistent evidence that child care per seis harmful to child development, regardlessof the age at which a child begins out-of-home care.5–6 It does, however, indicate thatvariations in the quality of child care are

27Long-Term Effects of Early Childhood Programs on Cognitive and School Outcomes

important determinants of the impact ofchild care. Higher quality child care is asso-ciated with better cognitive and social devel-opment both while children are in childcare and during their first few years ofschool.5–8 (See also the article by Frede inthis journal issue for a discussion of the com-ponents of quality care.)

A recent investigation found that age atentry to or years of experience in child careduring the preschool years influenced thereading and math achievement of childrenat ages five and six, but differently for chil-dren from high- and low-income homes.9For children from impoverished homes, ear-lier entry and/or more years in care pro-duced a larger effect on reading scores thanfewer years. Conversely, effects were negativefor children in the highest-income families.The key may be differences in the quality ofthe children’s home environments ratherthan income per se: children whose homeenvironments were very highly supportive ofcognitive development and socializationactually had lower scores if they had been incare outside their homes, while childrenwhose home environments were relativelypoor gained the most from outside care.

Short-Term Effects ofModel InterventionsSeveral reviews of the intervention researchliterature have been written over the pastdecade.10–12 The authors of those reviewsconclude that programs designed for disad-vantaged children, including those that areroutinely provided on a large scale, can pro-duce immediate boosts in IQ equivalent toabout eight IQ points. This is a meaningfulimprovement in cognitive ability and canhave important implications for children interms of academic performance and place-ments in special education classes.

Effects of similar magnitude were foundon preschool and kindergarten achieve-ment measures. Changes of somewhatsmaller magnitude were found for socio-emotional outcomes such as self-esteem, aca-demic motivation, and social behaviorimmediately after the end of the interven-tion. On average, these effects declined overtime and were negligible several years afterchildren exited the programs. However,some programs produced sizable gains thatpersisted at least into the first few years of

school for IQ, achievement, and school out-comes such as grade retention and specialeducation placement.

A variety of ECCE intervention approach-es (for example, one-to-one tutoring, half-day preschool education, and child care)—some emphasizing direct instruction andothers emphasizing child-initiated activi-ties—for disadvantaged children seem tohave worked equally well. However, themagnitude of the effects appears to be atleast roughly related to the intensity,breadth, and amount of involvement withthe children and their families.10 One spe-cific type of intervention, home visiting(alone or in combination with a center-based program), may be relatively ineffec-tive in directly improving children’s devel-opment, although it may be more successfulin improving maternal and child health,providing social support, or reducing childabuse and neglect.13–14

Since most of those reviews were writ-ten, four significant research projects havebeen undertaken to investigate the effectsof ECCE for disadvantaged children:Project CARE, the Infant Health andDevelopment Program (IHDP), EvenStart, and the Comprehensive ChildDevelopment Program (CCDP).14–15 Theprograms are all still too new to have gen-erated any long-term results, but theirshort-term results are fairly consistent with

those of earlier studies. Project CARE andIHDP, which involved high-quality center-based care, produced substantial immediateincreases in IQ and some benefits in childlanguage skills and child behavior. EvenStart and CCDP, which offered weaker earlychildhood services for children but addedparenting education or job skills trainingservices for parents, had smaller benefits onchild development but positive effects onoutcomes such as parental expectations forchildren’s academic success and enhancedutilization of other community social, health,and education services. (For additional

Programs designed for disadvantagedchildren can produce immediate boosts inIQ equivalent to about eight IQ points.


information about these programs, see thearticles by Yoshikawa and St. Pierre and col-leagues in this journal issue.)

Results of the two research literaturesappear to be converging: ECCE has impor-tant impacts on cognitive development andsocialization of disadvantaged childrenimmediately and in the short term. Effectsdepend on program quality, and cross-studycomparisons indicate that effects are largerfor well-designed, intensive ECCE interven-tions than for ordinary child care. Fromsome studies, it appears that the effects ofECCE programs decline over time once chil-dren leave ECCE.

Research on Long-TermProgram EffectsThe children in many of the studies includ-ed in the preceding summary of short-termeffects of ECCE have been followed overseveral years to generate information aboutthe long-term effects of ECCE. This articlereviews 36 such studies, identified throughcomputerized and manual searches of theresearch literature, which meet four crite-ria: (1) the ECCE program studied began atage four or earlier (thereby excluding stud-

ies of kindergarten programs); (2) the targetpopulation for the ECCE program studiedwas children who were economically disad-vantaged; (3) at least one aspect of cognitivedevelopment, school progress, or socializa-tion was measured after age eight (thirdgrade or later); and (4) the research designemployed a no-treatment comparisongroup that was reasonably similar to thegroup of children who participated in theintervention.

The 36 studies were divided into two cat-egories for review based on the nature of theECCE program and the research design. In15 studies,16–30 researchers developed theirown ECCE programs to study the effects ofexemplary, or model, programs. In 21 stud-ies,31–51 researchers investigated the effects of

ongoing, large-scale public ECCE programs.Five large-scale studies examined state orlocal programs (including some supportedwith federal Title 1 funding),31–35 five studiedchildren who had attended Head Start andstate or local programs,36–40 and elevenexamined Head Start programs.41–51

Model ProgramsThe 15 studies of model programs are iden-tified and described in Table 1. Generally,these model ECCE programs were probablyof higher quality than the large-scale publicprograms. They may have had more highlyqualified staff, closer supervision of staff byexperts, lower child-staff ratios, and smallergroup size. These advantages were madepossible by much higher levels of fundingper child than are available to typical HeadStart and public school programs. (See alsothe article by Frede in this journal issue.)

Services Offered The model programs varied considerably inthe services offered, their onset and dura-tion, and when the programs operated(1962 to 1980). All but one (Verbal In-teraction Project) provided classroom ser-vices. Most also offered home visits. Pro-grams offered a range of other activities forparents, and, in three instances, these wereextensive enough that they might be calledparent support and development pro-grams.17,19,21

Study Participants In all but one study, most of the participa-ting children were African American. Theexception, the Houston Parent Child De-velopment Center, served Hispanic families.All but one program served boys and girls;the Harlem Training Project served onlyboys. All but two studies did not select par-ticipants on the basis of IQ: the PerryPreschool study selected children based onlow IQ scores (generally below 85), and theMilwaukee program served children whosemothers had low IQs (below 75).

The average level of mother’s educationwas under 12 years in all studies, and under10 years in five studies.17–19,23,26

Most of the children in comparison orcontrol groups began formal education atkindergarten, but, especially in the studiesconducted in later years when alternative

The model ECCE programs were probablyof higher quality than the large-scale publicprograms.


services were more widely available, a signifi-cant percentage of children in comparisongroups could have attended preschool orchild care programs.

Large-Scale ProgramsThe 21 studies of large-scale public ECCEprograms are identified and described inTable 2.

Services Offered None of the large-scale programs enrolledchildren before age three, and most servedchildren part day for one school year at agefour. Typically, the studies do not describethe programs so that one must depend ongeneral knowledge about public schoolpreschool and Head Start programs tounderstand the services offered. The HeadStart programs had broader missions thanmost of the other programs; their goalsincluded improving health and nutrition,and providing services to parents and thecommunity.52 However, Head Start pro-grams tend to have larger group sizes, payteachers much more poorly, and may pro-vide classes only part year.53–54 Classroomstend to serve 15 to 20 children with ateacher and an aide, but there is consider-able variation in the quality of educationalexperiences within these classrooms and inthe parent involvement and other servicesprovided.53–55

In three large-scale program studies,children who participated in the ECCEprograms also participated in other ser-vices. In the Cincinnati Title I study, mostfull-day kindergarten students had attend-ed preschool and most half-day kinder-garten students had not. In both ChildParent Center (CPC) studies, servicesbegan in preschool and continued to pro-vide enriched educational experiencesthrough third grade. In the FloridaLearning-to-Learn and Head Start study,children received intensive servicesthrough first grade.

Study Participants In all the large-scale studies, children andtheir families were low income. There is con-siderable ethnic and geographic diversityacross studies. Most children are minorities,from single-parent families, and most par-ents are less educated, but substantial num-bers of white children, two-parent families,

and parents who are high school graduatesalso participated.

Research DesignThe best research studies are those that areso tightly designed that one can concludewith confidence that the results obtained aredue only to the intervention. In the ECCEliterature, as in most, there is considerablevariability in the quality of the researchdesign, especially with respect to four keyaspects of research design: (1) the ways in

which the comparison groups were formed,(2) initial and follow-up sample sizes, (3)attrition, and (4) who was measured andhow to assess effects of the program. Each ofthese aspects of study quality has importantimplications for the interpretation of studyresults. These implications are discussedbriefly below, and the methodological con-cerns associated with each of the studies arelisted in Tables 1 and 2.

