Background

The broad focus of the Early Years Review Group is research on the impact of various policies that promote early education and care. In this third review, we focus on evidence from longitudinal studies of centre-based early childhood interventions. Our question is:

What is known about the long-term economic impact of centre-based early childhood interventions?

There is a substantial literature about cost-benefit studies of social welfare interventions. It is widely assumed, and widely quoted by politicians and policymakers, that early childhood interventions in particular are effective and bring returns in the order of seven dollars saved for every one dollar spent. These savings do not appear to be apparent until the children who received the intervention reach adulthood.

We wished to scrutinise this evidence in detail.

Methods

The Review Group operated at two levels: (i) a core group of academics and practitioners who undertook the research, and (ii) a wider peripheral group, mostly international, which was consulted at various stages in the procedure, and which helped to formulate the research question and write the protocol, and advising about specific issues. The core group and the peripheral group are concerned with early education and care, but because this review also included economic analysis, two economists were recruited, one to the core and one to the peripheral group.

Initial work concentrated on development of definitional statements, and inclusion criteria. Long-term was defined as more than ten years from the date of the intervention so that children should be aged at least 15. Economic impact referred to the outcomes for children (and for mothers of participants) to which a cost has been assigned, including long-term social integration or mental or physical health, rates of incarceration, remedial education, teenage pregnancy rates, employment and earnings. In addition, we limited the studies to those undertaken after 1950 and which were published in English. A search strategy was developed that included service categories, financial and related categories, and outcome/research categories.

The abstracts were scanned to make an initial decision about whether they met the inclusion criteria. Those where determination was positive or unclear were obtained, and where they still met the criteria on examination of the documents, they were keyworded using the standard EPPI Centre and review-specific keywords.

Following this exercise a map of relevant literature was produced. Searching – including database searching, internet searching, citation tracking and hand-searching – produced 4,893 reports. The complex nature of the search meant that it was not possible at the search stage to look with any precision for either long-term economic outcomes or early childhood interventions. As a result, 93% of the abstracts were excluded at the initial screening stage because, although they contained the search terms specified, they were not actually about either early childhood interventions or long-term economic outcomes. Only three studies met our criteria, but because the review examined longitudinal data, these were reported in 58 separate reports. There was not a second stage of inclusion/exclusion using further criteria as we had anticipated and described in our review’s protocol. Three review-specific questions were added to the EPPI Centre data-extraction tool:

A.1. What is the range of domains considered by the cost-benefit analysis?

A.2. What is the evidence used within those domains as a basis for the analysis?

A.3. On what basis are future projections of benefits made?

A synthesis of the results was undertaken based on the quality of the evidence provided.

Results

We found only three studies which deal with the long-term economic outcomes of centre-based early childhood interventions, although there is a considerable secondary literature which refers to them. These studies were the Perry High/Scope, the Abecedarian and the Chicago Child-Parent Centers (CPC), all undertaken in the United States.

Type of study

Two of the studies, Abecedarian and Perry High/Scope, described themselves as randomised controlled trials, where participants were allocated at random to the early years intervention or a comparison condition. They involved small, single site samples. The third study, the CPC, was a matched controlled trial that identified children at entry into kindergarten (i.e. aged 5-6), who had experienced the CPC programme and established a matched comparison group of children who had not. This study had a much larger sample and was multisite.

Timescale of studies

Perry High/Scope was initiated in the 1960s, Abecedarian in the 1970s, and the CPC was an evaluation commencing in the 1980s, of a cohort attending a programme which had been first introduced in the 1960s.

Nature of the intervention

The programmes offered to children were all high quality in that they provided good staff-child ratios and carefully structured curricula. The Abecedarian programme offered longer hours for all children and accepted children into the programme at a very young age. The other two programmes offered part-time, school-term nursery education to children aged three and four years, and parenting support, either through home visits (Perry High Scope) or through required attendance at child-parent centers (CPC).

Study samples

All three studies drew almost entirely on low-income African-American populations. There are some differences in the samples (Table 4.1). The Abecedarian study drew on an especially high risk population, and the sampling frame was based on referrals from other agencies.

Calculation of costs and benefits

Relatively similar programme cost information was given for the three studies, although the costs are not directly comparable since the levels of intervention differed considerably; different information was collected about participant outcomes. The outcomes studied relate in part to the kind of provision under offer. The Abecedarian study offered full-time provision for children aged 0-5 and collected information about mother’s subsequent education and employment patterns, whereas the other two studies assumed that mothers would be available to take part in parent programmes and home visiting; they therefore did not aim to track mothers’ education or employment. The Abecedarian study also included a youth risk assessment survey, which enabled the authors to calculate health benefits, especially from cessation of smoking. The other two studies did not include measurement of health benefits. There were also other differences between the three studies in terms of the outcome measures used and the range of projected savings resulting from the interventions. The projected savings were calculated on the basis of a range of US datasets, not necessarily the same ones across the three studies.

