ArticlePDF Available

Preschool Education and Its Lasting Effects: Research and Policy Implications

January 2008

January 2008

Authors:

William Steven Barnett

Rutgers, The State University of New Jersey

Content uploaded by William Steven Barnett

Content may be subject to copyright.

Preschool Education and Its Lasting Effects:

Research and Policy Implications

W. Steven Barnett, Ph.D.

National Institute for Early Education Research

Rutgers, The State University of New Jersey

September 2008

1 of 35

Preschool Education and Its Lasting Effects:

Research and Policy Implications

W. Steven Barnett, National Institute for Early Education Research

Executive Summary

Over the last several decades, participation in center-based preschool

programs has become much more common, and public support for these

programs has grown dramatically. Nevertheless, participation remains far

from universal, and policies vary across states, as well as across options

such as private child care, preschools, Head Start, and state pre-K. Since

policy makers typically have more alternatives than money, they face key

questions about the value of preschool education, whom it should serve or

subsidize, and which program designs are best. This brief reviews the

research regarding the short- and long-term effects of preschool education

on young children’s learning and development. A detailed and

comprehensive assessment of evidence yields the following conclusions

and recommendations:

Conclusions

• Many different preschool programs have been shown to produce

positive effects on children’s learning and development, but those

effects vary in size and persistence by type of program.

• Well-designed preschool education programs produce long-term

improvements in school success, including higher achievement test

scores, lower rates of grade repetition and special education, and higher

educational attainment. Some preschool programs are also associated

with reduced delinquency and crime in childhood and adulthood.

• The strongest evidence suggests that economically disadvantaged

children reap long-term benefits from preschool. However, children

from all other socioeconomic backgrounds have been found to benefit

as well.

• Current public policies for child care, Head Start, and state pre-K do

not ensure that most American children will attend highly effective

preschool programs. Some attend no program at all, and others attend

educationally weak programs. Children from middle-income families

have least access, but many children in poverty also lack preschool

experiences.

• Increasing child care subsidies under current federal and state policies

is particularly unlikely to produce any meaningful improvements in

children’s learning and development. Given the poor quality of much

child care, it might instead produce mild negative consequences.

2 of 35

• Increasing public investment in effective preschool education programs

for all children can produce substantial educational, social, and

economic benefits. State and local pre-K programs with high standards

have been the most effective, and such programs need not be provided

by public schools. Public schools, Head Start, and private child care

programs have produced similar results when operating with the same

resources and standards as part of the same state pre-K program.

• Publicly funded pre-K for all might produce a paradoxical but

worthwhile effect in terms of educational gains. Disadvantaged

children benefit (in comparison to their gains with targeted programs),

but so do more advantaged children. Accordingly, while such universal

programs may result in higher levels of achievement for the

disadvantaged, they might leave a larger achievement gap. If a

universal preschool program substantially increased the enrollment of

disadvantaged children, however, the achievement gap might also be

reduced. .

Recommendations

• Policy makers should not depart from preschool education models that

have proven highly effective. These models typically have reasonably

small class sizes and well-educated teachers with adequate pay.

• Teachers in preschool programs should receive intensive supervision

and coaching, and they should be involved in a continuous

improvement process for teaching and learning.

• Preschool programs should regularly assess children’s learning and

development to monitor how well they are accomplishing their goals.

• Preschool programs, in order to produce positive effects on children’s

behavior and later reductions in crime and delinquency, should be

designed to develop the whole child, including social and emotional

development and self-regulation.

• Because an earlier start and longer duration does appear to produce

better results, policies expanding access to children under 4 should

prioritize disadvantaged children who are likely to benefit most. More

broadly, preschool education policy should be developed in the context

of comprehensive public policies and programs to effectively support

child development from birth to age 5 and beyond.

3 of 35

Preschool Education and Its Lasting Effects:

Research and Policy Implications

W. Steven Barnett, National Institute for Early Education Research

Introduction

Rapidly evolving preschool education poses challenges for local,

state, and federal education policy. In 1960, just 10% of the nation’s 3-

and 4-year-olds were enrolled in any type of classroom. Less than a half

century later, nearly three-quarters of children enroll in a preschool

classroom at age 4 and about half do so at age 3.

These trends have been

accompanied by growth in private preschool education and child care,

state-funded pre-K, preschool special education, and the federal Head

Start program.

Public programs currently enroll about half of those in

programs at ages 3 and 4. Children are therefore served by programs that

vary widely in enrollment, program design and operation, and this is true

across and even within states. Issues of quality also arise out of this

miscellany. A recent study in California, for example, revealed that state

pre-K offered the highest educational quality, but that educational quality

averaged across all programs, public and private, was relatively low.

This policy brief summarizes research regarding the short- and

long-term effects of preschool education, with particular attention given to

what is known about influences on program effectiveness. This

information is relevant to public policy makers who must decide whether

and how much to support various types of preschool programs, what

standards to set for public programs, and how much funding to allocate.

A Brief Survey of the Preschool Landscape

Nationally, the largest public investments in early education are for

child care subsidies, state pre-K, Head Start, and preschool special

education. About 75% of the nation’s 4-year olds attend a preschool

center, as do 50% of 3-year-olds.

About half of all 4-year-olds, but fewer

than 20% of all 3-year-olds, are in public programs, while about 35% of

both age groups enroll in private preschool options.