Formation of Comparison GroupsGenerally, the strongest research designinvolves identifying a pool of potential par-ticipants and then randomly assigning somechildren to an experimental group andsome to a control or comparison group.This increases confidence that estimatedeffects in these studies are due to the pro-gram rather than to preexisting differencesbetween program and comparison groups.

If random assignment is not employed,then researchers usually attempt either (1)to construct a comparison group, matchedas closely as possible on a number of charac-teristics thought to be relevant (for example,maternal education, family income level,ethnic or racial background), or (2) to con-struct a comparison group but then also usestatistical techniques to control for initial dif-ferences on key characteristics. Unfortu-nately, in neither technique is it possible toknow with certainty that one has matchedor controlled for all the key characteristics,and that is why random assignment, whichpresumably equalizes the groups initially, is

Most of the large-scale programs servedchildren part day for one school year atage four.


Program Name/ Program Description Ages of Participation Research Design/Related Endnote Numberb Methodological Concerns

(Years of Operation)

Carolina Abecedarian16 Preschoolers: full-day Entry: 6 weeks to Randomized.child care 3 months

(1972–1985)Schoolage: parent Exit: 5 to 8 years

program

Houston Parent Child Home visits Entry: 1 to 3 years Randomized.Development Center17 Full-day child care High attrition.f

Exit: 3 to 5 years(1970–1980) Center-based program

for parents

Florida Parent Home visits Entry: 3 to 24 months Initially randomized with one group,Education Project18 Twice weekly part-day and additional control group

preschool (ages 2 to Exit: 3 years members added at 24 months.(1966-1970) 3 years) Not randomized.g High attrition.

School-administered tests.h

Milwaukee Project19 Full-day child care Entry: 3 to 6 months Groups of 3 to 4 children assignedalternately to E and C groups.

(1968-1978) Job and academic Exit: 5 years Small sample.itraining for mothers

Syracuse Family Home visits Entry: 6 months Matched comparison group Development Research Full-day child care selected at 36 months.Program20 Exit: 5 years Not randomized.

(1969-1975)

Yale Child Welfare Home visits Entry: Prenatal Two comparison groups for sameResearch Program21 Full-day child care neighborhoods for first follow-up.

Pediatric care Exit: 30 months Matched comparison group (1968-1974) Developmental selected for follow-up at 30 months.

screenings Not randomized. School-administered tests.

Curriculum Comparison Part-day preschool Entry: 4 years Post hoc comparison group from Study22 program original pool.

Kindergarten program Exit: 5 or 6 years Not randomized. School-(1965-1967) administered tests.

Table 1

Model Early Childhood Programsa

Notes a Programs are grouped such that those enrolling children younger than three years old appear first, followed by those enrolling children after age three.b See the related endnotes at the end of this article for complete citations of the reports and/or studies in which program methods and outcomes are described.c Throughout Table 1, E refers to the experimental or intervention group, and C refers to the control or comparison group.d IQs were measured using the WISC or WISC-R, unless otherwise noted.e Outcomes listed as E > C or E < C were statistically significant at the p < .05 level, at least (that is, likely to have occurred by chance no more than 5 times in 100).

In some instances, the difference between the E and C groups was fairly large but not statistically significant, perhaps because of small sample sizes. Such outcomes are indicated as “E = C, but positive trend.”

f Results may be biased because of high attrition rates.g Results may be biased because children were not randomly assigned to experimental and control or comparison groups.h Results may be biased because school-administered tests were used to measure achievement.i The small initial sample makes it difficult to demonstrate statistically significant effects of the program.


Initial Sample Follow-up Time of IQd,e School Outcomese

Sizec Sample Size Follow-up

E = 57 Age 8 8, 12, and 15 Age 12: E > C Achievement tests: E > C at age 15C = 54 E = 48 years E = 93.7

C = 42 C = 88.4 Special education: E< C at age 15:E = 24%, C = 48%

Age 15 Age 15: E = CE = 48 E = 95.0 Grade retention: E < C at age 15:E = 44 C = 90.3 E = 39%, C = 59%

E = 97 School data Grades 2 to 5 Not measured Achievement tests: E = C, but positive trendC = 119 E = 50

C = 87 Grades: E = C

IQ data Bilingual education: E < CE = 39 E = 16%, C = 36%C = 78

Special education: E = C in grades 2 to 5E = 27%, C = 31%

Grade retention: E = C in grades 2 to 5E = 16%, C = 29%

E = 288 E = 83 Grades 4 to 7 E = C (grades Math achievement: E > CC = 109 C = 24 4 to 7) Reading achievement: E = C

E = 83.1 Special education: E < C, grade 7C = 79.8 E = 23%, C = 54%

Grade retention: E = C, grade 7E = 28%, C = 29%

E = 20 E = 17 Grade 4 Grade 8: E > C Achievement tests: E = C, but positive trendC = 20 C = 18 Grade 8 E = 101

C = 91 Grades: E = C

Special education: E = CE = 41%, C = 89%


E = 82 Parents Grades 7 to 8 E = C, age 5 on Teacher ratings: E > C, but for girls onlyC = 72 E = 52 Stanford-Binet

C = 42 Grades: E > C, but for girls only

Children Attendance: E > C, but for girls onlyE = 49C = 39

E = 18 Age 7 to 8 Age 7 to 8 and E = C at age 10 Achievement tests: E = CC = 18 E = 17 age 10

C1 = 33 Attendance: E > CC2 = 31

Teacher ratings: E = C, but positive trend for boysonly

Age 10E = 16 Special education: E = C, but positive trend C = 16 for boys only

E = 25%, C = 50%

E = 244 E = 168 Post high school Not measured Special education: E= C, grade 12C = 68 C = 51 E = 32%, C = 63%


High school graduation: E = C

Program Name/ Program Description Ages of Participation Research Design/Related Endnote Numberb Methodological Concerns

(Years of Operation)

Early Training Project23 Home visits Entry: 4 to 5 years Randomized.Summer part-day School-administered tests.

(1962–1967) preschool program Exit: 6 years

Experimental Variation Preschool program Entry: 4 years Post hoc comparison group fromof Head Start24 same communities.

Exit: 5 years Not randomized. High attrition. (1968–1969) School-administered tests.

Harlem Training Project25 One-to-one tutoring or Entry: 2 to 3 years Comparison group recruited fromchild-directed play children born 1 to 2 months later.

(1966-1967) Exit: 4 years Not randomized. School-administered tests.

High/Scope Perry Home visits Entry: 3 to 4 years Randomized.Preschool Project26 Preschool program

Exit: 5 years

(1962-1967)

Howard University Preschool program Entry: 3 years Comparison group fromProject27 neighboring tracts.

Exit: 5 years Not randomized.(1964-1966)

Institute for Home visits Entry: 4 years Randomized.Developmental Studies28 Part-day preschool High attrition. School-

program Exit: 9 years administered tests.(1963-1967) Parent center school (K-3)

Philadelphia Project29 Home visits Entry: 4 years Matched comparison group from Part-day preschool same kindergarten classes.

(1963-1964) program Exit: 5 years Not randomized. School-administered tests.

Verbal Interaction Home visits Entry: 2 to 3 years Six groups with three matchedProject30 comparison groups.

Exit: 4 years Not randomized.(1967-1972)

Table 1 (continued)

Model Early Childhood Programsa

Notes a Programs are grouped such that those enrolling children younger than three years old appear first, followed by those enrolling children after age three. b See the related endnotes at the end of this article for complete citations of the reports and/or studies in which program methods and outcomes are described.c Throughout Table 1, E refers to the experimental or intervention group, and C refers to the control or comparison group.d IQs were measured using the WISC or WISC-R, unless otherwise noted.e Outcomes listed as E > C or E < C were statistically significant at the p < .05 level, at least (that is, likely to have occurred by chance no more than 5 times in 100).