Cost-benefit results

Each study made an overall estimate of the ratio of dollars spent to dollars saved, taking long-term projections of benefits into account. However, it should be noted that these were headline figures and, as such, did not fully reflect the range of variation in outcome measures within each study or across studies. Each study has a slightly different sample, starting in a different decade, varying in the nature of the intervention, and providing a different configuration of outcomes, and therefore potential savings. The savings were costed on the basis of different datasets for each study. The Abecedarian study claims a projected return rate to society of 1:3.78. The results are not highly sensitive to the presence or exclusion of any one outcome, and there were no savings on juvenile justice. The CPC study claims a projected return rate to society of 1:7.14. The most important single benefit comes from juvenile justice savings. The High/Scope study claims a return rate of 1:7.16 for every dollar invested. By far the largest benefit also comes from juvenile justice savings.

Discussion and conclusions

The difficulties of longitudinal studies

Longitudinal studies are problematic for several reasons. They may draw on ideas and assumptions that have become dated or irrelevant over the intervening years: for example, the studies in the review raised problems of relevance concerning race and motherhood, since understanding of these issues has considerably shifted since the 1960s. There is considerable reason to suppose that in the US, the climate of poor neighbourhoods has worsened, owing to drug subcultures, a rise in single parenthood and other socioeconomic issues (Bourgois, 1998; Katz, 2004; Sennett, 2005). In the three studies examined, the Abecedarian appeared to have the most high-risk sample, and showed no crime effects. It may be the case that in higher-risk groups, the juvenile justice savings are minimal or non-existent, so that a worsening of conditions means that interventions are likely to be less effective. For these kinds of reasons, strategies for early intervention that were considered appropriate in the 1960s and 1970s may not be relevant 30 or 40 years on.

Methodologies of interventions

There is also an argument about the need for educational and social interventions to bed down before being evaluated. Newly set up programmes (as with the Perry High/Scope and Abecedarian) may shift over time as the intervention becomes established, and initial difficulties are ironed out. Chatterji (2004) and Finn-Stevenson et al. (1998) have argued convincingly for educational interventions which incorporate multi-method analyses of contextual and site-specific variables over time, as well as an experimental design, such as a randomised controlled trial.

The strengths and weaknesses of cost-benefit analysis

The cost-benefit studies considered here provide sophisticated and complicated economic analyses of a complex range of data, in which many judgements are made about the basis of calculations. As Foster and Holden (2004, p 48) point out, ‘cost-benefit analysis provides a way of prioritizing a potentially very large body of information, focusing on those outcomes that have the greatest potential benefits or costs from a particular perspective’. They enable an intervention to be costed and rational decisions to be made about future investment. Some commentators argue that cost-benefit studies of early interventions represent a significant development in economic research (Currie, 2004; Heckman, 1999, 2000) and may play an important role in policymaking (Lynch, 2004).

But however complex the methodology, at best a cost-benefit analysis only aims to provide a range of plausible estimates, based on the outcomes reported in the studies. In undertaking a cost-benefit analysis, there are always decisions to be made about the timeframe of the analysis, the range of costs and benefits to be included (e.g. the range of crime savings) and the datasets used as a basis of projections. The calculations do not necessarily reflect all the variations in the original data, but the use of estimates and headline figures may give rise to an exaggerated sense of precision by those in the field who refer to, or extrapolate from, the data.

What cost-benefit studies leave out

Finally any cost-benefit analysis is limited to measurable costs and excludes, at least in most contemporary economic literature, more nebulous criteria such as child wellbeing. Recent international work arising out of the Convention on the Rights of the Child suggests that this is an increasingly important issue (CRC: General Comment 7, 2005). Distributional issues, such as social justice and universal access, may be seen as legitimate goals in themselves, regardless of strict economic efficiency (Karoly et al., 2001; Phipps, 2001; PricewaterhouseCoopers, 2004; Cleveland and Krashinsky, 2003).

Do early childhood interventions save money over the long term?

Although there seems to be a general indication from these three studies, and others such as Head Start, that early childhood interventions may make a long-term difference (Currie, 2000, 2004) and save money, the processes involved are relatively unclear. In these three studies of centre-based interventions, the processes appear to be somewhat contradictory. The Abecedarian study is the most extensive intervention, and shows the most marked effects between the experimental and control groups during schooling, but with no impact on crime ratings, which is the major source of cost savings in the other two studies. On the basis of this review, the widespread, international use of the most favourable headline findings, and in particular of the Perry High/Scope study, is unjustified. Apart from the variation within and between studies, and problems of interpretation of the results, especially crime figures, there is also a problem about the context in which these studies were carried out. The targeting of low-income African-American children in ghettoised neighbourhoods, in a period of considerable racial tension, leads to considerable doubts about the generalisability of these interventions outside their original context. For example, Zigler (personal communication, 1986) writes of the time at which the Perry High Scope study was carried out as a period of ‘anguish and change’ for the education of black children.

Policy

Given the much wider range of policy initiatives on early childhood care and education in the UK and OECD countries, compared with the US, the longitudinal cost-benefit studies of early childhood interventions add little to understanding outside of a US context. This is an important point, since the most generous headline figures from these three studies are so widely cited. There is undoubtedly a trend, reflected in many studies, to indicate that early intervention makes a difference to subsequent outcomes, but the misapplication of the findings from these three studies is likely to lead to a diminished, rather than enlarged, understanding of the processes involved, and the contexts in which they can be said to work.

Practice

The results of the three studies here can only be cited with caution. While there may well be long-term outcomes from early childhood interventions, these studies say little about processes, and are based on cost estimates and projections which do not appear to apply directly outside a US context. The implications for practice, especially in relation to race and ethnicity, have not only not been explored, but have been ignored. The argument against targeted interventions are that they are likely to be stigmatising, and therefore unpopular with recipients. Targeting black children, in particular, may lead to accusations of unequal treatment.

Research

There seems little point in trying to replicate longitudinal studies in the UK. Apart from the expense of such studies and the difficulty of obtaining conclusive results, the notion of targeted intervention is itself problematic. On the other hand, it is important to explore different models of providing and costing services. Cleveland and Krashinsky (2003) have suggested that governments have put forward a range of distributional justifications for expenditure on early childhood education and care, and relatively few developed countries have chosen to adopt the model of funded targeted provision within a private market context, as is the case in the US (OECD, 2000). Instead, distributional issues of access and social justice have been of more concern. PricewaterhouseCoopers (2004) have argued for the need to undertake more research on calculating the costs and benefits of universalised services in the UK. More radically, considerations about the rights and status of young children in the here and now, arising from the work on the Convention of the Rights of the Child, imply that the scope of cost-benefit studies in early childhood should be substantially rethought.

References

Bourgois P (1998) Families and children in pain in the US inner city. In: Scheper-Hughes N, Sargent C (eds) (1998) Small Wars. Los Angeles: University of California Press, pages 331-351.

Chatterji M (2004) Evidence on ‘what works’: an argument for extended-term mixed method (ETMM) evaluation designs. Education Researcher 33: 3-13.

Cleveland G, Krashinsky M (2003) Financing ECEC Services in OECD Countries. Paris: OECD.

CRC (2005) Convention on the Rights of the Child. Implementing Child Rights in Early Childhood. General Comment No. 7, 2005. New York: United Nations.

Currie J (2000) Early Childhood Intervention Programs: What Do We Know? Arlington, Virginia: Joint Centre for Poverty Research, National Science Foundation.

Currie J (2004) Viewpoint: child research comes of age. Canadian Journal of Economics 37: 509-527.

Finn-Stevenson M, Desimone L, Chung A-M (1998) Linking child care and support services with the school: pilot evaluation of the school of the 21st century. Children and Youth Services Review 20: 177-205.

Foster EM, Holden EW (2004) Benefit-cost analysis in the evaluation of child welfare programs. Focus 23: 44-49.

Heckman J (1999) Policies to Foster Human Development. Working paper 7288. Cambridge, Masssachusetts: National Bureau of Economic Research.

Karoly, L Kilburn R, Bigelow J, Caulkins J, Cannon J (2001) Assessing Costs and Benefits of Early Childhood Intervention Programs. http://www.rand.org/publications/MR/MR1336 (accessed 19 October 2004).

Katz C (2004) Growing up Global: Economic Restructuring and Children’s Everyday Lives. Minneapolis, Minnesota: University of Minnesota Press.

Lynch R (2004) Exceptional Returns: Economic, Fiscal, and Social Benefits of Investment in Early Child Development. Washington, DC: Economic Policy Institute.

OECD (2000) United States: Early Childhood Education and Care Country Note. Paris: OECD.

Phipps S (2001) Values, policies and the well-being of young children in Canada, Norway and the United States. In: Vleminckx K, Smeeding T (eds) Child Well-Being, Child Poverty and Child Policy in Modern Nations: What do we know? Bristol: The Policy Press.

PricewaterhouseCoopers (2004) Universal Early Education and Care in 2020: Costs, Benefits and Funding Options. Daycare Trust/Social Market Foundation.

Sennett R (2005) Respect: The Formation of Character in an Age of Inequality. London: Penguin.