The federal Head Start program serves comparatively few children:

11% of 4-year-olds and 8% of 3-year-olds. In 2006-2007, the federal

government spent about $6.2 billion on Head Start (and nearly $700

million on Early Head Start, which serves children younger than 3). State

pre-K programs enroll 22% of 4-year-olds and 11% of 3-year-olds. As

these figures suggest, far fewer 3-year-olds overall enroll in public

programs—8% in Head Start, as noted above, and an additional 3% in

4 of 35

state pre-K. Nationwide in 2006-2007, 38 states and the District of

Columbia funded pre-K programs, spending more than $3.7 billion on

them. Generally, the federal government is not a major funder of state pre-

K, although local education agencies have supplemented state funding.

Head Start and most state pre-K programs have eligibility

requirements based on family income. Head Start limits eligibility to

children in families whose income is below the Federal Poverty Level

(FPL), or who are eligible for public assistance, with exceptions for

homeless children and some others.

Some states use the FPL or some

multiple of it, while others use a percentage of the state median income

(SMI).

Several states have committed to provide access for all 4-year-

olds: Florida, Georgia, Illinois, Iowa, New York, Oklahoma, and West

Virginia. However, only Oklahoma is currently approaching universal

enrollment.

Generally, preschool enrollment rates are lowest for children

in families whose income is above the poverty line but in the lower half of

income distribution.

Children who attend preschool programs have widely varying

experiences.

Public programs vary considerably in operating schedules,

teacher qualifications, class size and ratio, auxiliary services (such as

health and social services, or parenting education), monitoring and

accountability, actual teaching practices, and effects on children’s learning

and development. Teacher qualifications in state pre-K programs range

from little more than a high school diploma to a four-year college degree

with specialized training in early childhood education. Head Start has

national standards for program structure and operation. Private programs

vary greatly as well. State child care regulations are weak everywhere, but

many centers exceed standards, even as others violate them.

With

programs varying so greatly, widely varied effects on children are to be

expected.

Effects across All Types of Programs: An Overview

A substantial body of research establishes that preschool education

can improve the learning and development of young children. Many

studies have investigated the immediate effects of preschool education for

children during their first five years of life. Some have compared the

outcomes for preschool education against other options: outcomes for a

true control group having no preschool education; outcomes for typical

experience (which includes child care outside the home); and outcomes for

more or less well-defined alternatives (for example, outcomes for

preschool education programs using different curricula). With such a large

number of studies, meta-analysis is a useful tool to summarize findings.

As a statistical procedure, meta-analysis can provide a transparent,

quantitative summary of findings that is easy to interpret. However, since

meta-analysis offers little nuance concerning the details, strengths and

weaknesses of each included study, it is also useful to consider a detailed

5 of 35

review of the most rigorous and relevant individual studies. The following

discussion includes both approaches.

Multiple meta-analyses conducted over the past 25 years have

found preschool education to produce an average immediate effect of

about half (0.50) a standard deviation on cognitive development.

This is

the equivalent of 7 or 8 points on an IQ test, or a move from the 30

to the

percentile for achievement test scores. For the social and emotional

domains, estimated effects have been somewhat smaller but still

practically meaningful, averaging about 0.33 standard deviations.

To put

these gains in perspective, it’s important to realize that on many measures,

a half standard deviation is enough to reduce by half the school readiness

gap between children in poverty and the national average.

Dozens of studies have examined preschool education’s long-term

effects, providing information on effects into elementary school and

beyond.

Recent meta-analyses of these find that preschool education has

significant lasting effects on cognitive abilities, school progress (grade

repetition, special education placement, and high school graduation), and

social behavior.

Estimated effects decline as students move from

immediate experience to elementary school, to adolescence, and to

adulthood follow-up. Thus, long-term effect sizes (reported as standard

deviation units for each measure) are smaller, and are roughly 0.10 to 0.20

for cognitive abilities, 0.15 for school progress, and 0.15 to 0.20 on social

behavior including delinquency and crime.

These effect size estimates are averages across studies that vary

widely in rigor and program types included. There is also some variation

in populations served, although most studies have focused on

economically disadvantaged populations. The strongest studies, which are

randomized trials, have examined programs ranging from intensive

“model” programs for children from birth to age 5 to typical Head Start

centers.

The largest estimated effects have been reported by these more

rigorous studies. Also, programs focused directly on educating the child

had greater effect than multi-purpose programs delivering a mix of

services to children and families.

Thus, the average effect sizes across all

studies summarized by meta-analysis are significantly smaller than the

average effect sizes found for well-implemented, intensive educational

programs. For a more finely grained picture, then, this review turns now to

the results of specific studies.

Child Care Effects

Ordinary child care is found to have the smallest initial effects on

children’s learning and development. Typically, family day care homes

show no effect on cognitive development, while child care centers produce

small short-term effects (0.10 to 0.15) on cognitive and language

development.

Several non-experimental studies in the United States and

Canada have found center-based child care to produce small negative

6 of 35

effects on social-emotional development and behavior—in particular,

increased aggression.

There is some evidence that negative effects

increase with number of years in care, but lessen when children attend

higher quality programs.

Higher program quality (defined in terms of

standards and observed practices) is also associated with larger gains in

cognitive and language abilities.

In addition, some studies, but not all,

find larger benefits for children from economically disadvantaged

families.

These findings are from non-experimental studies, however;

effects may be influenced by unobserved differences between children and

families who do and do not use child care.

Finally, evidence suggests

that child care subsidies increase employment for mothers of young

children.