Initial Sample Follow-up Time of IQd,e School Outcomese

Sizec Sample Size Follow-up

E = 44 E = 36 Post high school E = C at age 17 Achievement tests: E = C C = 21 C = 16 E = 78.7

C = 76.4 Special education: E< C, grade 12E = 5%, C = 29%

Grade retention: E = C E = 58%, C = 61%

High school graduation: E = CE= 68%, C = 52%

E = 116 E = 102 Post high school E > C at age 13 Achievement tests: E = C, butC = 24 C = 19 E = 85.0 positive trend

C = 91.0Special Education: E = C, grade 7

E = 13%, C = 15%


E = 244 E = 168 Grade 7 E = C at age 12 Math achievement: E > CC = 68 C = 51 E = 92.1 Reading achievement: E < C

C = 88.9Grade retention: E < C, grade 7

E = 30%, C = 52%

E = 58 E = 58 Post high school E = C at age 14 Achievement tests: E > CC = 65 C = 65 E = 81.0

C = 81.0 Grades: E > C

Special education: E = C, grade 12E = 37%, C = 50%


High school graduation: E > CE = 67%, C = 49%

E = 38 E = 30 Grade 4 Not measured Grade retention: E = CC = 69 C = 69 E = 33%, C = 47%

E = 312 E = 63 Grade 7 Not measured Special education: E = CC = 191 C = 34 E = 0%, C = 13%

Grade retention: E = CE = 23%, C = 43%

E = 60 E = 44 Post high school E > C at age 10 Achievement tests: E = C, but positive trendC = 53 C = 37 on Stanford-

Binet Special education: E = C, grade 12E = 98.4 E = 5%, C = 6%C = 91.7


E = 111 E = 79 Grade 3 E > C at grade 3 Achievement tests: E > CC = 51 C = 49 E = 101.9

C = 93.6 Special education: E > C, grade 7E = 14%, C = 39%



usually thought to be the most rigorousmethodological approach.

Of the 36 studies included in this review,7 of the 15 model program studies formedcomparison groups by random assign-ment.56 None of the 21 large-scale publicprogram studies used random assignment.

Sample SizeResearchers commonly use statistical tests ofdifferences between groups to estimate thelikelihood that findings are due to the inter-vention rather than to chance. These testscan help determine whether the sameeffects could be reproduced for the wholetarget population. A general principlebehind all of these tests is that it is muchharder to demonstrate a “statistically signifi-cant” difference between groups (usuallydefined as a result that would occur bychance no more than 5 times out of 100)when there are only a few participants thanwhen there are many.

Among the evaluations of model pro-grams included in this review, two experi-mental studies (Milwaukee and the EarlyTraining Project) began with extremelysmall sample sizes which provided thesestudies with very little statistical power to

detect even fairly large effects. The large-scale studies, just because more childrenare involved, are at an advantage in havinggreater statistical power to detect effects ofthe services offered.

AttritionAttrition (loss of study participants overtime) can be a serious problem for any study.First, because attrition means that fewer par-ticipants remain, it reduces a study’s statisti-cal power to detect effects. Second, itreduces confidence that (a) the final sampleis comparable to the initial sample and (b)the final program and comparison groupsare comparable to each other. If the finalsample differs substantially from the origi-nal, the results might not generalize to the

original target population. If treatment andcomparison groups lose comparability, thebenefits of random assignment are lost, andthe results of a comparison can be totallymisleading.

Among the model programs in thisreview, four had attrition rates so high thatinitial random assignment could have beeninvalidated.17,18,23,28 Among the large-scalepublic programs, attrition appears to havebeen a substantial problem for at least sixprograms.33,37,43–46

Measurement IssuesResearchers must make sure that the teststhey use to measure outcomes are adminis-tered accurately and fairly to all participantsin the study. In many studies of early child-hood programs, especially studies of large-scale programs, standardized tests routinelyadministered by schools often served as thesource of achievement test data for follow-up. Although this strategy provided data atlow cost, it had several unfortunate conse-quences, including (1) less uniformity of testadministration and (2) lost data becauseschools used different tests from year to year,and not all children—especially those whowere retained in grade or in special educa-tion programs—were tested.

Studies that relied on school-adminis-tered tests would, at best, have less reliabletest scores and smaller sample sizes. Atworst, they would systematically have lostmore poorly performing students fromeach year as the cumulative percentage ofchildren retained in grade and placed inspecial education increased. Even whenstudies administered their own achieve-ment tests, other research design flawssometimes produced a similar distortionof achievement comparisons over time.43,51

The effect on these studies would be togradually “erase” any differences betweenprogram and comparison groups withachievement test data as grade levelincreased.

Among the studies reviewed, measure-ment problems affect all of the large-scaleprogram studies31–51 and at least fourmodel program studies.18,21,28,30 It isimportant to note that measurementproblems do not affect the other findingsreported in those studies, such as effects


None of the 36 studies reviewed is perfect;however, it is important to look at theoverall picture.


on grade retention and special educationplacement.

In sum, none of the 36 studies reviewedin this article is perfect. However, while it isimportant to acknowledge their weaknessesand, perhaps, weigh studies of varying quali-ty differently, it is also important to look atthe overall picture. Each of these studies is avaluable addition to the literature andshould be reviewed with an eye toward whatit adds to our overall understanding of theeffects of ECCE programs.

Long-Term Study FindingsThis section discusses the long-term effectson cognitive development, school success,and socialization reported by each study.The results of each study’s longest follow-upare reported in Tables 1 and 2 (for modelprogram and large-scale program studies,respectively) for outcome measures that areeasily compared across studies—IQ, achieve-ment, grade retention, special educationplacement, and high school graduation.After the discussions of each type of out-come, two key issues are discussed across alloutcomes: Did effects vary by characteristicsof the children served (for example, by age,ethnicity, or gender)? Did effects vary withprogram characteristics?

Effects on IQAlthough the general public tends to thinkof intelligence as native cognitive ability

(how smart you are) and of IQ tests as mea-suring intelligence, there is considerabledisagreement among experts about whatconstitutes intelligence, the extent to whichits development is influenced by the envi-ronment, and what IQ tests measure.57

There are doubts about how completelyand accurately IQ tests measure generalintellectual ability. Nevertheless, researchersoften used IQ scores to gauge the success ofprograms, and those results are reported inthe following section.

Model Programs All of the model program studies reportedIQ gains at some point during or after chil-dren’s program participation. In mostinstances, effects were sustained untilschool entry at age five, at which time 10studies reported effects between 4 and 11IQ points,16,18,21,28–30,37–40 the Milwaukeestudy reported a gain of 25 points, and theSyracuse study reported no effect. Threestudies did not measure IQ at schoolentry.17,22,27

IQ effects persisted the longest (intoadolescence) in the two experimental stud-ies that enrolled infants in full-day educa-tional child care programs (Milwaukee andAbecedarian). Although two other studiesthat enrolled infants did not find persistentIQ effects, both were quasi-experimen-tal20,21 and one ceased serving childrenbefore age three.21

© J

eff

rey

Sylv

est

er/

FPG

Inte

rna

tion

al


Program Nameb Ages of Participation Design Initial (Years of Operation) Sample Sizec

Child-Parent Center31 Entry: 3 or 4 years Compared former CPC children with non- E = 684CPC children from same feeder schools. C = 304

(1965–1977) Exit: 9 years

Child-Parent Center II32 Entry: 3 or 4 years Compared former CPC children with Unknownseveral other groups.

(1983–1985) Exit: 9 years

Cincinnati Title I Preschool33 Entry: 4 or 5 years Compared children who attended full-day E = 688kindergarten and mostly had preschool C = 524

(1969-1970; 1970-1971) Exit: 6 years with children who attended half-daykindergarten and mostly had nopreschool.

Maryland Extended Entry: 4 years Compared attenders to nonattenders, UnknownElementary Pre-K34 including only children continuously

Exit: 5 years enrolled in school district (kindergarten(1977-1980) to grade 5).

New York State Experimental Entry: 3 or 4 years Compared attenders with children in same 1,800j

Prekindergarten35 district on waiting list and with children inExit: 5 years other districts with no prekindergarten

(1975-1976) program.

Detroit Head Start and Entry: 4 years Compared children who had attended UnknownTitle I Preschool36 Head Start or Title I preschool with children

Exit: 5 years who were eligible but did not attend.(1972-1973)

DC Public Schools and Entry: 4 years Compared children who attended public E = 372Head Start37 school preschool or Head Start with C = 89

Exit: 5 years children in same kindergartens who (1986-1987) had not.

Florida Learn to Learn and Entry: 4 years Compared children who attended LTL E = 45Head Start38 preschool or Head Start at age 4 (E) C = 45

Exit: 5 years with children who started school in (1986-1987) kindergarten (C).

Philadelphia School District Entry: 4 years Compared children in enriched K-3 program E = 1,082Get Set and Head Start39 (follow-through) who had and had not C = 1,615

Exit: 5 years attended preschool.(1969-1970; 1970-1971)

Table 2

Large-Scale Public Early Childhood Programsa

Notes a Programs are grouped such that public school program studies are listed first, followed by program studies involving both public school programs and Head Start, and

then all Head Start studies.b See the related endnotes at the end of this article for complete citations of the reports and/or studies in which program methods and outcomes are described.c Throughout Table 2, E refers to the experimental or intervention group, and C refers to the control or comparison group.d Outcomes listed as E > C or E < C were statistically significant at the p < .05 level, at least (that is, likely to have occurred by chance no more than 5 times in 100).


e Results may be biased because children were not randomly assigned to experimental and control or comparison groups.f No pretest was given to assess/control for initial differences between groups.g Results may be biased because school-administered tests were used to measure achievement.h Results may be biased because of high attrition rates.i Design flaws bias the estimated effect of the program on children’s achievement toward zero.j The numbers of children in experimental and comparison groups were not reported separately.


