Outcomes of early childhood education: Literature review

Publication Details

This literature review was commissioned by the Ministry of Education to provide policy makers with a synthesis of research that analyses the impact of early childhood education (ECE) for children and families.

Author(s): Linda Mitchell, Cathy Wylie and Margaret Carr, New Zealand Council for Educational Research.

Date Published: May 2008

Please consider the environment before printing the contents of this report.

This report is available as a download (please refer to the 'Downloads' inset box).  To view the individual chapters please refer to the 'Sections' inset box.

Chapter 1: Introduction

In this section, we describe the aim of this review of the outcomes of early childhood education (ECE), and outline the methods we used to select research to use in the review, synthesising what outcomes we covered and our approach to research findings for each outcome. Measuring the outcomes of ECE and establishing ECE's contribution is a complex area, so we then go on to discuss some of the issues around this, including the difficulty of measuring complex outcomes like dispositions and key competencies, although ethnographic studies highlight their importance. We also outline the meaning of effect sizes in quantitative analysis. Since this review is of more recent research, it is important to establish a kind of baseline in terms of the picture that emerges in earlier reviews about the impact of ECE, the differences in impact that occur between studies focused on substantial "interventions" for low-income children and families focused on ECE, and those that look at everyday ECE. We summarise the findings from these pre 1995 reviews for each of the main outcome areas included in this review. Finally, we describe a model from Cunha, Heckman et al. (2005) explaining how ECE effects might endure, and "how skills beget skills".


The intention of this review is to provide policy makers with a synthesis of research that analyses the impact of ECE for children and their families. It addresses three questions posed by the Ministry of Education. These are: 

  1. What (short-term and long-term) developmental, educational, social, and economic outcomes are associated with participation in ECE for learners and their families?
  2. Are different outcomes associated with different population groups and under different circumstances/ contexts? In particular:
    1. What types of institutional (e.g. differences in educational/care systems, types of services), funding and regulatory arrangements/frameworks are associated with achieving positive outcomes?
    2. When are adverse impacts more likely and for whom?
  3. How do different outcomes interact/relate with one another?
    1. What is the size/significance of the different impacts of ECE? How long do the effects last?  

Criteria for inclusion in the review

We include in this review reports of research that met the following criteria:

  • The study is reported after 1995, except for some key works that were reported up to 1995. These are described as "key" if they are closely related to the review topic, and address aspects on which we have limited evidence. We chose the date 1995, since many reviews on outcomes of ECE have examined pre 1995 evidence and we did not think it necessary to duplicate this work. Instead, we have summarised earlier reviews and discussed whether there is a general consensus from them about the impact of ECE. In addition, we were interested in studies that analysed the impact of ECE over and above family socioeconomic characteristics linked to differences in children's outcomes. There are more of these in recent years.
  • The report provides information from a primary study, and is not a literature review. We have used other reviews to: source studies for their insights; use them in this introduction to frame the state of knowledge about ECE impacts, and have copied some summary tables of effect sizes from earlier studies calculated or compiled by earlier reviewers in this field (Appendix B). We are aiming to build on the review work that exists, rather than repeat or ignore it.
  • Research methods, analyses and findings are sufficiently detailed to provide a sound basis for a judgment to be made about the robustness of the conclusions. This includes information on the characteristics of the sample.
  • Information is provided about characteristics of the context, such as types of service, quality, staffing, funding, and regulatory framework.
  • The evidence demonstrates valid linkages between ECE participation and outcomes. We included only studies where there was information about outcomes for learners or parents. We also aimed to focus on studies that provided analysis about the impact of ECE that allowed its effect to be separated from the family socioeconomic resources that are most likely to be associated with differences in children's outcomes. This was not always possible, particularly with cross-sectional studies of parent perspectives on their gains from ECE participation, and with qualitative studies.
  • Where methods are qualitative, the qualitative chain of evidence is robust. Where quantitative approaches are used, there are enough in the sample to justify the analysis and conclusions.

This review is focused on outcomes resulting from ECE experience, as they have been measured in research. This posed some issues for us, because the measurement to date in quantitative studies has been limited to traditional approaches to outcomes of education. Some of the work being done about children's development of valuable dispositions and attitudes that are linked elsewhere (e.g., Wylie, Hodgen, Ferral, & Thompson, 2006) to gains in learning in more traditional and quantitatively measurable areas, such as reading and numeracy, is qualitative. This leading edge work does not approach learning "outcomes" as distinct domains, with a clear separation between cognitive ("academic"), and dispositional and social-emotional domains. Indeed, this is the approach taken in New Zealand's ECE curriculum, Te Whariki, where the outcome areas are wellbeing, belonging, communication, contribution, and exploration. The key competencies in the draft New Zealand curriculum, included there because they are linked to the kind of critical enquiry and problem solving capacity we need for the 21st century (Gilbert, 2005), also cross the traditional distinctions between the cognitive and social-emotional areas. We have included some studies using this new approach in the section on children's dispositions and social-emotional outcomes.