Given the small initial effects of child care, it is not surprising that

the estimated long-term effects are small as well. Some children might

benefit from long-term increases in family income due to increases in

maternal employment, though work could lead mothers to reduce time

with their young children, perhaps partially offsetting income benefits.

The methodologically strongest long-term study of child care’s effects in

the United States finds that both positive and negative effects tend to

decline over time. Persistent effects are few. Greater child care quality is

associated with higher vocabulary scores through grade 5 (an effect size of

0.06), and more time in center-based care increases teacher-reported

behavior problems through grade 6 (an effect size of 0.08).

The weaker

Early Childhood Longitudinal Study Kindergarten Cohort of 1998 (ECLS-

K) studies find that center-based programs have small residual effects on

reading and math test scores until the end of third grade (about 0.05

standard deviations).

Head Start Effects

Two recent randomized trials and a regression-discontinuity study

(a rigorous alternative) have assessed the short-term effects of one year of

Head Start.

The strongest study to date is the Head Start Impact Study

(HSIS) of a large sample of children across the country randomly assigned

to attend Head Start or not at ages 3 and 4.

The estimated cognitive

effects of nine months of Head Start range from 0.05 to 0.25 standard

deviations. Effects tend to be smaller for cognitive measures of broad

domains and larger for measures of limited sets of literacy skills and

knowledge more easily taught and mastered in a brief time.

No evidence

was found of any negative effects on socio-emotional development, and

behavior problems and hyperactivity were significantly reduced (0.13 to

0.18 standard deviations) for 3-year-olds. Access to dental care was

improved and child health, as reported by parents, was modestly improved

(0.12 standard deviations) for 3-year-olds. An earlier, smaller, randomized

trial of Head Start for 4-year-olds in one region of the country found larger

effects on cognitive development. For example, on the Peabody Picture

7 of 35

Vocabulary Test (PPVT), gains were 0.32 standard deviations (compared

to HSIS’s 0.05 to 0.12). The study also found a very large effect on dental

care as well as positive effects on health care.

Another recent rigorous study of Head Start’s initial effects

employed a quasi-experimental, regression-discontinuity design in Tulsa,

Oklahoma. For one year of Head Start at age 4, this study found effects of

0.33 to 0.55 standard deviations on literacy and math assessments. Of

note, the Tulsa Head Start programs are not typical. All lead teachers in

Tulsa Head Start had four-year college degrees and early childhood

teacher certification, and they received public school salaries and benefits.

While it would therefore be inappropriate to generalize these results to

Head Start overall, these findings increase concerns that the Head Start

Impact Study (HSIS) might underestimate potential effects.

One reason for these concerns about HSIS is the influence of

“crossovers” – children whose experience didn’t actually correlate with

study design. At age 4, only 86% of the treatment group actually attended

Head Start, and conversely, 18% of the control group found their way into

a Head Start program elsewhere.

Adjustments for crossovers in the study

yield estimates of 0.08 to 0.36 on cognitive outcomes for children who

actually attended Head Start.

Even this finding is open to some question,

however, since much of the “no treatment” control group attended

programs other than Head Start: at least 22% of 3-year-olds and 30% of 4-

year-olds in the control group attended other center-based programs.

similar concern exists about the earlier, smaller randomized trial discussed

above, in which over a third of the control group similarly attended

another child care program. Overall, it seems reasonable to conclude that

one year of Head Start has initial effects on cognitive abilities that are at

least in the range of 0.10 to 0.30 standard deviations.

Several researchers have used data from the Early Childhood

Longitudinal Study Kindergarten Cohort of 1998 (ECLS-K) to estimate

the effects of Head Start.

Although the ECLS-K provides a large and

diverse sample and several family background measures for use as

statistical controls, the data set has significant limitations including

reliance on parent reports to identify Head Start participation, and the lack

of pre-tests for behavior and academic skills, to control for pre-existing

differences between children in the control and Head Start cohorts. These

studies find that for children entering kindergarten, Head Start has no

significant effects on cognition and negative effects on socio-emotional

development and behavior as children enter kindergarten. It is important to

note that the findings from these studies are precisely the opposite of the

findings of the national randomized trial. A reasonable conclusion is that

the ECLS-K findings differ because of the biases introduced by their

weaker research designs, casting doubt on their longer-term findings as

well.

Some studies using survey data have employed approaches

specifically designed to address problems in selection bias. These have

8 of 35

found positive Head Start effects on achievement, producing estimates of

initial impacts approximating those from randomized trials.

Several have

also looked at longer-term effects. A series of studies comparing children

within the same family who did and did not attend Head Start (as reported

by parents) found effects on achievement at kindergarten entry that were

similar for all ethnic groups. Nevertheless, later effects appear to vary by

ethnicity. They found lasting effects on achievement test scores and grade

retention for white and Hispanic children, but not for African-American

children. Grade repetition for Hispanic children age 10 and older was

found to be substantially reduced.

In addition, they found that Head Start

increased high school graduation rates by 22 percentage points for white

children and decreased arrest rates by 12 percentage points for African-

American children.

However, other researchers suggest that the methods

and data employed in these ESCL-K studies may introduce biases that

could account for some of the variations in outcomes by ethnicity.

One study, which approximates a randomized trial, relies on an

historical discontinuity in Head Start funding across counties to estimate

Head Start’s effects on children’s health and on their educational

attainment.

It finds that Head Start decreased mortality among children

ages 5 to 9 from causes plausibly affected by Head Start health services; in

addition, it increased high school graduation and college attendance.