Follow-up Time of Last School Outcomesd Methodological ConcernsSample Size Follow-up

E = 513 Post high Achievement tests: Not randomized.e No pretest.fC = 244 school E > C at grade 2 School-administered tests.g

E = C at grade 8

High school graduation: E > CE = 62%, C = 49%

E = 757 Grade 7 Achievement tests: E > C for grades K to 7 Not randomized. No pretest.C = 130 School-administered tests.

Special education: E < CE = 12%, C = 22%

Grade retention: E < CE= 24%, C = 34%

E = 410 Grade 8 Achievement tests: E > C for grades 1, 5, and 8 Not randomized. No pretest.C = 141 School-administered tests.



E = 356 Grade 8 Achievement tests: E > C for grades 3, 5, and 8 Not randomized. No pretest. HighC = 306 attrition.h

Special education: E < C, grade 8 School-administered tests.E = 15%, C = 22%

Grade retention: E < C, grade 8E = 31%, C = 45%

E = 1,348 Grade 3 Achievement tests: Not randomized. High attrition.C = 258 E > C in kindergarten

E = C in grade 1

Special education: E = CE = 2 %, C = 5%

Grade retention: E < CE = 16%, C = 21%

Unknown Grade 4 Achievement tests: E > C in grade 4 Not randomized. No pretest. School-administered tests. Bias toward no effect.i

E varies Grades 4 and 5 Achievement tests: E = C in grades 3 to 5 Not randomized. Bias toward no effect.C varies High attrition.



E = 44 Grade 6 Achievement tests: E = C Not randomized. No pretest.C = 39

Special education: E = C

Grade retention: E = C

E = 688 Grades 4 to 8, Achievement tests: E = C Not randomized. No pretest.C = 524 varies by Bias toward no effect. High attrition.

cohort Grade retention: E > C School-administered tests.


Program Nameb Ages of Participation Design Initial(Years of Operation) Sample Sizec

Seattle DISTAR and Head Start40 Entry: 4 years Compared children who had attended E = 92Head Start and DISTAR with matched C = unknown

(1970–1971) Exit: 5 years children from same school and grades.

Cincinnati Head Start41 Entry: 4 years Compared third graders who had attended UnknownHead Start with those who had not.

(1968–1969) Exit: 5 years

Detroit Head Start42 Entry: 4 years Compared children who had attended UnknownHead Start with children in Title I

(1969-1970) Exit: 5 years elementary programs.

ETS Longitudinal Study of Entry: 4 or 5 years Compared children who went to Head Start 1,875Head Start43 with children who went to other preschools

Exit: 5 or 6 years or no preschool.(1969-1970; 1970-1971)

Hartford Head Start44 Entry: 4 years Compared children who had attended 293Head Start with those who had not.

(1965-1966) Exit: 5 years

Kanawha County, West Virginia Entry: 4 years Compared children who had attended UnknownHead Start45 Head Start with low-income children

Exit 5 years who had not.(1973-1974)

Montgomery County, Maryland Entry: 4 years Compared children who had attended eight E = 1,915Head Start46 or nine months with those who had C = 619

Exit 5 years attended one month or less.(1970-1971; 1974-1975;1978-1979)

NBER-NLSCM Head Start47 Entry: 3 to 5 years Compared children who had attended UnknownHead Start with those who had not.

(1979-1989) Exit: 5 to 6 years

New Haven Head Start48 Entry: 4 years Compared children who attended Head E = 61Start with those who had not. C = 48

(1968-1969) Exit: 5 years

Pennsylvania Head Start49 Entry: 3 to 5 years Compared children who attended Head E = 98Start with children who had applied C = unknown

(1986-1987) Exit: 5 to 6 years but had not been admitted.

Rome, Georgia, Head Start50 Entry: 5 years Compared children who attended Head E = 130Start with all children in first grade C = 88

(1966) Exit: 6 years in disadvantaged schools in 1966.

Westinghouse National Entry: 4 or 5 years Compared children who attended Head UnknownEvaluation of Head Start51 Start with those who did not (matched

Exit: 5 or 6 years within grade).(1965-1966)

Table 2 (continued)

Large-Scale Public Early Childhood Programsa

Notes a Programs are grouped such that public school program studies are listed first, followed by program studies involving both public school programs and Head Start, and

then all Head Start studies.b See the related endnotes at the end of this article for complete citations of the reports and/or studies in which program methods and outcomes are described.c Throughout Table 2, E refers to the experimental or intervention group, and C refers to the control or comparison group.d Outcomes listed as E > C or E < C were statistically significant at the p < .05 level, at least (that is, likely to have occurred by chance no more than 5 times in 100).



Follow-up Time of Last School Outcomesd Methodological ConcernsSample Size Follow-up

E = 44 Grades 6 and 8 Achievement tests: E = C, but positive trend, in Not randomized. No pretest.C = 20 grades 6 and 8 High attrition. School-administered tests.

Unknown Grade 3 Achievement tests: E = C in grade 3 Not randomized. No pretest.Bias toward no effect.

Unknown Grade 4 Achievement tests: E > C in grade 4 Not randomized. No pretest.School-administered tests. Bias toward

no effect.

852 Grade 3 Achievement tests: Not randomized. High attrition.E > C in grade 1 Bias toward no effect.E = C in grades 2 and 3

E =148 Grade 6 Achievement tests: E = C in grade 6 Not randomized. No pretest. HighC = 50 attrition. School-administered tests.

Special education: E = CE = 5%, C = 10%


Unknown Grade 3 Achievement tests: E = C in grade 3 Not randomized. No pretest.School-administered tests. Bias toward

no effect.

E = 186 Grade 11 Achievement tests: Not randomized. Possibly no pretests. C = 112 E = C, but negative trend in most grades High attrition. School achievement tests.

E > C in grade 11

E = 747 Grade varies Achievement tests: E = C Not randomized. No pretest.C = 1,810

Grade retention: E > C, whites only

E = 35 Grade 3 Achievement tests: Not randomized. No pretest. High attrition.C = 26 E > C in grade 1 Bias toward no effect.

E = C in grade 3


E = 54 Grade 3 Achievement tests: E = C, but positive trend, Not randomized. No pretest.C = 18 in grades 2 and 3

E = 94 Post high Achievement tests: Not randomized. No pretest. School-C = 60 school E > C in grade 5 administered tests.

E = C in grades 6 and above

Special education: E < CE = 11%, C = 25%

Grade retention: E = CE = 51%, C = 63%

E = 1,988 Grades 1 to 3 Achievement tests: Not randomized. No pretest. BiasC = 1,992 E > C in grade 1 toward no effect.

E = C in grades 2 and 3

Large-Scale Programs Only one of the large-scale program studiesprovided IQ data using tests comparable tothose used in the model program studies.44

A small number of studies35,43,48 providedresults on a test of language ability, andanother51 administered a test of cognitiveability. In any event, only one large-scaleprogram study found effects on these vari-ous tests of linguistic or cognitive abilityafter children exited ECCE and enteredschool.47

Effects on AchievementAchievement tests measure ability andknowledge in the subjects of reading andmathematics. Compared with IQ tests, theyare more focused on academic accomplish-ment and the acquisition of what is taught inschool.

Model Programs Estimated effects on standardized achieve-ment tests varied among the model pro-gram studies. Five of 11 studies withachievement test data found statisticallysignificant positive effects beyond thirdgrade.16,18,25,26,30 Evidence of effects wasstrongest among the experimental stud-

ies that had used random assignment toform comparison groups. Achievementeffects were found through second grade(Milwaukee), fourth grade (Florida),and into junior high school (Abecedarianand Perry). In contrast, only one of thequasi-experimental studies of modelprograms found long-term effects onachievement.30

Large-Scale Programs The achievement test results of the large-scale program studies were quite variable.Four found no effects at any time.37,41,45,47

Five found initial effects that faded andceased to be statistically significant by theend of third grade.35,39,43,48,51 The othersfound statistically significant effects in thirdgrade or later, though the patterns of effectsover time are variable.

The variation in findings with respect toachievement could be the result of (1) thebasic design weakness of quasi-experimentalstudies, (2) exceptionally high attrition ratesfor achievement test data that both reducedsample size and biased comparisons towardfinding no effect,33,39,44 or (3) some of themeasurement problems described in thesection above.43,48,51

Effects on School Progress andPlacementSchool outcomes were also measured byrates of grade retention, special education,and high school graduation.