The new approaches to learning also emphasise the importance of a less linear approach to children's development, and more appreciation of the contextual nature of display or realisation of an outcome than is taken in much of the research included in this review. Much of the quantitative analysis by necessity uses statistical models that assume that children develop linearly, and treats ECE experience as an input to a later outcome, rather than as a context in which children form their learning identities-taking something of the context into these identities and also making something new of it in relation to what was already there. Thus the ECE experience is not standard, and its relation to children's learning identities is dynamic rather than fixed.

This review does not include studies of patterns of, or changes in, ECE participation rates unless these also included analysis relating these to outcomes. We have included less qualitative material than we anticipated, because much of it was about change in practice, and somewhat outside our brief. However, we do use some of these studies, where they help develop our understanding of why some aspects of ECE are important in enabling children and their parents to make the most of ECE experiences.

Our search, through the databases and contacts listed in Appendix C, included material we already had on hand, references in that material, and in the new material gained through the searching gave us over 300 articles. One hundred and seventeen of these met the criteria used for this review (p. 11 above). Final decisions on what to include were made by two of the research team. Because the research team included those also working on the Competent Children, Competent Learners study, Megan Clark of the Mathematics Department of Victoria University was asked to decide whether the findings from that study warranted inclusion. She concluded that the findings were robust (see Appendix D). Margaret Carr similarly reviewed three NZCER projects with data on outcomes for parents (her summaries are included in the summary section, Appendix A).

Outcomes for children

We summarised each of the reports of studies that we used in this review, and then used these to compile an overall picture of what the research could tell us (or not) about these outcomes for children:

Outcomes for children:
the traditional "academic" aspects, such as mathematics, literacy, intelligence tests, and also school readiness, grade retention, and special education placement-decisions usually based on cognitive/knowledge performance (though some aspects of dispositions and social-emotional outcomes will be associated with the latter)
Dispositional and Social-emotional:
learning dispositions and orientations such as independence, perseverance, participation, curiosity, and social competence (and negative outcomes of antisocial/worried behaviour)
cortisol levels, infections, and parent reports of health.
Three Outcomes for Parents:
interactions with the child, home learning environment, parental understanding of child learning and development
Parent Life Course Outcomes:
education and training, social support, cultural connectedness, community participation, confidence and family functioning
interactions with the child, home learning environment, parental understanding of child learning and development.
Two cost benefit outcomes for government:
  1. Outcomes from children's ECE participation; and
  2. Outcomes from increased taxation, through parental employment facilitated by access to child care.

We chose these outcomes on the basis of the kinds of evidence that was available in the research, and in consultation with the Ministry of Education.


The summaries of the studies used in this review are given in Appendix A. These summaries were designed for quick reading and reporting of some additional information. We have aimed to give a snapshot of research method, sample, and findings, with a final column which includes relevant material that may aid policy makers, researchers, and others in the ECE sector. Most of this final column material is related to the interpretation of particular study findings, including comments on design and context (sometimes the authors, sometimes our own, and for one outlier study, brief summaries of relevant material from other studies).

We then compiled the syntheses of main findings for each outcome that follow this introduction. We did this by sorting the findings for each outcome into whether they were positive, negative, or showed no impact; adding the number of findings in each category (positive, negative, no impact). We then briefly discuss the trends in the findings, and the overall picture they give us of the research knowledge for each outcome.

These syntheses are also succinct, to allow quick reading, and include a summary for that outcome area. We then bring the results for each outcome area together in the conclusion to provide overall answers to the Ministry of Education's three research questions.

Because the research around ECE outcomes is wide-ranging, and study results can show differences according to the nature of the ECE experiences, the comparisons made, and how something was measured, we move next to discussing some of the issues around estimating the effects of ECE.

Sizing up the ECE contribution to outcomes

ECE provision has become a major policy plank in many countries. This means that there are now more studies being undertaken, in a wider range of social contexts, and using a wider range of methods of analysis. The studies in this review include longitudinal studies of interventions, longitudinal studies of everyday ECE experiences, analysis of databases that include some information about ECE participation (usually limited to whether there is ECE participation or not, length of participation, sometimes type of ECE), survey information from parents using ECE, and cross-sectional studies, often focused on ECE quality. They include studies that take an experimental intervention approach, randomly assigning children from similar backgrounds who were then given different ECE experiences or none; studies that use already existing variations in ECE experience to compare groups with different experiences; and studies that focus on description, particularly in reporting parental perspectives.