Based on the health effects found in this study, an increase in Head Start

enrollment from 12,000 to 30,000 per 100,000 children can be expected to

result in one or two fewer deaths. Although no effects were found on

eighth grade test scores, the significance intervals are so wide that the

modest positive effects found in other studies can not be ruled out. The

significant long-term effects are found for boys and girls and for blacks as

well as whites. However, the estimates apply to children who attended

Head Start in the 1960s and 1970s, and it is possible that today’s Head

Start has smaller or larger effects.

State and Local Pre-Kindergarten Effects

Two studies of preschool education that began in the 1960s stand

out. They were well-implemented, randomized trials of public school

programs.

The two programs studied, which are discussed briefly below,

differ from typical state and local pre-K programs in scale and in the

extent of oversight, supervision, and teacher support. Both employed

public school teachers who received intensive coaching and supervision,

regular in-depth discussion, and feedback regarding teaching practices.

Other preschool programs with strong evidence of effectiveness have also

had such teacher support, and it has been suggested that strong teacher

support is likely to be important to replicating positive results.

Teachers

in both programs also conducted home visits. Among numerous other

studies of preschool interventions, only the two discussed below employed

9 of 35

a true experimental design to investigate the persistent effects of preschool

education provided by public schools.

The High/Scope Perry Preschool program randomly assigned 128

disadvantaged minority children to either a half-day preschool program

with home visits by the teachers or a control group.

Children attended

the preschool program for two school years beginning at age 3 (except for

a few who entered at age 4). Classes and ratios were much better than is

typical of most public programs: 12 or 13 children with two teachers. This

staffing made the Perry program considerably more expensive than the

typical public education pre-K program. However, initial effects on

language and general cognitive abilities after two years were impressive:

about 0.90 standard deviations, which is about the size of the typical

black/white test score gap.

The Perry study followed 123 children from preschool well into

adulthood. Researchers’ ability to study nearly the entire original group

over time allows confidence in long-term findings. The initial cognitive

advantage from the preschool program was seen to decline over time,

partly because the public school experience appeared to help the control

group catch up once they entered kindergarten.

While there was no

persistent effect on IQ, the study found a persistent effect on achievement

tests through middle school, a finding consistent with results from the

meta-analysis of all relevant research literature. In addition, the preschool

group had better classroom and personal behavior as reported by teachers,

less involvement in delinquency and crime, fewer special education

placements, and a higher high school graduation rate.

Through age 40,

the program was associated with increased employment and earnings,

decreased welfare dependency, and reduced arrests. Long-term effect sizes

are in the range from 0.30 to 0.50 standard deviations. High school

graduation increased from half to two-thirds, the number of arrests by age

27 fell by half, and employment at age 40 showed an increase of 14

percentage points.

Another study of public school preschool education was conducted

by the Institute for Developmental Studies (IDS). It included a larger

sample of 402 children who were randomly assigned to the public

program or to a control group.

Children entered the preschool program at

age 4 and attended for one year, and afterward entered an IDS

kindergarten program. A teacher and an aide staffed each preschool

classroom of 17 children. Estimated effects at the end of pre-K were just

over 0.40 standard deviations on measures of cognitive and language

abilities; an estimated effect of about 0.20 on these abilities persisted

through at least third grade. The IDS study also provides follow-up

analyses that indicate persistent effects to adulthood on achievement,

educational attainment, and employment. However, the study suffers from

severe attrition, which limits the confidence that can be placed in those

very long-term findings.

10 of 35

A third randomized trial deserving mention examined the effects of

one-year of a half-day university preschool program on 291 children from

advantaged backgrounds. This study experienced higher attrition than the

others, with only 196 (67%) of the original group remaining in the study at

second and third grade follow-up. Moreover, this study had no tests of

children’s abilities prior to program entry. Attrition rates in treatment and

control groups were apparently not affected by children’s gender, IQ, or

social competence. The average IQ of children in the study was 130,

which indicates very high abilities—the 97

percentile. No significant

program effects were found on IQ (although the estimated effect is 3

points, or 0.20 standard deviations), but significant effects were found on

measures of social competence and school readiness.

For second and

third grade, the study found statistically significant gains on achievement

tests for boys, but not girls. The estimated effect size for boys on the total

achievement battery was 0.82 standard deviations.

There are no randomized trials of large scale, state-funded pre-K

programs. However, recent studies have employed a regression-

discontinuity design (RDD) that emulates the results of a randomized trial

under reasonable assumptions.

These RDD studies make use of the birth

date cut-off for school entry to compare two groups of children who both

enter the program (thereby avoiding the selection bias problem) a year

apart, but who differ in age by only one day.

Such studies of the

universal pre-K program in Tulsa, Oklahoma, provide credible estimates

of short-term effects on subgroups, as well as for children generally. These

studies find effects on literacy and math achievement of 0.36 to 0.99

standard deviations.

These positive effects are found for boys and girls,

for white, black, Hispanic, and Native American children, and for children

who do and do not qualify for free and reduced-price lunches.

Effect

sizes are, however, somewhat larger for minority children (ranging from

0.40 to 1.3).

The RDD approach also has been used to estimate the initial

effects of one year of state pre-K on children’s cognitive abilities in

Arkansas, California, Michigan, New Jersey, New Mexico, Oklahoma,

South Carolina, and West Virginia.