Grade Retention and SpecialEducation in Model Programs Across all studies, the findings were relative-ly uniform and constitute overwhelming evi-dence that ECCE can produce sizableimprovements in school success. All but oneof the model program studies reportedgrade retention and special education rates,and in all of these the rates are lower for theprogram group. The one model programstudy that did not report rates (Syracuse)simply reported that there was no statistical-ly significant difference. The estimatedeffects for the model programs are notalways statistically significant given the smallsample sizes; but in most instances, they arelarge enough to be of practical importance.Despite small sample sizes, statistical signifi-cance on one or the other was found in fivemodel program studies,16,18,21,23,30 and anoth-er, the Perry Preschool study, found signifi-cant effects on the rate of placement formild mental retardation and for number ofyears of special education.

Grade Retention and SpecialEducation in Large-Scale ProgramsStatistically significant effects on graderetention or special education were foundin 8 of the 10 large-scale program studiesthat collected relevant data. The failure tofind significant effects in two studiesappears to be the result of relatively lowrates of retention and special educationplacement in the community as a whole(Cincinnati Title I) and/or initial differ-ences between the groups (Washington,DC). Both of these circumstances wouldmake it harder to demonstrate a differ-ence between the experimental and com-parison groups.37


The findings constitute overwhelmingevidence that ECCE can produce sizableimprovements in school success.

High School Graduation Three model program studies and twolarge-scale program studies had sufficientlylong follow-ups to assess effects on highschool graduation rates.22,23,26,31,50 All fiveestimated that ECCE had a large effect onthe graduation rate, though only the threestudies with larger sample sizes found theeffect to be statistically significant. However,added support is provided by the other stud-ies that find effects on achievement, graderetention, or special education place-ment—all of which are predictive of highschool graduation.26,58,59

Effects on SocializationAlthough the primary focus of long-termstudies of ECCE has been on cognitive devel-opment and school success, socialization(the learning and adoption of sociallyaccepted values and behavior) has receivedsome attention, particularly in the modelprogram studies.

Increased aggression at school entry hadbeen found for three studies in which chil-dren began child care as infants,20,21,60 butthere is no evidence that this aggression per-sisted. Indeed, two of those studies foundthat program children had better classroombehavior later,20,21 and two other studiesreported that children were rated by ele-mentary school teachers as better adjustedsocially.17, 49 Two studies found no significanteffects on primary grade teacher ratings ofclassroom behavior.18,26

Long-term positive effects on social-ization were evident not only in teacherratings, but also in parent ratings in onestudy20 and in data on delinquency andcrime in the only two studies that soughtthese out.20,26 In addition, several modelECCE programs were found to increasepride in school achievement.61 The PerryPreschool study provides the longest andmost intensive follow-up study of effectson socialization.26 It found that ECCEwas associated with increased commit-ment to school, better relationships withfriends and neighbors, greater adult eco-nomic success, and, for girls, increasedmarriage and fewer out-of-wedlock births.(For further review of this set of out-comes, especially effects on crime anddelinquency, see the article by Yoshikawain this journal issue.)

Child Characteristics and EffectsOne of the most important policy questionsis whether ECCE programs appear to havedifferent effects on different groups of chil-dren. This sort of question usually can beaddressed by examining the results within astudy (if the researcher actually comparedgroups of children) or by examining resultsacross studies that served different groups ofchildren.

In general, there do not appear to belarge variations in effects for children fromlow-income families, though this questionhas not received much attention. Outsidethis range, family income becomes relevant;the same effects would not be expected forchildren from higher-income families whosedevelopment is not impacted by poverty.

One possible exception to this generalrule is gender. Four experimental studiesof model programs (Abecedarian, Houston,Perry, Early Training Project) found largereffects on achievement test scores forlow-income girls than boys, though the dif-ferences were not necessarily statistically sig-

nificant. Two of these studies found thatgraduation rates were higher for girls thanfor boys (Perry, Early Training Project).Results of the quasi-experimental studies ofmodel programs are less consistent with thispicture, and none of the large-scale studieswhich explicitly tested for gender differ-ences found any.

While it is true that these reported gen-der differences could be statistical flukes,the findings occur with enough consistencyin some of the best quality studies that it isreasonable to conclude that the possibilityof gender differences warrants furtherattention.

Program Characteristics and EffectsAnother important policy question iswhether particular types of programs or con-stellations of services appear to be more


Four experimental studies of model programsfound larger effects on achievement testscores for low-income girls than boys.

effective than others. Again, one can try toanswer this question by examining resultsboth within and across studies.

Comparison of Model Programs with Community Child Care Of the 36 studies in the review, only theAbecedarian study investigated the poten-tial effects of the comparison group’s par-ticipation in other forms of ECCE.62

Comparison group children who hadattended community ECCE programs thatmet federal guidelines for quality63 werefound to have higher IQ scores than com-parison group children with little or noECCE experience. The estimated effect atschool entry was roughly half the size ofthe effect of the Abecedarian program.

Head Start and Other Early Childhood Programs Another issue addressed by some within-study comparisons is the relative effective-ness of Head Start compared with otherECCE programs, usually public schoolpreschool programs. Public school pro-grams might be thought to be more effec-tive because they pay much higher salariesthan Head Start and can attract better-

qualified staff, but Head Start offers abroader range of services. The studiesreviewed here all reported smaller effectsfor Head Start.32,36,38,39 Of course, HeadStart children tend to be more disadvan-taged; therefore, with one exception, thesecomparisons could be affected by preexist-ing differences between the two popula-tions. The exception, a study that randomlyassigned children to either a model pro-gram which continued through first gradeor to Head Start, found that the modelprogram produced a larger effect on long-term achievement.38

Age of Entry into the ECCE Program A number of the older model program stud-ies were designed to investigate the effects ofage of entry and duration of services onchild development.18,23,25,30 None found sig-

nificant effects from earlier entry (for exam-ple, from entry at age two rather than threeyears).64 Unfortunately, these studies tendedto have such small sample sizes and suchhigh attrition that only very large effectscould have been detected, and the resultscannot be considered very conclusive. TheChild Parent Center II study32 was the onlyother study to estimate the long-term effectsof variations in age of entry, and it found noadvantage for children who entered at agethree compared with children who enteredat age four.

Effectiveness of Add-on Schoolage Services Intuitively, it makes sense that, if preschool isgood for children, then continuing to pro-vide them with enhanced services duringtheir subsequent school years ought to beeven better. Two studies provide informationabout the effects of such extended elemen-tary programs.

In the Abecedarian study, half of the pro-gram and control groups were randomlyassigned to a special schoolage program atage five, enabling the researchers to com-pare the effects of ECCE alone, ECCE plusan enriched schoolage program, theschoolage program alone, and no interven-tion. The schoolage program was providedfor the first three years of elementary schooland consisted of biweekly home visits inwhich teachers provided individualizedsupplemental activities in partnership withparents and social supports for families. Byadolescence the results were clear.Substantial effects on IQ, achievement, andschool progress were produced by ECCEalone. The schoolage program was largelyineffective and, as an add-on to ECCE, hadno effects on IQ and only mixed effects onschool success and achievement.

In contrast, the CPC II study found thatenriched elementary school services addedsubstantially to the effects of ECCE, with thesize of the effect increasing directly withthe number of years of enhanced elemen-tary services. One possible explanation forthe difference in results between theAbecedarian and the CPC II studies is thatthe two research teams studied different pro-grams. While the Abecedarian program wasa modest supplement to the children’sschool experience, the CPC II program

To have any effect at all, schoolage servicesmust be more than add-ons to a preschoolprogram.


changed the elementary school in markedways: classes were smaller, additional class-room and support staff were added, andparent involvement was emphasized. Analternative explanation is that the parentswho sought out and continued participationin CPC II somehow differed from those whodid not (that is, they differed from the con-trol group), and it is those differences ratherthan the intervention that led to theobserved benefits.

In any event, it seems clear that, to haveany effect at all, schoolage services must bemore than add-ons to a preschool program.These services must actually change thelearning environment in some significantways before they can be expected to pro-duce benefits in addition to those producedby ECCE.

ConclusionsMany studies have investigated the long-termeffects of preschool programs on disadvan-taged children, but they are far from equal intheir capacity to inform public policy. Oneconservative view of the literature would bethat only two studies provide sufficiently validestimates of the effects of ECCE—theAbecedarian and Perry Preschool studies.The other experiments are impaired by vari-ous methodological problems such as smallsample size, attrition, and selection bias.

Of course, this is much too narrow a view.The Abecedarian and Perry programs eachserved African-American children in smallcities, so focusing on them limits the conclu-sions that we can draw to those populations.The other studies can add a great deal to ourunderstanding of ECCE programs in termsof types of programs, populations served,and social and historical context. Thus, infor-mation from these other studies may andshould be used to supplement the informa-tion from the stronger studies, especially iftheir short-term findings, at least, are consis-tent with those of existing experiments.