Learning dispositions are more difficult to measure than cognitive outcomes. Some quantitative studies have included aspects of these (e.g., Competent Children, Competent Learners). We have also included robust qualitative studies for this outcome. These are longitudinal and ethnographic, where researchers follow closely children's learning pathways, seeking perspectives from families, teachers, and the children themselves. These studies become robust (trustworthy, accountable, empirically valid) through a range of processes that include: unobtrusive data gathering ("natural" social processes are undisturbed); respondent validation (subjects recognise and affirm the findings); triangulation (a variety of types of data are collected); clearly outlining the role of the researcher; the collection of additional structured data (e.g. interviewing the children); the robust nature of the primary data; taking a comparative approach, e.g. "case studies" within the same site; combining the analytic and the systemic (an early childhood setting as a complex system of interdependent and nested sub-units); and using a theoretical framework or model against which to "test" the data and to provide valuable insights for policy.

Some reviews have restricted themselves to only those that use an experimental intervention approach. A recent review of long-term economic impact of centre-based early childhood interventions is the most narrow (Penn et al., 2006), using only three studies. The disadvantage of this idealistic, but also mechanical, approach is that it sets artificial limits on the understanding that can be gained by making connections across different studies, imperfect though they may be judged according to one set of criteria.

Most reviews have done what we do here, and have set the scope wide enough to gain as much understanding of what could be happening as possible. The design of the study and the method of analysis used are not the deciding factors: what matters is whether the study was done well (and reported clearly) within the framework of its design.

The studies that report the greatest impact from ECE are those with the most comprehensive and direct data about ECE experience, particularly its quality, and that follow children over time. Intervention studies, where the ECE experience has been particularly rich, and the target population particularly impoverished, with the most to gain from complementary and different experiences outside the home, show the largest and most long lasting impacts through comparisons of matched learners who had the intervention ECE experience, than those who did not. These studies involve random assignment to either the intervention or control group. They are also the most expensive, and so they are rare.

Recent studies that analysed the impact of ECE over and above family socioeconomic characteristics linked to differences in children's outcomes are valuable. These are more likely to use quantitative analysis.

However, like other aspects of education, it is difficult to isolate the impact of ECE, or to be conclusive that ECE alone has "caused" outcomes for learners and parents. The factors that contribute to children's development over time are manifold, making it impossible to include all of them in any one study. With ECE, "selection" factors are also at work: the things that have influenced whether a family decide a child shall take part in ECE (since it is voluntary), and if so, what kind of ECE is available). Again, research is increasingly seeking to gather information about likely main factors, but cannot always include them for both reasons of cost and respondent burden. The focus of a study can also limit the range of factors included. For example, while early intervention programmes with random allocation to the intervention and to the control group offer one of the best means of distinguishing ECE impacts from the impact of other factors, they are likely to be targeted to high-poverty learners and families, and therefore of less use in finding out whether differences in the size of the impact of ECE are related to large differences in family income. In short, there is no perfect study in this area (nor is there ever likely to be one). Thus it is important to consider a range of studies together, and build connections between the findings of different studies with different designs, focus, and groups involved, as we have aimed to do in this review.

Effect sizes in quantitative studies

Just as each study has included different factors and measures of outcomes, there is also variation in how the size of the impact of ECE has been reported in quantitative studies. Some studies have reported the size in terms of a unit of the measure being used (e.g. scores on a reading test); others have given it in terms of percentage points relative to the mean of a measure (e.g. a difference as a percentage of a mean score); some have quoted one of the possible measures of effect size (see below); and some have given inadequate information on actual size, reporting only which differences were statistically significant. Where samples are small, statistical significance is likely to indicate a non-negligible difference (although in small samples even relatively large difference can be just by chance), but with large samples, very small sized differences can be statistically significant (and moderate to large differences are unlikely to be just by chance).

Where possible with the quantitative studies, we report the size of any impact in terms of effect sizes. There are several different possible measures of effect size, and these different measures need to be interpreted slightly differently. The most commonly used is d, which is a standardised difference between means (the difference between two group means divided by the common standard deviation). Examples of its use are mean cognitive scores in an experimental and control group, or the difference between the mean cognitive score between the highest and lowest socioeconomic groups. Also presented by some authors is f, which measures the trend, upward or downward, across three or more groups (e.g. levels of maternal qualification, or socioeconomic status). Some authors used d where f may be more applicable, as d only measures the difference between the most extreme groups, and f uses information from all groups. Both d and f can be considered to measure the ratio of "signal to noise", however the values (?Table 1) that are commonly used to benchmark a large, medium, or small effect are not the same.

Other measures of effect size include estimates of the proportion of variation in a model that is accounted for by each of the explanatory variables in the model, or, alternately, the extra amount of variability accounted for in a model if a particular variable is included. The correlation between two variables included in a model can also be used to indicate effect size.

We have calculated values of d for studies that did not report effect sizes, but did include enough information for the computation. This was not possible for all studies that did not report effect sizes.

A guide as to how to interpret the different possible effect sizes is given in Table 1. 

Table 1: Guide to interpreting effect sizes
d 0 0.2 0.5 0.8 No limit
f 0 0.1 0.25 0.4 No limit
r (correlation) 0 0.1 0.25 0.4 1.0
% variation accounted for 0 1 6 14 100

Effect sizes in different studies in the same area (the same or similar outcome and explanatory variables) are unlikely to be exactly the same. The main reasons why they vary are the sample size and variability in the explanatory variable. Effect sizes, like other statistics calculated from sample data, will vary from sample to sample, and this variation will be more marked in small samples than in large samples. Effect size measures for cognitive scores across different levels of family socioeconomic status, or quality of child care, would typically be bigger in studies where there was a wide range in socioeconomic status, or in quality of care. Calculation of d for the most extreme groups present in a study that only includes the lowest socioeconomic status families would be expected to give a value that is lower than that from a study that includes a wide range of families (if there is an association between the outcome and socioeconomic status).

The NICHD ECCRN team has noted that:

of the practical importance of research findings that are modest in magnitude are not straightforward, because effect size estimates are affected by measurement, design, and method (McCartney & Rosenthal, 2000). In the health domain, small effects are taken seriously. Consider the fact that the effect of aspirin on reducing heart attack is statistically very small (r2 = 0.001, with corresponding r = 0.034; Rosenthal, 1994), yet the findings have influenced medical practice (NICHD ECCRN, 2003b, p. 1001).

They concluded that:

Even small effects, when experienced by many children, may have broad-scale implications for larger policy discussions (Fabes, Martin, Hanish, & Updergraff, 2000; Jeffrey, 1989). Indeed, the detected effects may have no implications for how any individual child should be cared for or how any individual family functions, but could have implications at broader levels concerning how classrooms, communities, and even societies operate (NICHD ECCRN, 2003b, p. 1002).

What we aim to do in this review is provide a summary of the consistencies in findings around the impact of ECE, and advance likely reasons for differences in findings by looking at differences in the ECE context, or in the study design and analysis.

We start developing this picture by summarising the conclusions of earlier reviews.

Is there a general consensus in reviews of pre 1995 studies about the impact of ECE?

Cognitive outcomes-early intervention programmes 

One of the reasons for the growth in policy interest in ECE is that there is growing consensus that ECE can have positive impacts for children's cognitive growth and school performance. The strongest evidence for this comes from U.S. early intervention programmes. These are the programmes most likely to be covered in reviews that use "scientific rigour" (random assignment to intervention and control groups) as the main criterion for inclusion (e.g. Anderson et al., 2003; Karoly et al., 1998). Anderson et al.'s appendices are particularly useful as they have calculated effect sizes for earlier studies pre 1995, and so we attach them in our appendices. Their estimates for a range of U.S. interventions (Abecedarian, Perry preschool, and various Head Start programmes) give a positive medium effect size overall of around d=0.35 for academic achievement tests, d=0.38 for school readiness tests, and d=0.43 for IQ test scores, a (positive) decrease of around d=0.13 in grade retention, and of around d=0.14 of placement in special education.

Barnett's (1995) summary of reviews of the impact of ECE written in the 1980s &-; early 1990s was that there were meaningful impacts on cognitive ability (the equivalent of an increase in 8 IQ points, and of "similar magnitude" on preschool and kindergarten achievement measures. There were also positive impacts for "socio-emotional outcomes such as self-esteem, academic motivation, and social behaviour" at the end of the interventions. These effects declined and were negligible several years afterwards. However, some interventions did show more persistent effects. These were ones that had more "intensity, breadth, and amount of involvement with the children and their families" (p. 27).

Barnett's own 1995 review of ECE impact after age 8 included 36 studies of programmes targeted at low-income children, that compared reasonably similar groups of children (those who had the programme, and those who did not). The programmes fell into two groups: model programmes, usually of higher-quality and including some parent involvement; and everyday, large-scale (mostly Head Start).

Barnett notes that none of the studies were perfect. For example, seven of the 15 model programmes used random assignation to form comparison groups; but four of these "had attrition rates so high that initial random assignment could have been invalidated". Several had small sample sizes that made it difficult to "detect even fairly large effects". However, his perspective, as with most reviewers, is that it is the overall picture built up through (imperfect) studies that is important.

The overall picture he reports shows:

  • variation in cognitive gains retained using IQ and school tests;
  • Five of 11 model programme studies with achievement data showed gains beyond age 8; of the 21 large-scale programmes, four found no cognitive effects (either short- or long-term), five found only short-term gains (no longer evident at age 8/3rd grade), and 12 found some positive cognitive effects at this age or later;
  • Reductions in grade retention and special education programme assignment;
  • These were found in all the model programme and large-scale programmes that collected this data (or were in environments where these were common practices); and
  • High school graduation-a "large" effect shown for the five studies that had data (three model and two large-scale programmes).

Girls did better than boys on achievement test scores in four model programmes using experimental design (random allocation) and on graduation rates (two of these four model programmes), but there were no gender differences found in the other model programme or large-scale programme studies.

Within these studies, there was limited analysis of different experiences of ECE quality or duration. The few analyses of quality show higher-quality programmes had larger impacts. The Chicago Child Parent Centre study was the only one that allowed comparison of different starting ages. It found no greater gain for children who started the programme at age 3 compared with those who started it at age 4.

Karoly et al. (1998) reviewed the studies of nine model U.S. programmes and evaluations of Head Start programmes (which are less well resourced) to estimate the size of the gains for children, parents, and government (through eventual savings). Estimation of gains is done by comparison; most of the studies used random allocation to the programme and non programme groups. Five of these model programmes were included in Barnett's review. The main findings of interest here are:

  • All the programmes that measured IQ showed short-term gains for model programme participants; none show long-term gains;
  • All six of the programmes that collected educational achievement data showed gains for participants at ages ranging from 7 to 15;
  • Special education placement decreased in three of the seven programmes that measured this (over the course of school education);
  • Grade repetition decreased in two of the seven programmes that measured this (six over the course of school education);
  • One of the two programmes that collected data on school graduation showed a gain for programme participants; and
  • One of the four programmes that collected data on behaviour showed gains (ages 4-7); one showed no effect; and two showed mixed results. Of the four that collected data on crime/delinquency, two showed gains, and two showed mixed results.

They report that the "size of many of the differences could be fairly characterised as substantial" (1998, p. xv). These reviewers also make the point that because most studies measure only a few outcomes, we do not know the full extent of ECE outcomes. Like other reviewers, they also note the differences found between programmes in terms of outcomes may reflect variation in the implementation of individual programmes, as well as differences in programme emphasis.

A recent update of this review included 20 programmes, with similar findings (Karoly, Kilburn, & Cannon, 2005). Three features of programmes that had larger or longer-lasting positive outcomes for children were identified: better trained staff than others; smaller staff/child ratios; and greater intensity (but no optimal number of programme hours could be estimated). The authors caution that while these programmes generally improved outcomes for participants, "they typically do not fully close the gap between the disadvantaged children they serve and their more advantaged peers" (p. xix).

Other reviews of ECE interventions cover much the same studies, and come to similar conclusions: that overall, there are positive outcomes for children from low-income homes who participate in intervention programmes, of a modest to large size; and that the size of the outcomes is related to the quality of ECE experienced.

Summarising eight U.S. ECE intervention programmes, Frede (1995) noted that part-day and full-day programmes seemed to be as effective; that while most programmes ran for two years or more, some ran for less than this. The lowest staff: child ratio of any of these programmes was 1:8; most were around 1:5, and less than that for infants and toddlers. Frede suggested that "Intensity may encompass more than time, also including the concentration that comes from low ratios, home visiting, and coherent curricula" (p. 123). She noted that these ratios were generally better on the whole than everyday ECE (including Head Start), and that the contact with parents was generally more frequent and focused than in everyday ECE: "the approaches identified as effective all increased the contact between teachers and children and gave the teachers greater knowledge about the children in their care, permitting the teachers to tailor their teaching styles to meet each child's individual needs."

Waldfogel (2002) cites evaluations of early interventions in France and Ireland that show some cognitive short-term gains, followed by the "wash-out" effect over the first few years of school, then a longer-term positive impact. The Irish study underlines the importance of context, including local peer culture and opportunities. The Rutland Street project began in 1969 in an area of high unemployment and poverty. It was a two-year programme, providing half-day ECE, with a cooked lunch, for 3-4-year-olds. The evaluation compared the outcomes for programme participants with a control group of others in the same neighbourhood. At the end of the two years, the participants had higher IQ scores and higher scores on measures of preschool readiness; they continued to have higher IQ scores after three years of school (though their scores did fall), but there was no difference in reading performance at age 8. However, at age 16 they were much more likely (two and three times) to take state examinations at secondary school, and just under a tenth took the leaving certificate, compared with none of the control group. There were no differences in school absenteeism or social deviance (Waldfogel, 2002; Nicaise et al., 1999).

Cognitive outcomes associated with everyday ECE

Pennet al. (2006) reviewed nine studies that included child outcomes related to integrated care and education; these were drawn from the U.S., France, Israel, Korea, Norway, and Sweden. Their primary criteria were that the studies were clearly of ECE that was institutional, open for six hours a day, five days a week, with a formally agreed curricular framework and delivery of activities, and included analysis of outcomes. They were particularly interested in studies that provided analysis in terms of differences in attendance, as well as age, social characteristics, and health.

This review is very cautious in drawing universal conclusions from its suite of studies, and emphasises the importance of context. Thus, it notes that the four studies showing that children who had attended all-day ECE from an early age had better cognitive and socio-emotional outcomes than their peers who had stayed at home or received another kind of care, with similar outcomes for children from different home backgrounds were all in countries where the all-day ECE "was part of a universal service in which especially trained staff offered an explicitly educational curriculum to the children. In addition, good parental leave arrangements meant that the children did not usually enter the setting before the age of one year (p. 39).

Boocock's (1995) review of research on ECE in non-U.S. countries also concluded that differences in context made a difference for ECE outcomes. She concluded that ECE benefited children and their families, particularly in countries "with a national policy of providing preschool services to all children and a tradition of ensuring the quality of those services through enforceable regulations" (p 110). The gains were usually greater for children from poor homes.

Leseman (2002) also notes the importance of context, in relation to the question of why some intervention studies targeted at disadvantaged children show clear and sizeable benefits, and others do not: "As programmes … differ considerably in basic strategic design, structural quality, programme content and process quality, this may explain the mixed findings and point out the ways to improve efficacy" (p. 23). He cites Gilliam and Zigler's (2000) statistical meta-analysis (combining results from individual U.S. state-funded half-day preschools and then analysing as a single study) that found while overall these everyday centres showed little impact on children's outcomes, in some states the average effect size did approach the average 0.50 effect size (or half a standard deviation) reported in McKey et al.'s (1985) analysis of Head Start evaluations. These states had higher regulatory standards for staff training, group size, and staff: child ratios.

Leseman cites a 1998 meta-analysis he led that calculated effect sizes for 18 evaluations of centre-based preschool programmes, published between 1985 and 1996. This found:

  • medium effect sizes of d=0.41 for cognitive performance and d=0.49 for verbal performance favouring experimental preschool programmes with everyday preschool or kindergarten;
  • larger effect sizes, e.g. d=0.67 for IQ measures favouring experimental preschool programmes compared with no preschool participation;
  • smaller effect sizes (d=0.20) favouring experimental preschool programmes (the comparative group is not given) for socio-emotional measures;
  • effect sizes were stronger if children started ECE before age three, if they worked with professionals rather than paraprofessionals or parents, and if their ECE programme was developmental rather than didactic; and
  • greater intensity (covering both duration and hours of attendance) was beneficial when outcomes were looked at long term-but not if outcomes were looked at short term

He cites two other meta-analyses reaching similar conclusions about the value of programme intensity long term; with one of these showing little gains from programmes that lasted for two hours or less a week.

One area where there is less consistency of findings is around ECE experiences within the first year of life. Waldfogel (2002) notes that several U.S. studies find negative effects from starting ECE in the first year of life. However, she reports that these negative effects depend on the nature of the ECE experience, particularly its quality, type, and whether it was full- or part-time. Type of care can be a particularly important factor to have clear information about, since many studies of everyday experience are studies of out-of-home or non-maternal care, rather than formal ECE that is provided within policy or regulatory settings.

Learning dispositions and social-emotional outcomes

The U.S. intervention studies have some information about social-emotional outcomes. They predate the recent and growing interest in the development of learning dispositions. Of those studies that do include social-emotional outcomes, there are mixed findings: some showing very positive outcomes; some with no effect; and a couple with short-term negative outcomes. Again, quality of ECE (and linkages between ECE and home) emerge as distinguishing features of ECE experiences with positive outcomes for children.

Anderson et al. (2003), noted that only the Perry preschool study met their criteria for inclusion in relation to this impact; effect sizes from this one study were medium and indicated positive associations with ECE experience: d=0.38 for assessments of social competence and d=0.60 on assessment of risk behaviours.

Yoshikawa (1995) covered 40 programmes for which there were comparison groups in his review of the long-term effects of early childhood programmes on social outcomes and delinquency. Eight were ECE programmes, 23 were parent-focused family support programmes, and 11 provided both ECE and family support. Eleven of these studies provided measures of antisocial behaviour. There was no difference between the programme participants and the control groups for four of the programmes; the programme participants were rated as more aggressive than the control groups at school entry for two of the ECE programmes. In one parent-focused family support programme, parents were less avoidant and angry than the control group at the end of the programme.

However, in all four of the programmes providing both ECE and family support, the long-term outcomes were positive and effect sizes medium to large for the children, in terms of parent or teacher ratings of behaviour, official delinquency, and criminal reports. The four programmes were the Perry preschool (d=0.42 overall effect size for behavioural outcomes), the Yale child welfare project (d=1.13 effect size), the Syracuse University family development project (d=0.48 effect size), and the Houston PCDC project (also d=0.48 effect size). Two of these started in the first year of the child's life; all involved ECE 4-5 days a week, some half-day, some full-day. These were quality child-centred ECE programmes, with strong theoretically based curricula, staff: child ratios of 1:3 for infants and toddlers, and 1:6 for 3-4-year-olds; staff had pre-service and in-service professional development. The programmes' home visits were regular (some weekly, some monthly).

McCartney (2004) notes "increasing evidence that hours in child care may constitute a risk factor for the development of behaviour problems, including aggression" (p. 3). This conclusion seems to be largely based on the NICHD study. McCartney observes in relation to this study's finding of increased problem behaviour at 54 months and in kindergarten (the first year of school in the U.S.) that "the effects are relatively small, that most children with extensive child care experience do not have behaviour problems, and that the direction of such effects is not clear-in other words, parents with more difficult children may enrol their children in child care for more hours" (p. 3). McCartney emphasises the need to understand why this may be occurring-since the effect was unrelated to the quality of non-maternal care-and speculates about whether large group sizes may "increase the frequency of acting out behaviours that go unnoticed, and therefore uncorrected, by caregivers" (p. 3).

The link between ECE quality and children's behaviour appears differently in Peisner-Feinberg's (2004) summary of research. She cites 13 studies (both intervention and everyday) showing "modest to moderate" links between good quality ECE and social skills at the time of ECE attendance, and two showing this in early school years. She also cites four studies that found little effect of ECE at the time, and two in early school years. The reasons for these different findings may be that the six studies showing no effect (either on social skills or cognitive) had more limited variability in ECE quality or relatively small sample sizes, or in the outcomes measured (e.g. measuring social skills in terms of "very low-frequency behaviours such as social withdrawal (p.3).

Love et al. (2003) amplify this point about how different findings reflect different contexts by contrasting findings from the NICHD study, Israel and Australia, and showing how they reflect differences in ECE quality and government regulation. These studies are included in our review.

Outcomes for parents

Evidence from interventions

There are fewer analyses of outcomes for parents, with most evidence available from intervention studies. The size of these effects is usually smaller than the size of effects for children. Karoly et al. (1998) found no negative outcomes for parents from the nine intervention studies included in their review. Of the six studies that had analysed changes to parenting behaviour, three showed improvements for the parents involved in a programme. Two of the four studies that had data on maternal educational attainment showed gains for mothers when the child was aged 5.

ECE and maternal employment

Labour force participation rates are generally lowest for mothers of preschool children. Childcare use allows maternal employment for this group. Reviews of ECE contribution to maternal employment related to everyday ECE have noted the importance of seeing ECE as part of a complex jigsaw determining whether mothers of preschool children will participate in the labour market. ECE affordability, accessibility, and quality play a part in employment decisions, but so too do other factors such as employment conditions, pay rates, parental leave policies, welfare benefits, taxation, and views about maternal roles (Gustafsson & Stafford, 1995; Wylie, Podmore, & Murrow, with Meagher-Lundberg, 1997). Gustafsson & Stafford report several studies showing that reducing ECE costs to parents in three countries (U.S., Sweden, and Netherlands) was more likely to increase ECE use, or a shift to higher-quality ECE, among those already using it than among those not using it. Studies included in this review also note increases in ECE use and a shift to more formal types of ECE among current users when ECE is made more affordable.

Two of three intervention studies included in the 1998 Karoly et al. review showed increased maternal employment when the children were taking part in the intervention. The authors also cite a 1992 review of 27 early intervention programmes (some home-based, some centre-based) that found increased maternal employment at the time of the programme in 10 of 11 studies that measured this; the impact was "modest" (Karoly et al., 1998, p. 70, referring to Benasich, A.A., Brooks-Gunn, J., & Clewell, B.C. (1992). How do mothers benefit from early intervention programs? Journal of Applied Developmental Psychology, 13, pp. 311-362).

ECE costs and maternal employment

Kohen, Forer, and Hertzman (2006) note the finding from analysis of two Canadian national datasets showing that the labour force participation of this group in the late 1980s was influenced by child care costs, particularly in relation to taking full-time work. Centre-based care in this policy environment was more likely to be chosen by mothers who worked full-time, had higher levels of maternal income, and whose youngest child was 2-3 years old, rather than an infant.

A review of more recent U.S. research on the relationship between child care subsidies targeted to low-income mothers and their labour force participation argues that employment and child care use decisions are probably made together, and that it is therefore difficult to assign causality to child care subsidies in relationship to employment. These child care subsidies seemed mostly to take the form of reduced fees, that is, they address access but not necessarily provision or quality. That said, Lawrence and Kreader (2006) conclude from the studies they reviewed that low-income mothers who use child care subsidies are more likely than their peers who do not to be in paid employment, work more hours, have more stable employment, with standard hours, and earn more. They were also more likely to return to work earlier after childbirth. These studies also found that mothers with low education levels were most likely to increase employment, as were single women. Lawrence and Kreader note that the research they cover is not experimental, and that the wide range of effect sizes found in this research is likely to reflect different approaches to modelling (including simulated subsidy use, where data on actual child care subsidy use was unavailable). They also note that different policy settings are also likely to play a role in different findings. Three experimental studies currently beginning in the U.S. should allow more testing of the direction (or simultaneity) of the relationship between child care subsidies and employment.

In our review, we have been able to include several quasi-experiments following policy changes from other countries that do indicate that improving access to affordable ECE supports maternal employment, though not necessarily immediately (suggesting that the employment/child care use decisions may not be as intertwined as Lawrence and Kreader surmise). Other 1990s U.S. studies cited by Queralt, Witte, and Griesinger (2000) also suggest that the availability of affordable and reliable ECE affects employment decisions.

Understanding how ECE outcomes can endure

In a review of 153 studies of empirical literature on skills formation, Cunha et al. (2005) developed a formal model of the economics of investing in human capital that describes two mechanisms: self-productivity and complementarity. These are multiplier effects which explain how "skills beget skills". Self-productivity says that skills that develop in one period persist into future periods; skills are self-reinforcing. "For example, self-control and emotional security may reinforce intellectual curiosity and promote more vigorous learning of cognitive skills" (p. 5). Also "Students with greater early cognitive and non-cognitive abilities are more efficient in later learning of both cognitive and non-cognitive skills. Thus the enriched early environments of the Abecedarian, Perry and CPC programs promote greater efficiency in learning in high schools and reduce problem behaviours" (p. 65). Complementarity implies that early investment has to be followed up by later investment in order for the early investment to be productive. Facilitating environments have to follow facilitating environments.

This model also emphasises the role of non-cognitive skills. "Non-cognitive skills (perseverance, motivation, self-control and the like) have direct effects on wages (given schooling), schooling, teenage pregnancy, smoking, crime and achievement tests" (p. 85). Cunha et al. emphasise the contribution of family environments and add that the returns to investing early in the life cycle are high.


  1. Originally, we used the three research questions to order this report, but this fragmented the evidence relating to each outcome, making it harder to get an overall picture of the impact on ECE for each outcome. We therefore made the outcomes the focus, and provided pictures of the research evidence relating to each outcome, before bringing this evidence together for each research question in the conclusion.
  2. Shpancer (2006, pp. 227- 237) notes that "effect size is still reported primarily by referring to Cohen's traditional small-medium-large power distinction-an expert opinion based on neither a mathematical formula nor specific and clearly established links to relevant developmental outcome." Shpancer's sceptical view of whether statistical research can establish firm links between ECE and outcomes includes a reminder of the weight that social context and changes, values, and multiple needs of parents, children, and society play in the reality of ECE experiences and its role in any given society, as well as in different children's and families' lives.
  3. Leseman (2002) notes, however, that few children from very poor homes or from ethnic minorities were included in these programmes.
  4. This is only available in Dutch, and we were therefore unable to get a more complete picture of its findings.
  5. In Leseman's references, this study is shown as submitted for publication, but we have not been able to track it down.

Contact Us

Education Data Requests
If you have any questions about education data then please contact us at:
Email:      Requests EDK
Phone:    +64 4 463 8065