The average effect sizes across these

eight states were 0.23 for general cognitive and language ability, 0.31 for

math, and 0.79 for print awareness.

Effects on a test of general cognitive

and language abilities can be directly compared to those in the Perry and

IDS studies. The average initial effect size for one year of these state pre-

K programs is about half that of IDS and one-quarter that of Perry, with

the top-performing state pre-K programs approaching IDS effectiveness.

These eight state programs are not necessarily representative, but they do

provide a broad sample what state pre-K can produce on a large scale.

Unfortunately, this particular RDD approach cannot be used to

estimate long-term effects, so some studies have supplemented it with

other research designs in order to compare children who attended state

pre-K with similar children from the same localities who did not. There is

11 of 35

some evidence that this less rigorous study design underestimates the

initial effects of pre-K, sometimes by as much as half.

Even so, they

have found statistically significant effects persisting through kindergarten

and first grade, although those effects may decline a bit over time.

Also,

these persistent gains are larger on broader domains of learning that are

not quickly mastered.

Children with two years of New Jersey’s relatively

intensive Abbott pre-K program starting at age 3 sustained cognitive gains

through the end of kindergarten (0.42 effect size on the PPVT) that are

comparable to end of kindergarten gains from the Perry Preschool

program.

Other studies of state and local school pre-K programs provide

additional evidence, including longer-term results, but unfortunately most

used non-experimental research designs that offer less protection from

selection bias.

Two of the methodologically stronger state evaluations

used waiting lists to construct comparison groups (thereby creating a

“control” group that is also characterized by parental interest in preschool

education). A South Carolina study using a pre-test of children’s abilities

found a positive effect (0.33 standard deviations) on cognitive abilities at

kindergarten entry, particularly for children whose classrooms had higher

quality as measured by direct observation (0.44 standard deviations).

New York study found initial positive effects on general reasoning, verbal

concepts, and school-related knowledge and skills.

Both studies found

some persistence of effects into the elementary school years; the New

York study additionally found evidence of less retention in grade.

Several literature reviews have summarized long-term findings

from the large body of research literature evaluating state and local pre-K

programs using quasi-experimental methods. Such summaries find a

mixed pattern of positive and null findings on achievement tests, but a

more uniform pattern of significant reductions in special education and

grade retention.

The substantially lower rates of grade retention and

special education for children attending pre-K reported by several studies

actually explains some of the null findings on achievement test scores.

Most studies compare children in the control group with children receiving

pre-K education within each grade level; such comparisons, however, do

not take into account the absence of low-performing children from the

control group who were retained in an earlier grade or moved into special

education.

The most rigorous long-term, large-scale study of pre-K is the

Child Parent Center (CPC) study.

Chicago’s public schools operated the

CPC program beginning in the late 1960s. CPC provides low-income

children with a half-day preschool, kindergarten, and a follow-on

elementary school component. Some 55% of CPC students attended CPC

preschool for two-years beginning at age 3 (the remaining students

attended the preschool for one year beginning at age 4). The preschool

program had a certified teacher and an assistant in each classroom of 18

children, and a relatively strong parent outreach and support component.

12 of 35

This program design is similar to “better” state programs in terms of basic

design and cost. Multiple evaluations have been conducted on the CPC

program, all finding positive effects on children’s learning. The most

recent study constructed a comparison group at kindergarten entry drawn

from children in similar neighborhoods, some attending the same

elementary schools as the CPC children. Estimated effects on test scores at

kindergarten vary from 0.35 to 0.77 standard deviations depending on the

measure and analysis procedures (estimated effects of just one year of

preschool attendance are about 0.20 to 0.65 standard deviations).

The CPC study has a relatively large sample and has maintained its

sample integrity into adulthood; it estimates program effects on a broad

range of outcomes through age 21. Because the half-day CPC program is

broadly similar to the Perry Preschool program (albeit less intensive), it

can be considered a replication of the Perry study, and the extent to which

the CPC study’s findings confirm the long-term findings of the Perry

Preschool study is important. The CPC study finds positive effects on the

following outcomes: test scores through at least middle school, arrests for

delinquency and crime, special education, and high school graduation. The

estimated effects are remarkably similar to those in the Perry Preschool

study, though sometimes smaller. In addition, it finds a significant

reduction in grade retention.

This pattern is what one would expect from

a somewhat less intense dose of the same “treatment” in a different

location. As in the Perry Preschool study, effects on cognitive abilities

decline over time, but as late as eighth grade they are still nearly 0.20

standard deviations. The effects on schooling outcomes are substantial

(effect sizes of 0.23 to 0.34) with a 15 percentage point reduction in grade

retention, a 10 percentage point reduction in special education placements,

and an 11 percentage point increase in high school graduation by age 20.

Additional estimates of the effects of state and local pre-K have

been produced by other studies using the large national ECLS-K data set.

Although the size and diversity of their samples are appealing, such

studies have serious methodological limitations. As noted earlier, studies

employing large national data sets have generally been found to produce

more biased estimates than studies using smaller samples with local

comparison groups.

Moreover, ECLS-K uses only parental reporting to

identify the type of program, and parents were not asked if the program

was a state or local education program. This is a serious problem for

studies trying to look closely at state pre-K, which can encompass private

centers, Head Start programs, and public school programs.

These studies

find positive effects on cognitive development of just under 0.20 standard

deviations and negative effects on social skills and behavior of 0.12-0.29

standard deviations. Longer-term, these studies find pre-K to reduce grade

retention and to increase test scores in third grade by about 0.05 standard

deviations.