It is from this perspective that the con-cluding section returns to the policy ques-tions posed at the beginning of this article.

What Are the Effects of ECCEand How Long Do They Persist?The weight of the evidence establishes thatECCE can produce large effects on IQ dur-

ing the early childhood years and sizablepersistent effects on achievement, graderetention, special education, high schoolgraduation, and socialization. In particular,the evidence for effects on grade retentionand special education is overwhelming.Evidence is weaker for persistent achieve-ment effects, but this weakness is probablythe result of flaws in study design and follow-up procedures. Evidence for effects on highschool graduation and delinquency is strongbut based on a smaller number of studies.

These effects are large enough andpersistent enough to make a meaningfuldifference in the lives of children fromlow-income families: for many children,preschool programs can mean the differ-ence between failing and passing, regularor special education, staying out of troubleor becoming involved in crime and delin-quency, dropping out or graduating fromhigh school.

Do Effects Vary with thePopulation Served or Type of Program?Benefits from ECCE programs appear to beproduced via a number of different types ofprograms and across a number of differentgroups of children. Indeed, the best predic-tor of the size of program effects may be thesize of the gap between the program and

home as learning environments, rather thanwhether a child is a member of a particulargroup. Thus, effects might be expected to belargest for the most disadvantaged, thoughthere is no evidence that meaningful effectscease if a child’s family moves above thepoverty line. Indeed, there is even some sug-gestion at the other end of the income spec-trum that children from very well-off familiesmay suffer from ECCE inferior to that pro-vided by their homes.

The most interesting hint with respect tovariations in effects with child characteristicsis that long-term effects on educational


Preschool programs can mean the differencebetween failing and passing, regular orspecial education, or staying out of trouble.

achievement and attainment might begreater for girls than for boys. The reason isunclear, but because boys from low-incomefamilies fare so poorly in the educationalsystem (twice as many boys as girls are inspecial education), further research on thistopic is warranted. Possibly, teaching meth-ods could be altered in ECCE or in elemen-tary schools to better accommodate theneeds of these boys.

With respect to program characteristics,evidence about when programs shouldbegin and how long they should last ismixed. The notion that development ismore easily influenced earlier suggests thatearlier programs should have larger effects.This notion is confirmed by cross-study com-parisons but receives little support fromwithin-study comparisons. One possibleexplanation for the apparent contradictionis that within-study comparisons have tendedto look at very limited age differences suchas the effect of beginning at three ratherthan four years of age. The more importantdifference may be between beginning ininfancy versus beginning later. Also, researchon brain development indicates that thesame effects on brain development can pro-duce different effects on cognitive and socialdevelopment depending on the age atwhich they occur.65 Thus, ECCE beginningwith infants and continuing to kindergartenmay be required to produce persistenteffects on IQ and may produce larger effectson academic success and, perhaps, on social-ization as well.

Guidance with respect to the effects ofenriched elementary school programs is lim-ited and mixed. Despite the intuitive appealof the idea that fade-out in the benefits ofECCE might occur without prolonged inter-vention, the empirical support for this view isextremely weak. Fade-out is more apparentthan real for all measures except IQ: pro-longed effects on achievement, school suc-cess, and socialization occur withoutschoolage intervention. The only direct sup-port for the need for prolonged, schoolageintervention comes from a single quasi-experimental study.32 However, improve-ments in elementary education for childrenwho attend poor quality schools would beexpected to contribute independently tochild development including IQ, achieve-ment, and school success and might be a way

to improve long-term educational outcomesfor boys.66

To What Extent Can Findings BeGeneralized to Existing Publicand Private Programs?Research supports the view that large-scalepublic ECCE programs can produce long-term cognitive and academic benefits fordisadvantaged children. Comparison of esti-mated long-term effects between modelprograms and large-scale programs indi-cates that the latter tend to have smallereffects, perhaps because model programsprovided higher quality services than manyof the large-scale public programs. Ofcourse, the extent to which the large-scalepublic programs that were studied accurate-ly reflect today’s large-scale public programsgenerally is unknown. Nevertheless, there isa risk that today’s public programs will notproduce the desired benefits because theyare lower in quality (larger classes, fewerstaff members, less educated staff, poorersupervision) than the model programs.

Cross-study and within-study compar-isons suggest that Head Start has been lesseffective than better-funded public schoolprograms, although these comparisons suf-fer from methodological problems thatreduce confidence in the results. Neverthe-less, these findings are consistent with theview that quality matters, a view recentlyendorsed explicitly by a national panelexamining the future of Head Start.67

Costs, Benefits, and FinancingIt should be evident that ECCE can producesubstantial improvements in the cognitivedevelopment and educational success of dis-advantaged children. The best-knownsource of support for this view, the PerryPreschool study, does not stand alone, andits effects are not unusually large. Much larg-er effects might be produced by theAbecedarian program model. From this per-spective, the benefit-cost analysis of the PerryPreschool program provides a conservativeestimate of the potential returns to publicinvestment in ECCE.68 Results of this benefit-cost analysis are presented in Table 3.

As can be seen in Table 3, the presentvalue of benefits (the current value of afuture stream of costs of benefits)69 of the


Perry program greatly exceeded costs forboth program participants and the generalpublic. This is true even if all benefits fromreductions in delinquency and crime weretotally omitted. The national cost of failingto provide at least two years of quality ECCEis extremely high, on the order of $100,000for each child born into poverty, or $400 bil-lion for all poor children under fivetoday.68,70 An immediate and substantialincrease in public support for ECCE is war-ranted, therefore, on economic groundsalone. However, the appropriate public pol-icy response is more complex than provid-

ing two years of quality Head Start to everypoor child.

Thirty years ago, when fewer than 25% ofmothers of children under six were in thework force, a two-year half-day preschool pro-gram for poor children might have seemedlike an appropriate response. Today, it doesnot. In 1990, nearly half of the childrenunder age three were cared for by someoneother than a parent. (See the article byHernandez in this journal issue.) For poorchildren, welfare reform that requires moth-ers to work outside the home will greatly


Recipients of Costs and BenefitsCost or Benefit

Whole Preschool GeneralSociety Participants Public

Preschool Costa -$ 12,356 $ 0 -$ 12,356

Measured BenefitsChild Care 738 738 0K - 12 Education 6,872 0 6,872Adult Education 283 0 283Collegeb -868 0 -868Employmentc 14,498 10,269 4,229Crime 49,044 0 49,044Welfare 219 -2,193 2,412

Benefit Subtotal $ 70,876 $ 8,814 $ 61,972

Projected BenefitsEarnings 15,833 11,215 4,618Crime 21,337 0 21,337Welfare 46 -460 506

Total Benefits $108,002 $ 19,569 $ 88,433

Net Present Value $ 95,646 $ 19,569 $ 76,077

a Costs and cost increases appear as negative numbers.

b Some small portion of college costs is likely to have been borne by the participants, but this could not be esti-mated from the available information.

cThe benefits reported for employment include all costs paid by the employer to hire a participant. Allocationto participants and the general public assume that: (1) the marginal tax rate is 25%, (2) the value of fringe ben-efits received by the employee equals 10% of salary, and (3) the value of other fringe benefits paid by theemployer (for example, the employer’s share of Social Security payments) equals 10% of salary.

Table 3

Present Value of Costs and Benefits per Child in 1990 Dollars Discounted at a Real Rate of 3%

increase the numbers in the future.Moreover, whether or not children are poor,the quality of the ECCE services they receiveis important for their development. Poorquality ECCE could be detrimental to thedevelopment of any child at any age. It is nolonger just the benefits of quality ECCE fordisadvantaged children that are at stake.

Bringing ECCE services to all childrenwho could benefit from them will not becheap. Realistically, the cost of serving allpoor children under age five years in qualitypart-day or full-day (depending on need)ECCE programs could be as high as $25 bil-lion or $30 billion per year. If to this amountwere added sizable subsidies to nonpoorfamilies to encourage them to purchasequality ECCE, the total cost could approach5% of the federal budget (though the costcould be shared by state government, as

well). However, based on the evidence pre-sented above, these costs would be offsetover time by reductions in social problemsthat cost society far more each year.

A more comprehensive strategy is need-ed to increase the public and privateresources devoted to ECCE. Such a strategymight include a public information cam-paign to explain the importance of ECCEquality to parents, paid parental leave forparents of children under one year of age,and public funding for accredited ECCE ona sliding scale with full funding of qualitycare for children in poverty and partial fund-ing for many more children.70 Other alter-natives are available, but the important pointis that the nation needs to move ahead withpublic support for ECCE. Current policiesare penny wise and pound foolish, inexcus-ably costly in human and financial terms.


1. Consortium for Longitudinal Studies, ed. As the twig is bent . . . lasting effects of preschool pro-grams. Hillsdale, NJ: Erlbaum, 1983.

2. Haskins, R. Beyond metaphor: The efficacy of early childhood education. American Psychologist(1989) 44:274–82.

3. Locurto, C. Beyond IQ in preschool programs? Intelligence (1991) 15:295–312.

4. Seitz, V. Intervention programs for impoverished children: A comparison of educational andfamily support models. Annals of Child Development (1990) 7:73–104.

5. Lamb, M., and Sternberg, K. Do we really know how day care affects children? Journal ofApplied Developmental Psychology (1990) 11:351–79.

6. Zaslow, M. Variation in child care quality and its implications for children. Journal of SocialIssues (1991) 47,2:125–39.

7. Helburn, S., and Culkin, M.L. Cost, quality, and child outcomes in child care centers: Executive sum-mary. Denver: Economics Department, University of Colorado at Denver, 1995.

8. Phillips, D.A., McCartney, K., and Scarr, S. Child-care quality and children’s social develop-ment. Developmental Psychology (1987) 23:537–43.

9. Caughy, M.O., DiPietro, J., and Strobino, M. Day-care participation as a protective factor inthe cognitive development of low-income children. Child Development (1994) 65:457–71.

10. Ramey, C.T., Bryant, D.M., and Suarez, T.M. Preschool compensatory education and the mod-ifiability of intelligence: A critical review. In Current topics in human intelligence. D. Detterman,ed. Norwood, NJ: Ablex, 1985, pp. 247–96.

11. McKey, R., Condelli, L., Ganson, H., et al. The impact of Head Start on children, families, and com-munities. Final report of the Head Start Evaluation, Synthesis, and Utilization Project.Washington, DC: U.S. Department of Health and Human Services, 1985.

12. White, K., and Casto, G. An integrative review of early intervention efficacy studies with at-riskchildren: Implications for the handicapped. Analysis and Intervention in DevelopmentalDisabilities (1985) 5:7–31.

13. Olds, D., and Kitzman, H. Review of research on home visiting for pregnant women and par-ents of young children. The Future of Children (1993) 3,3:53–92.

14. Wasik, B.H., Ramey, C.T., Bryant, D.M., and Sparling, J.J. A longitudinal study of two earlyintervention strategies: Project CARE. Child Development (1990) 61:1682–96.

15. Infant Health and Development Program. Enhancing the outcomes of low-birth-weight pre-mature infants. Journal of the American Medical Association (1990) 263,22:3035–42; St. Pierre, R.and Lopez, M. The comprehensive child development program. Presentation to the National

Research Council, Board on Children and Families. Washington, DC, December 16, 1994; St.Pierre, R., Swartz, J., Murray, S., et al. National evaluation of the Even Start family literacy program:Report on effectiveness. Washington, DC: U.S. Department of Education, Office of Policy andPlanning, 1993.

16. Campbell, F.A., and Ramey, C.T. Mid-adolescent outcomes for high risk students: An examina-tion of the continuing effects of early intervention. Paper presented at the biennial meetingof the Society for Research in Child Development. New Orleans, March 26, 1993; Campbell,F.A., and Ramey, C.T. Effects of early intervention on intellectual and academic achievement:A follow-up study of children from low-income families. Child Development (1994) 65:684–98;F.A. Campbell: Unpublished analyses of Abecedarian data. F.A. Campbell generously providedto the author unpublished analyses of the Abecedarian data reporting means for academicand school success measures by preschool treatment group and gender.

17. Andrews, S.R., Blumenthal, J.B., Johnson, D.L., et al. The skills of mothering: A study of par-ent child development centers. Monographs of the Society for Research in Child Development. SerialNo. 198 (1982) 46,6; Johnson, D., and Walker, T. A follow-up evaluation of the HoustonParent Child Development Center: School performance. Journal of Early Intervention (1991)15,3:226–36.

18. Jester, R.E., and Guinagh, B.J. The Gordon Parent Education Infant and Toddler Program. InAs the twig is bent . . . lasting effects of preschool programs. Consortium for Longitudinal Studies,ed. Hillsdale, NJ: Erlbaum, 1983, pp. 103–32.

19. Garber, H.L. The Milwaukee Project: Prevention of mental retardation in children at risk. Washington,DC: American Association on Mental Retardation, 1988.

20. Lally, J.R., Mangione, P., and Honig, A. The Syracuse University Family DevelopmentProgram: Long-range impact of an early intervention with low-income children and their fam-ilies. In Parent education as early childhood intervention: Emerging directions theory research and prac-tice. D. Powell, ed. Norwood, NJ: Ablex, 1988.

21. Seitz, V., Rosenbaum, L.K., and Apfel, N.H. Effects of family support intervention: A ten-yearfollow-up. Child Development (1985) 56:376–91; Seitz, V., and Apfel, N.H. Parent-focused inter-vention: Diffusion effects on siblings. Child Development (1994) 65:677–83.

22. Miller, L.B., and Bizzell, R.P. The Louisville Experiment: A comparison of four programs. InAs the twig is bent . . . lasting effects of preschool programs. Consortium for Longitudinal Studies,ed. Hillsdale, NJ: Erlbaum, 1983, pp. 171–200; Miller, L.B., and Bizzell, R.P. Long-term effectsof four preschool programs: Sixth, seventh, and eighth grades. Child Development (1983)54:725–41; Miller, L.B., and Bizzell, R.P. Long-term effects of four preschool programs: Ninthand tenth grade results. Child Development (1984) 55,6:1570–87.

23. Gray, S.W., Ramsey, B., and Klaus, R. From 3 to 20: The Early Training Project. Baltimore, MD:University Park Press, 1982; Gray, S., Ramsey, B., and Klaus, R. The Early Training Project,1962–1980. In As the twig is bent . . . lasting effects of preschool programs. Consortium forLongitudinal Studies, ed. Hillsdale, NJ: Erlbaum, 1983, pp. 33–70.

24. Karnes, M.B., Shwedel, A.M., and Williams, M.B. A comparison of five approaches for educat-ing young children from low-income homes. In As the twig is bent . . . lasting effects of preschoolprograms. Consortium for Longitudinal Studies, ed. Hillsdale, NJ: Erlbaum, 1983, pp. 133–70.

25. Palmer, F. The Harlem Study: Effects by type of training, age of training, and social class. InAs the twig is bent . . . lasting effects of preschool programs. Consortium for Longitudinal Studies,ed. Hillsdale, NJ: Erlbaum, 1983, pp. 201–36.

26. Weikart, D.P., Bond, J.T., and McNeil, J.T. The Ypsilanti Perry Preschool Project: Preschool years andlongitudinal results through fourth grade. Ypsilanti, MI: High/Scope Press, 1978; Schweinhart,L.J., Barnes, H.V., Weikart, D.P., et al. Significant benefits: The High/Scope Perry Preschoolstudy through age 27. Monographs of the High/Scope Educational Research Foundation.No. 10. Ypsilanti, MI: High/Scope Educational Research Foundation, 1993; Barnett, W.S.,Young, J., and Schweinhart, L.J. How preschool education contributes to cognitive develop-ment and school success: An empirical model. Paper presented at the Rutgers InvitationalSymposium on Education. New Brunswick, NJ, October 28, 1994.

27. Herzog, E., Newcomb, C.H., and Cisin, I.H. Double deprivation: The less they have, the lessthey learn. In A Report on longitudinal evaluations of preschool programs. Vol. I. S. Ryan, ed.Washington, DC: U.S. Department of Health, Education, and Welfare, 1974.

28. Deutsch, M., Deutsch, C.P., Jordan, T.J., and Grallo, R. The IDS Program: An experiment inearly and sustained enrichment. In As the twig is bent . . . lasting effects of preschool programs.Consortium for Longitudinal Studies, ed. Hillsdale, NJ: Erlbaum, 1983, pp. 377–410.


29. Beller, K. The Philadelphia Study: The impact of preschool on intellectual and socio-emotion-al development. In As the twig is bent . . . lasting effects of preschool programs. Consortium forLongitudinal Studies, ed. Hillsdale, NJ: Erlbaum, 1983, pp. 133–70.

30. Levenstein, P., O’Hara, J., and Madden, J. The Mother-Child Home Program of the VerbalInteraction Project. In As the twig is bent . . . lasting effects of preschool programs. Consortium forLongitudinal Studies, ed. Hillsdale, NJ: Erlbaum, 1983, pp. 237–63.

31. Fuerst, J.S., and Fuerst, D. Chicago experience with an early childhood program: The specialcase of the Child Parent Center Program. Urban Education (1993) 28:69–96.