When estimates are limited to preschool programs housed in

the same elementary school a child attended for kindergarten, negative

effects on social skills and behavior disappear.

As with Head Start,

13 of 35

comparison to the results of randomized trials suggests that the ECLS-K

estimates are biased downward and assign an inaccurate negative in the

social-emotional domain. Thus, the ECLS-K studies add to the evidence of

persistent positive effects despite their short-comings, but their specific

estimates should carry little weight in policy debates.

Perhaps because many preschool education programs have been

half-day, few studies have estimated their effects on maternal

employment. One study finds that public preschool programs and cheaper

private programs increased employment of single and married mothers of

3- and 4-year olds and that public kindergarten increased employment of

single and married mothers of 5-year-olds.

These estimated increases are

large: from 6 to 15% for employment, hours, and earnings for mothers of

5-year-olds, and over 20% for mothers of younger children. Another study

of the effects of public kindergarten on maternal employment finds

smaller effects for single mothers and no effects for married mothers.

Neither study accounts for whether public kindergarten was part-day or

full-day, a potential problem as half-day kindergarten might be expected

have little or no effect on employment. A study of single mothers who had

received public assistance in Massachusetts found that both the

availability of Head Start and state funding for preschool education for

low-income children increased maternal employment.

Studies of the

effects of preschool education on maternal employment in other nations

have also found positive effects.

International Evidence

Research in other countries confirms many of the U.S. findings

regarding short- and long-term outcomes of pre-K. A randomized trial

with long-term follow-up of high-quality half-day pre-K in Mauritius finds

short-term improvements in children’s learning and behavior, and reduced

crime rates into adulthood.

Rigorous quasi-experimental studies in Latin

America find increased test scores through third grade, as well as

decreased school failure, increased educational attainment, and positive

effects on attention, class participation, and discipline,.

Several studies in

the United Kingdom that are similar to the National Institute for Child

Health and Human Development (NICHD) and ECLS-K studies in the

United States find modest positive effects on cognitive development that

persist for at least several years into school and mixed (but weak) effects

on social development and behavior.

In the UK, entering school prior to

age 5 has been linked with cognitive gains through age 16 and increased

employment at age 33.

Both early schooling and preschool attendance

were associated with increased wage rates (about 3%) at age 33.

Effects

in these studies are similar for children from all economic strata.

A New

Zealand study finds positive long-term effects on cognitive abilities and

some protection from getting into trouble through age 16 for children from

all backgrounds.

UK and New Zealand studies found larger gains when

14 of 35

programs contained more middle class families.

Finally, international

comparisons find that increased duration of preschool education is

associated with higher achievement test scores through age 15 and that

very high participation rates are associated with less within-country

inequality in test scores.

Effects of Very Early Educational Intervention

Researchers also have studied the effects of early education and

child care beginning before age 3, sometimes during the first year of life.

The Abecedarian study employed a randomized trial to evaluate the effects

of a full-day (six to eight hours) year-round educational program from

about four months of age to kindergarten entry. This study followed 111

children from program entry through age 21 with a largely intact sample.

The Abecedarian program produced large initial gains in IQ that have

declined over time, with effect sizes of about 0.75 at age 4, 0.50 at age 5,

and 0.33 at ages 15 and 21. Effects on reading and math achievement

averaged about 0.40 standard deviations from ages 8 to 21, with only a

very slight decrease in magnitude over time. The program had large

effects on grade retention and special education, reducing each by 23

percentage points. The study reports no statistically significant effect on

high school graduation, but high school graduations (excluding GED)

through age 19 were 67% for the treatment group and 51% for the control

group, a 16 percentage point difference. Attendance at a four-year college

was significantly different, 36% for the program group versus 14% for the

control group.

There were broader effects, as well. At the young adult follow-up,

the program group was more likely to have a skilled job, less likely to

have become teen parents, and less likely to smoke marijuana. Effects

were not found on social development or behavior during the program or

in later delinquency and crime. However, control group involvement in

crime and delinquency was low, making it difficult to improve on an

already good outcome.

A 16 percentage point difference in adult

smoking is not statistically significant, but is similar to the observed

difference in the Perry Preschool study. The program group also reported

fewer depressive symptoms at age 21 (effect size of 0.42). Finally, the free

child care appeared to have improved mothers’ long-term employment

opportunities and earnings.

Other randomized trials replicate or partially replicate the

Abecedarian findings. The Project CARE study compared essentially the

same program as Abecedarian to a home visiting model and a no-

treatment control group.

This study is limited by a smaller sample size

(only 16 children in the center-based program, with 14 included at long-

term follow-up), and thus it has sometimes been combined with the

Abecedarian data for analysis. The CARE study finds essentially the same

pattern and magnitude of preschool effects on IQ over time. The home

15 of 35

visitation program had no significant effects. Program effects on

achievement and school progress have not been reported. However, young

adult outcomes are quite similar for four-year college enrollment and

highly skilled employment. When the data from the two studies are

pooled, there is a statistically significant effect on years of education

(effect size 0.43). CARE did not replicate findings regarding smoking or

teen parenting.

Another small-scale replication is a Milwaukee study of a program

providing full-day educational child care to highly disadvantaged children

from before age 1 through kindergarten. This study had only 53 children at

start and 40 for longer-term follow-up, but it approximated a randomized

trial. The study found large initial gains in IQ that declined when the

control group entered school, but remained at about 10 points (0.67

standard deviations) through age 14. Effects on achievement test scores

appeared to decline rapidly during the early school years. Though not

statistically significant, the effect on reading achievement was an

estimated 0.68 grade equivalent, or 10 percentiles. Math scores were

essentially equal for the two groups by grade 4. However, there were half

as many grade retentions (10 v. 5) and substantially fewer special

education placements for the program group by grade 4. The subsequent

absence of the additional retained and special education students from the

control population may have adversely affected the achievement test

comparisons.

A much larger, but less exact replication is the Infant Health and

Development Program (IHDP), a multi-site randomized trial of home

visitation for the first year followed by educational child care modeled on

the Abecedarian program from ages 1 to 3. Children and families received

no services after age 3.

All 985 children in the IHDP study were low

birth weight (LBW), but the sample was divided into low and high LBW

groups. The heavier low birth weight group is more socio-economically

diverse and less educationally disadvantaged than samples in most other

randomized trials.

At age 3, the treatment group had a higher IQ (10

points, effect size 2/3), fewer behavior problems, and a higher rate of

maternal employment. The IQ gain was larger (14 points, 0.93) for the

heavier LBW program group.

Although no effects persist beyond age 3 for the sample as a

whole, long-term gains are found for the heavier LBW IHDP participants.

They have cognitive advantages of about 0.30 to 0.45 standard deviations

at ages 5 and 8, and higher math (but not reading) achievement at age 8.

At age 18, the heavier LBW program participants also had higher scores

on one measure of cognitive ability and language (0.25), but not on IQ, as

well as higher math achievement test scores (0.34) and less self-reported

risky behavior.

No significant effects were found on grade retention,

special education, high school graduation, or arrest rates. However,

estimated effects for the heavier children were similar in size to estimated

effects for the children in the Chicago Child Parent Center study for

16 of 35

special education (reduced from 24% to 17%) and for arrests (reduced

from 26% to 21%).

100

In contrast to child effects, the maternal

employment gain persisted only for the lighter LBW group.

101

Multiple randomized trials have been conducted on early

childhood interventions that seek to comprehensively intervene with

parents and children beginning in the first years of life. These do not

typically provide an intensive educational program to children. Overall,

they have not produced the same magnitude of effects as programs

focused on educating the child, and studies do not find substantive

cognitive effects persisting to kindergarten entry. Examples include the

Comprehensive Child Development Program (CCDP), Avance family

support program, Child and Family Resource Program, New Chance, and

Even Start.

102

The recent large-scale multi-site randomized trial of Early

Head Start (birth to age 3 programs delivering center-based and home

visitation services) is perhaps the most positive. It found small effects on

development (0.10 for cognitive and language abilities) for children age 3

and for parent outcomes.

103

At age 5, no effects persisted on cognition,

language, or achievement; however, small improvements (0.10 standard

deviations) were found for children’s behavior problems, parenting, and

maternal depression.

104

Implications for Policy Implementation

As described above, early care and education programs have

positive effects on young children’s cognitive and social development, and

these effects can be substantial. Rigorous studies find not only immediate

gains, but lasting benefits for learning and educational achievement,

school progress and educational attainment, and social behavior, including

delinquency and crime. The methodologically strongest studies find the

largest effects, and the earliest and smallest studies have been replicated

repeatedly and on a large scale. Research from other countries indicates

that findings of long-term educational and social benefits generalize across

a tremendous range of social, political, and educational environments.

This evidence indicates that policies supporting the provision of effective

preschool education can produce important improvements in children’s

learning and development. However, two policy questions immediately

arise. Who should receive the programs? And, what characteristics should

programs have in order to be effective?

In the United States, the question about who should receive

programs translates to the question of whether programs should be

targeted. One way to address the question is to ask what groups preschool

is known to help. Positive effects have been found for boys and girls,

children from all socio-economic strata, and children from all major ethnic

backgrounds.

105

While the size of some effects varies with children’s

characteristics (including their family background), the largest and most

rigorous studies do not consistently find that effects are limited to one

17 of 35

gender or only to children from low-income families. Larger effects do

seem likely for children from disadvantaged backgrounds on threshold

measures like grade repetition, special education placement, high school

graduation, and criminal arrests. However, about 1 in 10 children from the

middle 60% of the family income distribution repeats a grade and a similar

fraction drops out of high school; due to their sheer numbers it is these

middle-income children who account for most of the school failure

problem.

106

Multiple benefit-cost analyses have been conducted on three of the

intensive educational programs that have been studied long-term—

Abecedarian, Perry Preschool, and CPC.

107

All three find that benefits

substantially exceed costs. Thus, these programs not only achieve

important educational goals, but are sound public investments even if they

are far from optimal, or even if they serve populations with relatively less

to gain than the cohorts studied in these three programs. The value of the

benefits is so high that even if more advantaged children gained as little as

one half—or even one tenth—of the benefits disadvantaged children gain,

a one- or two-year preschool program for them would be a worthwhile

public investment.

There are, of course, other reasons for preferring either a targeted

or a universal program.

108

Some derive from political principles, like

assertions that smaller government is preferable to a larger one, or that

public education in a democracy should be equally available to all. Such

philosophical issues are beyond the scope of this review. In addition,

practical issues complicate the choice. Most targeted programs use family

income to determine eligibility. However, family income presents a

moving target due to the transience of poverty for many families. Many

families move from one side of the cut-off to the other during the school

year; some families manage to enroll their children despite having

incomes above the cut-off; and others who qualify are not even identified

as eligible. By the end of a program year, half of those enrolled may

exceed the income cut-off.