32. Reynolds, A.J. Longer-term effects of the Child Parent Center and Expansion Program. Paper present-ed at the annual meeting of the Chicago Association for the Education of Young Children.February 4, 1994; Reynolds, A.J. Effects of a preschool plus follow-on intervention for chil-dren at risk. Developmental Psychology (1994) 30:787–804; Reynolds, A.J. One year of preschoolintervention or two: Does it matter? Early Childhood Research Quarterly (1993) 10:1–33.

33. Nieman, R.H., and Gaithright, J.F. The long-term effects of ESEA Title I preschool and all-day kinder-garten: An eight-year follow-up. Cincinnati, OH: Cincinnati Public Schools, 1981.

34. Eckroade, G.A., Salehi, S., and Carter, J.L. An analysis of the midterm effects of the extended elemen-tary education prekindergarten program. Baltimore: Maryland State Department of Education,1988; Eckroade, G., Salehi, S., and Wode, J. An analysis of the long-term effect of the extend-ed elementary education prekindergarten program. Paper presented at the annual meetingof the American Educational Research Association. Chicago, 1991.

35. State Education Department, University of the State of New York. Evaluation of the New YorkState experimental prekindergarten program: Final report. ERIC Document Reproduction ServiceNo. ED 219 123. Albany, NY: New York State Education Department, 1982.

36. Clark, C.M. Effects of the project Head Start and Title I preschool programs on vocabularyand reading achievement measured at the kindergarten and fourth grade levels. Unpublisheddoctoral dissertation cited by McKey and colleagues 1985. Wayne State University, 1979.

37. Marcon, R.A. Early learning and early identification: Final report of the three year longitudinal study.Washington, DC: District of Columbia Public Schools, 1990; Marcon, R.A. Early learning andearly identification follow-up study: Transition from the early to the later childhood grades 1990–93.Washington, DC: District of Columbia Public Schools, Center for Systemic Change, 1993.

38. Sprigle, J.E., and Schaefer, L. Longitudinal evaluation of the effects of two compensatorypreschool programs on fourth- through sixth-grade students. Developmental Psychology (1985)21,4:702–8.

39. Copple, C.E., Cline, M.G., and Smith, A.N. Path to the future: Long-term effects of Head Start inthe Philadelphia School District. Washington, DC: U.S. Department of Health and HumanServices, 1987.

40. Evans, E. Longitudinal follow-up assessment of differential preschool experience for lowincome minority group children. Journal of Educational Research (l985) 78,4:197–202.

41. Pinkleton, N.B. A comparison of referred Head Start, non-referred Head Start and non-HeadStart groups of primary school children on achievement, language processing, and classroombehavior. Unpublished doctoral dissertation cited by McKey and colleagues 1985. Universityof Cincinnati, 1976.

42. O’Piela, J.M. Evaluation of the Detroit Public Schools Head Start program, 1975–1976, 1976.

43. Shipman, V.C. Disadvantaged children and their first school experiences: ETS-Head StartLongitudinal Study. Preliminary description of the initial sample prior to school enrollment. ETSTechnical Report Series, PR-70–20. Princeton, NJ: Educational Testing Service, 1970;Shipman, V.C. Stability and change in family status, situational, and process variables and their rela-tionship to children’s cognitive performance. Princeton, NJ: ETS, 1976; Lee, V.E., Brooks-Gunn, J.,Schnur, E., and Liaw, F.R. Are Head Start effects sustained? A longitudinal follow-up compari-son of disadvantaged children attending Head Start, no preschool, and other preschool pro-grams. Child Development (1990) 61:495–507.

44. Goodstein, H.A. The prediction of elementary school failure among high-risk children.Unpublished paper cited by McKey and colleagues 1985. Connecticut University, 1975.

45. Kanawha County Board of Education. Kanawha County Head Start evaluation study.Unpublished report. 1978.

46. Hebbeler, K. An old and a new question on the effects of early education for children fromlow income families. Educational Evaluation and Policy Analysis (1985) 7,3:207–16.


47. Currie, J., and Thomas, D. Does Head Start make a difference? Working Paper No. 4406. Boston:National Bureau of Economic Research, 1993.

48. Abelson, W.D. Head Start graduates in school: Studies in New Haven, Connecticut. In A reporton longitudinal evaluations of preschool programs. Vol. I. S. Ryan, ed. Washington, DC: U.S.Department of Health, Education, and Welfare, 1974, pp. 1–14; Abelson, W.D., Zigler, E., andDeBlasi, C.L. Effects of a four-year follow through program on economically disadvantagedchildren. Journal of Educational Psychology (1974) 66,5:756–71.

49. Reedy, Y.B. A comparison of long range effects of participation in Project Head Start and theimpact of three differing delivery models. Unpublished paper for the Graduate Program inSchool Psychology, Pennsylvania State University.

50. McDonald, M.S., and Monroe, E. A follow-up study of the 1966 Head Start program, Rome CitySchools. Unpublished paper, 1981.

51. Westinghouse Learning Corporation and Ohio University. The impact of Head Start: An evalua-tion of the effects of Head Start on children’s cognitive and affective development. Vols. 1 and 2. Reportto the Office of Economic Opportunity. Athens, OH: Westinghouse Learning Corporationand Ohio University, 1969.

52. Zigler, E., and Styfco, S.J. Using policy research and theory to justify and inform Head Startexpansion. Social Policy Report (1993) 8,2.

53. Slaughter, D.T., Washington, V., Oyemade, U.J., and Lindsey, R.W. Head Start: A backwardand forward look. Social Policy Report (1988) 3,2:1–19.

54. Willer, B., Hofferth, S.L., Kisker, E.E., et al. The demand and supply of child care in 1990.Washington, DC: National Association of the Education of Young Children, 1991.

55. Barnett, W.S., Frede, E.C., Mobasher, H., and Mohr, P. The efficacy of public preschool pro-grams and the relationship of program quality to efficacy. Educational Evaluation and PolicyAnalysis (1987) 10,1:37–49.

56. This includes the High/Scope Perry Preschool Project in which two departures from per-fect adherence to random assignment were made to keep siblings in the same group andto shift two children of working mothers from the program to the control group. Neitherchange had a significant effect on the results (see note No. 26, Schweinhart, Barnes,Weikart, et al., 1993).

57. Sternberg, R., and Detterman, D., eds. What is intelligence? Norwood, NJ: Ablex, 1986.

58. Ekstron, R., Goertz, M., and Rock, D.A. Education and American youth. Philadelphia: FalmerPress, 1988.

59. Roderick, M. Grade retention and school dropout: Investigating the association. AmericanEducational Research Journal (1994) 31,4:729–59.

60. Haskins, R. Public school aggression among children with varying day care experience. ChildDevelopment (1985) 56:689–703.

61. Royce, J.M., Darlington, R.B., and Murray, H.W. Pooled analyses: Findings across studies. InAs the twig is bent . . . lasting effects of preschool programs. Consortium for Longitudinal Studies,ed. Hillsdale, NJ: Erlbaum, 1983.

62. Burchinal, M., Lee, M., and Ramey, C. Type of day-care and intellectual development in disad-vantaged children. Child Development (1989) 60:128–37.

63. The Federal Interagency Day Care Requirements (rescinded in 1980) represented expertopinion about program quality and covered group size, staff-child ratio, and professionalqualifications of staff.

64. Lazar, I., Darlington, R., Murray, H., Royce, J., and Snipper, A. Lasting effects of early educa-tion: A report from the Consortium for Longitudinal Studies. Monographs of the Society forResearch in Child Development. Series No. 195 (1982) 47,2–3.

65. Kolb, B. Brain development, plasticity, and behavior. American Psychologist (1989)44,9:1203–12.

66. Husen, T., and Tuijnman, A. The contribution of formal schooling to the increase in intellec-tual capital. Educational Researcher (1991) 20,7:17–25.

67. Advisory Committee on Head Start Quality and Expansion. Creating a 21st century Head Start.Final Report of the committee. Washington, DC: U.S. Department of Health and HumanServices, 1993. See also the article by Frede in this journal issue.

68. Barnett, W.S. Benefit-cost analysis of preschool education: Findings from a 25–year follow-up.American Journal of Orthopsychiatry (1993) 63,4:500–508.


69. Present value is calculated by discounting future costs and benefits by an annual real rate ofinterest according to the formula PV=Benefit/(1+r)n where r is the discount rate (in this case5%), and n is the number of years into the future.

70. Barnett, W.S. New wine in old bottles: Increasing coherence in early childhood care and edu-cation policy. Early Childhood Research Quarterly (1993) 8,4:519–58.


Long-Term Effects of Early Childhood Programs on Cognitive and ...

Documents

Transcript of Long-Term Effects of Early Childhood Programs on Cognitive and ...