109

Moreover, family income is far from a

perfect indicator of who will benefit from preschool education. Others

have suggested developing measures of “poor parenting” for use in

targeting, but this seems even more intrusive and impractical than means-

testing. In addition, the evidence reviewed here suggests that program

effects on disadvantaged children may be larger when programs serve

children from diverse backgrounds. If the primary concern about

implementing a universal program is the cost of subsidies to higher-

income families, the use of a sliding fee scale can reduce public cost while

accommodating all children.

Regarding how children should be served, it is clear that all

programs do not produce the same gains. Current child care policies and

programs do not provide services of the educational intensity needed to

produce substantial short-term gains, much less long-term gains.

Moreover, there is reason for some concern that under current policies

18 of 35

child care has small negative effects on social and emotional development.

These undesirable outcomes are an important concern because most young

children who are in child care attend private programs with little or no

quality assurance: at best, they are regulated and subsidized as child

care—not education. Head Start and other public programs produce larger

positive effects, and no negative effects. Although Head Start’s effects on

learning may be smaller than those of many state and local pre-K

programs, rudimentary calculations for Head Start based on research

reviewed here suggest that it passes a cost-benefit test.

110

The better state

pre-K programs should pass as well. Private child care is not necessarily

inferior: private child care centers operating under state pre-K standards

with state pre-K funding can produce the same large positive effects as

programs in public schools.

111

In general, research finds that the programs with the largest and

longest lasting effects are the most educationally intensive and expensive.

Additional guidance from research regarding program design is limited.

Others things equal, programs that begin earlier appear to have greater

long-term effects. Only programs that begin at age 1 or earlier have had

persistent effects on IQ, but this does not necessarily translate into greater

long-term benefits across the board. Programs starting in the first year of

life and continuing to kindergarten may or may not have higher rates of

return than programs starting at ages 3 or 4. Programs for infants and

toddlers are much more expensive, and some quite expensive very early

interventions largely failed, apparently because they did not deliver

enough education.

Both half-day and full-day programs have produced strong results,

but only full-day programs produce economic benefits from increases in

parental employment. Employment and earnings benefits suggest the

value of jointly developing child care and preschool education policies.

Examples include extending the length of preschool education day to six

or more hours, offering wrap-around child care in public pre-K, funding

private child care centers to offer pre-K that meets high education

standards, coordinating child care subsidies and public preschool funding,

and connecting family home day care providers and public pre-K in ways

that facilitate children’s participation in pre-K without disrupting (or even

improving) their family home day care arrangements. As a bonus, one

small randomized trial found that an extended day and extended year

produced greater learning gains.

112

Strong guidance regarding other issues of preschool program

design is difficult to glean from existing research. Many studies have

measured the associations between children’s learning and development

and such program characteristics as teacher qualifications and training,

teacher salaries, class size, and adult-child ratio.

113

A number have failed

to find relationships between any of the structural characteristics of

programs and children’s learning.

114

However, none have been

randomized trials. The scant evidence from randomized trials that we do

19 of 35

have contradicts the results from weaker designs. The Tennessee class size

experiment, for instance, found that smaller classes (around 15) led to

greater learning gains in kindergarten.

115

A randomized trial within the

National Day Care Study found that smaller classes and better adult-child

ratios produced more learning.

116

As this review shows, weaker quasi-

experimental designs often produce misleading results and can even get

the direction of an effect wrong (e.g., finding a negative effect when the

true effect is positive). Recent analyses call into question the validity of

even sophisticated value-added studies as a means to address these

questions.

117

Until we have more randomized trials, policy makers would be

well advised to use caution when departing from models demonstrated to

be effective. The research literature does establish that programs with

well-educated, adequately paid teachers, small classes (no more than 20

children), and reasonable staff-child ratios (less than 1:10) have repeatedly

produced strong short- and long-term educational gains. Programs putting

fewer resources into the classroom often have failed to achieve similar

results.

Of course, structure and resources alone are not sufficient for high

levels of educational effectiveness. Careful attention must be paid to the

how teachers actually teach, what children experience and learn in the

classroom, the provision of supervision and coaching, and the engagement

of teachers and those who support them in a continuous improvement

process.

118

A program’s goals and implementation matter a great deal,

though these are complex and not necessarily well-specified by the formal

curriculum. Recent randomized trials comparing curricula produced few

strong findings, with the most frequent result by far no significant

difference. Nevertheless, several of these studies add to the evidence

reviewed above that intentional focus on specific learning goals does

matter.

119

One of the most important matters in this respect is to pay

attention to all of children’s needs for learning and development. Healthy

social and emotional development is as important in life as academic

success. An intentional focus on enhancing social and emotional

development is required for success in this domain, and curricula have

been found to differ in their effectiveness in this domain.

120

Physical

health and nutrition deserve attention, as well.

121

Conclusions and Recommendations

A substantial body of research is available regarding the effects of

preschool education on young children’s learning and development,

including long-term outcomes. Much of the evidence is from rigorous

studies, and findings have been replicated with considerable variations in

program design, populations served, and social context. These studies

provide a sound basis for conclusions about the benefits of publicly

funded preschool education, and they can help inform key decisions about

20 of 35

who to serve and how programs should be designed. Based on a detailed

and comprehensive review of the evidence the following conclusions and

recommendations are offered: