Child and mother holding hands

This report studies the impact of Sure Start, which supported families of under-5s, on children’s behaviour, youth offending and social care contacts.

Downloads

Executive summary

Introduced in 1999, Sure Start was an ambitious, large-scale early years programme in England aimed at improving the life chances of children, particularly those growing up in poverty. The programme’s reach peaked in the late 2000s, with a network of around 3,300 centres operating as ‘one-stop shops’ for families with children under 5.

Sure Start centres offered a wide range of services, from baby weighing clinics to childcare provision to employment support for parents. These services were designed primarily to target school readiness and children’s health, and recent evidence suggests the programme was successful in achieving these aims: in a series of reports, Cattan et al. (2022) and Carneiro et al. (2024a) document positive impacts of Sure Start for child health and school attainment, particularly for children from the most disadvantaged backgrounds.    

Given the efficacy of the Sure Start programme for health and educational outcomes, a natural question is whether it had broader impacts on children. This report details the findings from a robust evaluation of the impact of access to Sure Start on children’s absence and suspensions at school, youth offending and contacts with the children’s social care system.

Missing school, committing a crime or experiencing social services involvement can entail significant welfare costs for children. There is a case that investment in joined-up services and early intervention can prevent children from experiencing these poor outcomes. For instance, the Independent Review of Children’s Social Care (MacAlister, 2022) highlighted the potential of tailored services based in community settings to contribute to earlier identification of families in need and reduce social services intervention. It is important to understand whether an integrated early years programme delivered in local neighbourhoods, such as Sure Start, was able to influence the need for costlier interventions, such as those delivered through children’s social care and the youth justice 

Key Findings

1. Access to a nearby Sure Start centre between ages 0 and 4 significantly reduced youth crime that resulted in convictions or custodial sentences. Living within 2.5 kilometres of a Sure Start centre reduced the share of 16-year-olds who had ever received a criminal conviction by 13%. Meanwhile, custodial sentences – the most severe sanction – fell by a fifth due to access to Sure Start. Reductions in youth offending were concentrated on convictions for theft, the most common category of offence (20% reduction), and for drug offences (20% reduction).

2. While access to Sure Start reduced serious youth crime, it had more mixed impacts on less severe contact with the criminal justice system. Those with access to Sure Start committed offences earlier – a 10% increase in less serious misdemeanours by age 12 – and saw rises in cautions for criminal damage and violent crime, although overall numbers of young people experiencing cautions by age 16 were unchanged.  

3. Misbehaviour also increased within school settings: the proportion of children suspended from secondary school increased by 10%, and absence rates increased by 7%. Part of the increase in poor behaviour, both in schools and for younger adolescents in the criminal justice system, may reflect a diversion of children away from more severe offences towards lower-level infractions, but it also likely represents an increase in misbehaviour for some children. This could align with evidence that group-based childcare, a key component of Sure Start’s services, can adversely affect the behavioural development of some children.

4. Access to Sure Start had no significant effect on referrals to children’s social services or on receiving support as a child in need (CIN) or as a child looked after (CLA) between ages 7 and 16. Children in care during late primary school (age 7 to 11) did spend around 13% less time being looked after if they had access to Sure Start during their first five years of life, potentially indicating that children’s needs were somewhat less severe or that they benefited more quickly from support from social services. 

5. The youth justice system and children’s social care involve significant costs for government, as well as the individuals involved. We estimate that for every pound spent at its peak in 2010, Sure Start averted approximately 19 pence in public spending on youth justice and children’s social care, equivalent to £500 million (in today’s prices) of savings per cohort attending at the time. Savings mostly come from costs of youth custody and children looked after, reflecting the high costs of these intensive interventions (and so the large financial benefits of reducing need for these institutions). Future work will provide an overall cost–benefit analysis of the programme, incorporating the effects on educational achievement and health identified in our previous work, while taking account of how these different domains relate to one another to avoid double-counting benefits.

1. Introduction

Early childhood is a critical period for developing the skills that lay the foundations for positive outcomes later on in life. Inequalities emerge early on, particularly between more- and less-deprived children. In the UK, for instance, by age 3, children from the richest fifth of families are more than three times as likely to have high cognitive skills and fewer emotional and behavioural difficulties compared with children in families from the poorest fifth (Cattan et al., 2024). International evidence over the past 50 years has demonstrated that policy can be effective at helping to ameliorate some of the consequences of poverty and deliver long-lasting benefits for children.

Sure Start, a major early years programme in England initiated from the late 1990s, is one such policy. Initially targeted at deprived neighbourhoods, Sure Start delivered a network of centres offering integrated care and services for families with children under the age of 5. It was designed as a single location in which parents could access a wide range of services supporting mother and child health, well-being and development. Decades on from the introduction of the programme, a number of evaluations have built a picture of the medium-term benefits of Sure Start, which narrowed the disadvantage gap in academic attainment (Carneiro et al., 2024a) and health outcomes (Cattan et al., 2022), with effects sustained many years after children attended the centres.1

This report extends our understanding of the impacts of Sure Start by studying the effect of the programme on school absences and suspensions, youth offending and children’s contacts with social care.2 We focus on these outcomes as they are important for the welfare of children: children who come into contact with social services or the youth justice system, or who miss school, are often vulnerable and are having difficult times in their lives. Other important negative consequences of crime include damage to its victims, and the costs of the criminal justice system.

There is also evidence that crime outcomes and social services contacts can be amenable to interventions early on in life, including (for crime) evidence from Head Start, the integrated early years programme in the United States from which Sure Start was adapted (Anders, Barr and Smith, 2023). Meanwhile, evidence from Germany finds that access to early years childcare reduces the number of child protection cases (Sandner, Thomsen and González, 2024). Finally, interventions in the youth justice and children’s social care systems are costly from a public expenditure perspective, making it important to understand what works in order to avert the need for these interventions.

2. History of Sure Start

Services offered at Sure Start

Sure Start centres were intended to act as a ‘one-stop shop’ for young children and their families. This meant they delivered a holistic set of services covering many different aspects of children’s and parent’s well-being. These included family and community health services (e.g. ante-natal and post-natal services), early learning and play opportunities (e.g. stay-and-play sessions, toy libraries), childcare places, parenting support, parental employment support, as well as additional support for children with special educational needs.

Evidence from Meadows et al. (2011) on the centres’ budgets and from the Evaluation of Children’s Centres in England (Goff et al., 2013) on usage of services suggests that play, learning and childcare, and healthcare services accounted for a large share of Sure Start activity. While some Sure Start services were new to an area, in many cases the role of Sure Start was to make already existing types of support for parents of young children more accessible by bringing them all together under one roof, so that those attending for play or childcare would have easy access to other services they might otherwise be less likely to use.

Sure Start rollout

The rollout of Sure Start began in 1999, with the opening of 22 Sure Start centres, known as Sure Start Local Programmes (SSLPs). By 2003, 437 centres had opened. From 2004 onwards, the government rebranded SSLPs to Sure Start Children’s Centres (SSCCs), which entailed some changes to the offering, and a much faster expansion of the programme.3 By 2010, the number of centres had increased more than sevenfold, to 3,290. Carneiro et al. (2024a) show that these centres, which opened after 2004, had smaller positive effects on academic outcomes than those that opened earlier.

In this report we are not able to fully distinguish the impacts of these two types of programmes. In particular, some of the main outcomes considered in this report are only measured in late adolescence, such as juvenile crime or contact with social care at 16. These outcomes can only be observed for participants in the earlier versions of the programme, primarily SSLPs (in the future, with the ageing of more recent participants in Sure Start, we will be able to estimate the impacts of SSCCs on these outcomes and compare the impacts of these two generations of Sure Start).

Box 2.1. Prior Sure Start research

There have been two government-commissioned evaluations of Sure Start, which evaluated the short-term effects, as well as previous work looking at the medium-term effects in recent years.

The National Evaluation of Sure Start (NESS) focused on the effect of SSLPs on outcomes up to age 3. It found no significant overall impact on children’s cognitive outcomes, but found evidence of an improved home environment. It also found variation in the effect on behavioural problems by background – children of teenage mothers who lived near an SSLP developed worse behaviour, while other children saw improvements in behaviour. 

The Evaluation of Children’s Centres in England (ECCE) focused on later SSCCs. It found that use of SSCCs was correlated with improvements in the early home learning environment, and use of childcare specifically was correlated with improved cognitive outcomes. It also found improvements in the mental health of mothers, and some improvements in behaviour at age 3, attributed to the improved home learning environment.

More recently, Cattan et al. (2022) showed there was a short-run increase in hospitalisation in infancy, followed by a larger reduction in hospitalisations throughout childhood and adolescence – effects were concentrated in the most disadvantaged households. Carneiro et al. (2024a), meanwhile, found significant improvements in educational outcomes, particularly in disadvantaged households.

 

The rollout of Sure Start started with more-deprived areas: SSLPs aimed to reach the 20% most-deprived wards. The rollout of SSCCs achieved this goal by 2006, after which the 30% most-deprived wards were targeted. By its final phase, from 2008 to 2010, the rollout aimed to open centres in all but the 30% least-deprived areas. However, many less-deprived areas, which were near to more-deprived areas, would have had access to a Sure Start centre much earlier because centres had an open access policy.4

3. School absences and suspensions, youth offending and children’s social care in England

In this chapter, we provide context around the outcomes we are studying, including a discussion of how school absences and suspensions, youth offending and children’s social care contacts materialise, the systems in which they operate and how many children are affected. Where possible, we use statistics from cohorts of children born just before the opening of a Sure Start centre in their area to give a sense of how prevalent these outcomes were on the eve of Sure Start’s introduction. While the prevalence of some of these outcomes has changed over the last 25 years, the institutional background remains broadly similar today.5

3.1 Absences and suspensions

Suspensions from school are one of two types of exclusion: exclusions can refer to temporary exclusions (suspensions) or permanent exclusions from school. Formal permanent exclusions are very rare, with less than 0.1% of students being permanently excluded from school. However, students may also leave a school after poor behaviour without being permanently excluded, either to move to another school or to move a Pupil Referral Unit (PRU), though it is not possible to track these moves in the version of the data we use.

Suspensions from school are much more common: around 5% of all students in state secondary schools were suspended from school at least once per school year, for the cohorts born just before the introduction of Sure Start.6 Suspensions have a range of official causes, but consistently the most common causes are persistent disruption, verbal assault or harassment towards an adult at the school, and physical assault of a fellow student. This highlights that suspensions from school often result from a different type of misbehaviour to criminal sentencing – they are almost exclusively given for acts that take place within school or relate to other members of the school, with a focus on those types of misbehaviour that reduce other pupils’ ability to learn (such as persistent disruption). This means that while suspensions and criminal behaviour are correlated, we should not expect changes in criminal behaviour to necessarily imply changes in the kind of behaviour that would lead a child to be suspended from school, or vice versa.

Absences, in contrast, have a wide range of justifications. School absences can be separated into those that are authorised by schools, and those that are not authorised. Table 3.1 shows the full breakdown of justifications in secondary schools, where unauthorised absences are more common. Authorised absences are most frequently reported as being a result of illness, while unauthorised absences often stem from children deliberately skipping school – though they can also reflect parents taking their children out of school without authorisation from the school. Thus, changes in levels of authorised absences from school can reflect changes in children’s health (or changes in parents’ approach to children’s health), while changes in levels of unauthorised absences most likely reflect a change in misbehaviour by children.

Table 3.1. Secondary school absence rates by reason, Autumn 2009

 

Absence rate

Overall absence rate

6.9%

  of which authorised:

5.6%

     – of which illness

4.2%

     – of which medical appointments

0.4%

     – of which other authorised

1.0%

of which unauthorised:

1.3%

     – of which unauthorised holiday

0.1%

     – of which other unauthorised

1.2%

Note: 2009 is chosen as the last year for which all secondary school pupils were born before the introduction of Sure Start.

Source: Department for Education (2024a).

3.2 Youth justice system

The broad purpose of the youth justice system is to prevent children offending, and to respond when juvenile crimes are committed. From age 10, the age of criminal responsibility, young people can receive a punishment for any offence they are adjudged to have committed. As with the wider justice system, contact with the youth justice system can occur at different levels of severity (though guidelines for young offenders generally promote less severe sanctions). A stylised overview of the youth justice system is provided in Figure 3.1.

Figure 3.1. Youth justice system in England  

Note: Stylised illustration of pathways through the youth justice system. Green-shaded pathways are observed in administrative criminal justice system data available to researchers. Yellow-shaded pathways are not observed.

  • Young people who offend may not be caught, or – conditional on being caught – the police officer responding may choose to take no formal action, instead giving them an informal warning or other non-statutory disposal. These cases will not be captured in our dataset. 
  • The lowest tier of formal sanction is a youth caution. This acts as a formal warning: it constitutes an admission of guilt and appears on the young person’s criminal record but does not carry any additional punishment. Youth cautions are dispensed at the discretion of the police and are often the preferred response where offences are less serious (for example, low-level criminal damage, possession of controlled drugs, or low-level theft) and the offender does not have a previous criminal record. 
  • If the offence is too severe for a police caution, or if an offender already has a significant history of criminal behaviour, then the accused has to make a court appearance, where they can be found either guilty or not guilty. A guilty verdict normally comes with a sentence. More than half of these are community sentences, but others can include fines, conditional discharges or, in serious cases, custodial sentences. If the individual does not receive a guilty verdict, the court appearance is not captured in our data. 

By age 16, just over 10% of young people born in the years just before the opening of a Sure Start centre – the baseline cohorts in this report – had some formal sanction by the criminal justice system. Of these, 65% had only received a caution, while the rest had been convicted of a crime in court by the age of 16. Among the same cohort, 0.5% of all young people had received a custodial sentence before the age of 16. In the last ten years, rates of entry into the youth justice system have fallen by 70%, though there has been a spike in entry for certain types of crime, particularly weapons offences (Ministry of Justice, 2024).  

3.3 Children’s social care system

Children’s social services, delivered by local authorities, comprise a continuum of support for children and families designed to safeguard and promote the welfare of children. The interventions available to social services include not only the removal of children into the care of the local authority, which is relatively rare, but also the provision of specialist services for children and families and support from a social worker for children classed as ‘children in need’.

Figure 3.2 provides an overview of how children come into contact with the children’s social care system and what types of support they can receive. It abstracts from movement between different types of intervention by the social services, which is also common. Children are brought into the purview of social services following a referral, which typically originates from public services such as police, health services, schools and early years settings where staff are trained to identify safeguarding concerns. While estimates vary, it appears that a significant proportion of children experience an interaction with social services during childhood. One model estimate suggests that around 2 in 5 children are referred to children’s social care by age 16.7

Figure 3.2. Children’s social care system in England 

Note: Stylised illustration of pathways through children’s social care. Omits movement between CIN and CLA and within CIN. Green-shaded pathways are observed in administrative children’s social care data available to researchers. Yellow-shaded pathways are not observed.

Following a referral, children undergo a brief initial assessment designed to triage children into appropriate services. Around the time social care outcomes are observed for cohorts born before Sure Start’s introduction, 14% of referrals are judged to require no further action by social services; the remainder progress to a more in-depth needs assessment (Department for Education, 2015).8 At this stage, there is another opportunity to ‘step-down’ children who do not meet the threshold for statutory support (around 20% of referrals; see Department for Education, 2015). Many of these will be directed towards targeted family services (early help) and universal services, including Sure Start.

For children assessed to be in need (CIN), local authorities are required to set out which services children will receive under a CIN plan or child protection plan. Support for families may include advice and counselling, financial support, provision of housing services, visits from healthcare professionals and respite care. At a given point in time, around 1 in 29 children9 (or roughly one child in every class) are in receipt of services from children’s social care. In the most extreme cases – and subject to legal proceedings – children may be taken into care of the local authority, most commonly in the form of foster care. This intervention is much rarer, with 1 in 175 children looked after (CLA) around the time pre-Sure Start cohorts were interacting with the social care system (Department for Education, 2009). Although providing support for CIN and CLA is a statutory responsibility for local authorities, the levels of need at which children receive different types of interventions vary, both across local authorities and over time, depending on capacity in the system (Webb and Bywaters, 2018).

Children are assessed as in need of support from children’s social care for a number of broad reasons, as pictured in Figure 3.3, which breaks down children in need by primary need assessed.10 By far the most common reason for receiving support is due to risk of abuse or neglect, which includes children living in households with domestic violence present. This makes up around half of children in need. Intervention by social services also arises where parenting is deemed chronically inadequate (family dysfunction) or where parenting capacity is temporarily diminished (family in acute stress), for instance due to poor parental mental health. Less prevalent reasons for interacting with social care include parent disability or illness, low income, absent parenting and socially unacceptable behaviour. A minority of children in need (around 6%) receive support via social services where there is no safeguarding concern but they have high health needs or disabilities. While Figure 3.3 highlights the ways in which the parent and family environment shape the need for intervention by social services, it is also true that child-level factors such as cognition and behaviour play a role in influencing the responses of adults around them, particularly if parents share some of the vulnerabilities with their children (Viding and McCrory, 2024).

Figure 3.3. Children in need by primary need assessed

Figure 3.3. Children in need by primary need assessed

Note: Authors’ analysis of CIN census. For children aged 5 to 16 born in 2002. ‘Other’ includes parent’s disability or illness, socially unacceptable behaviour, absent parenting, low income, as well as cases other than CIN and not stated.

3.4 How might Sure Start have affected these outcomes?

Sure Start was designed as a holistic intervention with services aimed at supporting children and families across a number of dimensions. Its overarching purpose was to provide investments into children to support their early development and to facilitate wider investments in children by supporting families. Centres did this both through the direct provision of services and through connections to wider partner services. These ranged from parent and child health services, such as enhanced health visiting and ante-natal classes, to stay-and-play sessions and childcare to parent-focused activities, including parenting programmes and links to Job Centre Plus.

There are several reasons to think Sure Start may have had an impact on absences from school, school suspensions, offending and contacts with children’s social care. While some potential mechanisms are only relevant to a particular outcome (for instance, staff at Sure Start centres referring children to social services), these outcomes could also be driven by a shared set of mechanisms, such as changes in children’s behaviour and home environment.

Table 3.2 provides a mapping of services offered by Sure Start to potential channels through which Sure Start may have influenced absences and suspensions, youth offending and social services interactions.

Table 3.2. Possible impacts of Sure Start on school absences, school suspensions, offending and contact with children’s social care

Mechanisms

Absences and suspensions

Offending

Social care contacts

Early learning and childcare

Changes in children’s socio-emotional skills

↓ or ↑

↓ or ↑

↓ or ↑

Health services

Early identification and support for children with high health need and disabilities

 

 

Parenting support

Improvements in parenting and home environments

Outreach and changed peer relationships

Engaging hard-to-reach families and changing social links within communities

↓ or ↑

↓ or ↑

↓ or ↑

Identification of unmet need

Centre staff referring children to social services

NA

NA

Note: A downwards arrow indicates a fall in outcome; an upwards arrow indicates an increase. ‘NA’ indicates that the mechanism is not relevant for the outcome.

First, Sure Start centres delivered activities to support children’s early development and provided childcare, potentially replacing lower-quality informal care. These interventions can be effective at improving early skills, particularly social and emotional development, which previous research shows are an important determinant of outcomes such as criminal behaviour (e.g. Heckman, Stixrud and Urzua, 2006; Blattman, Jamison and Sheridan, 2017). Indeed, Head Start – a similar programme to Sure Start in the United States – was found to reduce criminal behaviour in adulthood for poor children attending the programme (Anders et al., 2023). One evaluation of pre-school in the United States studying similar outcomes found fewer school suspensions and lower incarceration rates for children attending (Gray-Lobe, Pathak and Walters, 2023). At the same time, however, there is evidence that childcare can lead to a worsening of behavioural outcomes, particularly for children from more-advantaged backgrounds (e.g. Baker, Gruber and Milligan, 2008; Fort, Ichino and Zanella, 2016). The effect of Sure Start on children’s behaviour and, in turn, on absences and suspensions, offending and need for intervention by social services is therefore ambiguous and may differ for different groups of children.

Second, Sure Start brought together a range of health services for parents, babies and children, from ante-natal classes to baby weighing and health visitor checks. These services could lead to earlier identification and support for children with high health needs and disabilities, reducing health-related absences from school or the need for support via the social care system. Earlier research finds that Sure Start reduced the share of children receiving support for special educational needs and disabilities (SEND) and boosted educational attainment (Carneiro et al., 2024a). These improvements in children’s human capital (health and education) may have also made crime less attractive to adolescents due to improved earnings prospects, thus resulting in falls in offending.

Third, centres offered programmes to improve parenting skills and had links to local job centres to provide assistance with parental employment. Previous evaluations suggest parenting programmes contributed to improvements in the quality of the home environment and parenting (Meadows et al., 2011; Maisey et al., 2013), reinforcing direct investments in children from Sure Start services and facilitating longer-lasting benefits beyond the period children were attending the centres. These improvements in the family environment could translate into fewer absences and suspensions, less offending behaviour and less need for intervention by social services. Increases in parental employment may have increased financial resources in the household, although Cattan et al. (2022) find no impact of Sure Start on maternal employment and income.11

There were also features of Sure Start that may have amplified these effects. An element of the programme, which was shown to be particularly important in centres opened under the first phase of the programme (SSLPs), was the emphasis on outreach and efforts to engage hard-to-reach families. Carneiro et al. (2024a) find that these approaches may have contributed to deliver greater benefits for the poorest children, among whom rates of absence, offending and social care contacts are elevated. Centre-based programmes can also increase interactions between children and families within the local community. Sure Start may have therefore changed or strengthened children’s peer groups, in turn shaping young people’s behaviour. These peer effects could run in either direction: early exposure to less-disruptive peers could improve children’s behaviour, while stronger social connections or exposure to lower-skilled peers could make some forms of misbehaviour more common (e.g. Carrell and Hoekstra, 2010; Lavy, Paserman and Schlosser, 2012; Carneiro et al., 2024b).

At the same time, there is a question of whether a programme such as Sure Start – which was designed to address gaps in child development and improve school readiness – affected more severe experiences for children. Evidence from a similar programme in the United States, Head Start, found that early investments in children did result in large falls in adult criminality (Anders et al., 2023). However, more intensive interventions (such as the Family Nurse Partnership home-visiting programme) in England have had no impact on child maltreatment or social care usage (Robling et al., 2021). Children who are suspended from school or who engage with children’s social care often have complex needs, and hence a lighter touch intervention may struggle to shift the dial on the drivers of suspensions and intervention by social services.

In the case of children’s social care contacts, an additional consideration is the potential of Sure Start to have resulted in higher referrals to children’s social care, as staff had opportunities to screen for concerning parent–child interactions and to identify families who might benefit from social services support. As will be discussed in the next chapter, unfortunately, the data in our study mean we are unable to look at children’s social care contacts under the age of 5 and hence we cannot provide direct evidence for this mechanism. Support for this channel, however, comes from another study that finds local authorities with greater exposure to SSCCs had higher rates of children entering care under the age of 5 (Anderberg and Olympiou, 2023).

A final important takeaway from Table 3.2 is that mechanisms could push in different directions, meaning that the overall effect (if any) on these outcomes is ambiguous. Where possible, we try to shed light on which mechanisms might be at play by studying outcomes more directly linked to particular mechanisms.

4. Methodology

4.1 Evaluation approach

This report aims to estimate the causal impact of children and their families having access to Sure Start services during the first five years of their life on a range of outcomes, across criminal behaviour, school-based discipline and children’s social care usage. To do this, we follow the Carneiro et al. (2024a) definition of ‘access’ to a Sure Start centre, and the same empirical approach to estimating the effect of access, with some adjustments.

Defining access to a Sure Start centre

Carneiro et al. (2024a) define having access to Sure Start as living within 2.5 kilometres of a Sure Start centre between the ages of 0 and 4. This is in line with the aims of the Sure Start programme, which aimed to help families within ‘pram-pushing distance’ (HM Treasury, 1998). We follow the same approach, and their implementation of it. We focus mostly on children who were ‘fully exposed’ to Sure Start, meaning that there was a Sure Start centre within 2.5 km for the entirety of the first five years of their life, though we also produce estimates for ‘partially exposed’ children in Appendix A.1. Because of data limitations, we use information on where a child lived at age 5, the earliest we observe their postcode, as a proxy for where they lived between the ages of 0 and 4.

As in previous work, this approach brings difficulties, as we should not expect every child within 2.5 km of a centre to have used Sure Start, and we should expect some children living further away to have regularly attended Sure Start centres. As Carneiro et al. (2024a) show, this should mean that the effect of having access to Sure Start, as we define it, is smaller than the effect of actually using a centre.

Estimating the causal effect of access to a Sure Start centre

As explained above, the allocation of Sure Start centres across locations is not random. The first Sure Start centres were in the poorest areas, and as the programme expanded, it moved to less-deprived areas. This means that in order to obtain causal estimates of the effect of Sure Start on crime, absences, suspensions and children’s social care use, we cannot simply compare children and families residing close to or far from a Sure Start centre. Such a comparison would confound the impact of the programme with differences in the underlying characteristics of children and families living in more- and less-deprived areas. Similarly, it is also invalid to compare outcomes of children in a given area before and after a centre opened – such a comparison would confound the impact of Sure Start with the changes in outcomes of different cohorts, which would have been observed even in the absence of Sure Start.

Therefore, for a credible evaluation of Sure Start, it is necessary to abstract from pre-existing differences in areas, and from unrelated trends in local areas. To do this, we use a difference-in-differencesmethodology. This methodology compares the outcomes at older ages of children fully exposed to Sure Start with both the outcomes of children in the same area born earlier (and so not exposed to Sure Start) and those of children in other areas without centres born at the same time. It compares differences in the outcomes of children between cohorts born before (both partially exposed and not exposed) and after (exposed) the opening of a Sure Start centre in their area of residence (i.e. within a radius of 2.5 km), with the differences in the outcomes of the same two groups of children residing in an area where there was no new opening of a Sure Start centre (i.e. differences across cohorts that are observed independently of the existence of Sure Start), isolating the changes in outcomes that are unique to children in areas that had received a Sure Start centre. For these difference-in-differences models to estimate the causal impact of Sure Start, we require that the trend of outcomes in areas with and without Sure Start centres would have been similar in the absence of the policy.

While we can never test this assumption directly, we can look at trends in outcomes across different areas beforeSure Start was introduced to give us a sense of its plausibility. We find that this assumption does not hold in the raw data: areas that would soon have a Sure Start centre saw the crime outcomes of their children improving more quickly than areas that had longer to wait, for instance.

Fortunately there are some procedures that help correct these imbalances in the data. We follow the approach proposed by Callaway and Sant’Anna (2021) and reweight our treatment and comparison groups, weighting more heavily the areas with Sure Start centres that look more like those without centres in terms of deprivation and high rates of low birthweight and teenage pregnancy, and weighting more heavily, on the same measures, areas without centres that look more like the areas that have centres. For criminal outcomes, we also weight more heavily areas with similar trends in youth crime rates in generations not exposed to Sure Start. As we show in Appendix A.1, with such adjustment, pre-treatment trends in outcomes in areas that would soon have a Sure Start centre become a lot more similar to trends in areas that had longer to wait. A more detailed explanation of this method we use can be found in Carneiro et al. (2024a).

4.2 Data

The analysis conducted for this report stems from five administrative data sources. This includes data on the opening of Sure Start centres, data on children’s home postcode at age 5, and different administrative data sources for each set of outcomes we consider.

Data on proximity to Sure Start

To construct the distance between each postcode in England and each Sure Start centre, we rely on data provided by the Department for Education on the exact location and opening date of all SSLPs and SSCCs. We then aggregate this to an LSOA (Lower Super Output Area) level and define an LSOA as exposed to Sure Start at a particular point in time if all postcodes are within 2.5 km of a Sure Start centre at that time.12 Using this definition of what a treatment is means we have comparable birth cohorts over time in each area, so that we can account for area-level differences over time.

In order to obtain data on proximity of individual children to Sure Start, we use data from the school census dataset in the National Pupil Database (NPD). This includes data on children’s home postcode in each school year, starting from age 5. Combining this with data on proximity of different postcodes and LSOAs to Sure Start, we have data on whether the address each child lived in when they were 5 was exposed to Sure Start when they were aged between 0 and 4. Because we do not have addresses of children earlier than age 5, we use these as a proxy of where children lived between the ages of 0 and 4.

Data on children’s characteristics

The school census dataset also gives us useful demographic information on children, which we incorporate into our analysis of which groups are most effected. This includes data on gender, ethnicity, whether English is their first language, and eligibility for free school meals (FSM). The last of these gives us information on whether children were in the approximately 15% of families in England with the lowest incomes at age 5. We also link the NPD to official statistics on area-level deprivation, rates of teenage pregnancy and rates of low birthweight, which were used as criteria for selecting the locations of Sure Start centres.

We focus on cohorts in the NPD born between 1990 and 2009. This means we observe some birth cohorts born before 1995, who would not have lived near a Sure Start centre at any point during their first five years of life, and so are only in the control group.

Crime outcomes

In order to estimate the effect of Sure Start on criminal behaviour, we use the link between the NPD and the Police National Computer (PNC) database from the Ministry of Justice. The PNC database includes the criminal history of all children matched to the NPD, covering all recordable offences for which they received formal sanction between 2000 and 2017. This enables us to look at offending outcomes for children born up to 2007 as they reach the criminal age of responsibility (age 10) in 2017. However, for most of the report, we focus specifically on crimes committed up to the age of 16 – this involves only looking at the outcomes of those children born up to 2001, and so living in an area that received a Sure Start centre in the first three years of the rollout. This restriction is what prevents us from comparing the effectiveness of early and late versions of the Sure Start programme, as in Carneiro et al. (2024a).

The offences captured in the PNC database include both those for which the offender received a police caution, and those for which they received a guilty verdict in a court appearance.

For guilty convictions in court, the PNC database also includes data on severity of sentence received. In order to get a sense of the most severe offences, we categorise sentences into custodial and non-custodial, where a sentence is custodial if it involves detention in either a young offender or adult institution. We use custodial sentences as a marker of more severe criminal behaviour – generally, this reflects more serious offences, but it can also reflect a string of other offences.

The PNC database also includes data on the specific offence for which individuals were sentenced or cautioned. We group these into broader categories of offences. The most common categories of offence we consider are theft, drug offences, battery, weapons offences and violent crime.13

Absence and suspension outcomes

For absences and suspensions, we use specific datasets within the NPD, which track information on individuals’ absences from school and spells of suspension from school. For absences, this allows us to construct variables indicating what percentage of time in school children miss, both in total and broken down between unauthorised and authorised absences. We report these absence rates at the end of primary school and the end of secondary school.

For both suspensions and permanent exclusions, we combine data on each individual exclusion with information on the total school population from the NPD, taken from examination data at ages 11 and 16. This gives us rates of exclusion and suspension from school at these two ages, which are the final year of primary and secondary school, respectively. For suspensions, we also try to capture the severity and frequency of suspensions from school, by estimating the effect of Sure Start on the total number of sessions (half-days) of school missed due to suspension from school.

Children’s social care outcomes

To measure children’s contacts with the social care system, we rely on two administrative datasets containing records of children in contact with social services, linked to school census records.

The first dataset is the child in need (CIN) census, which contains case information for children referred to children’s social services for a needs assessment. As can be seen in Figure 3.2, this will exclude the 14% of referrals that do not meet the threshold for a needs assessment (Department for Education, 2015). The data contain the date of the referral, the outcome of the needs assessment and the closure date of the episode.

Using information on children’s date of birth (from the school census) and the dates of social care episodes, we can identify whether a child has a referral to social care at each age from 5 to 16. By additionally excluding referrals that are assessed as ‘not in need’ at the point of the needs assessment, we can measure whether a child has an episode as a CIN at a given age. The data also record the primary need assessed for these children and we can measure the duration spent as a CIN, providing an indication of the severity of the case. We group the reasons for receiving support from social services into abuse/neglect, family dysfunction, family in acute stress, child disability and illness, and other/missing.

This categorisation of need is imperfect. In practice, there is substantial overlap between categories but only the most severe reason is recorded as the primary need. In addition, recording practices may vary across both local authorities and social workers, meaning that primary need assessed may reflect practice and recording culture as well as underlying need (Preston et al., 2021). While it is less likely that Sure Start directly affected recording practices, the effects of Sure Start on the composition of primary need cannot be straightforwardly interpreted as changes in underlying need.

The second dataset is the child looked after (CLA) dataset, which contains a subset of children from the CIN census who are in the care of the local authority. The information provided is similar to the CIN data: in each year, it contains the start date of the most recent period of care, the duration of the period of care and primary need assessed. We use these data to construct whether a child experiences a period in care between ages 5 and 16, and the average duration of the most recent period of care, and we can split by the primary need attached to the period in care. Because the CLA census contains the most recent period of care in a given year, this may miss episodes where children leave and re-enter within the same year. Mc Grath-Lone et al. (2020) find that among children born between 1992 and 1994, around two-thirds of children did not re-enter out-of-home care, suggesting that this limitation does not affect the majority of children who experienced care. We outline our approach to cleaning the CIN and CLA census data in Appendix A.1, as well as broader caveats around data coverage in the early years of the CIN census.

Limitations of the data

While these data provide micro-level data on a wide range of outcomes, there are some limitations that mean there are some effects of Sure Start that we cannot precisely estimate. While the data can capture suspensions and formal permanent exclusions, we cannot observe attendance at PRUs. Students often move to PRUs for poor behaviour even if they have not formally been excluded, but we are not able to capture this in the data.

For crime outcomes, we face the natural problem that we only observe the subset of criminal behaviour that ends in individuals being apprehended. Thus, estimates in this report only capture the full effect on all criminal behaviour, but only that effect on that subset of criminal behaviour.

There are also limitations with the social care data. The most important is that social care information for under-fives is incomplete as we require children to have started school (or to have attended an early years setting recorded in the early years census) before we can link in their school record. Accessing children’s school records is essential to our research design as this provides information about where a child lived at age 5, from which we can derive how close they lived to a Sure Start centre. Our analysis, therefore, considers the medium- to long-run impact of Sure Start on children’s social care contacts during primary and secondary school rather than during the period children were attending centres. This is a limitation of the study given that rates of social care intervention are high amongst infants and very young children (Department for Education, 2022).

An additional limitation of the CLA and CIN censuses is that different birth cohorts have different coverage at each age. Studying the impact of Sure Start on social care contacts at different ages, therefore, draws on different sets of birth cohorts. We study social care contacts during Key Stage 2 (KS2; ages 7–11) using children born from the early 2000s to 2009, and we study contacts with social care during secondary school (ages 12–16) using children born between the mid-1990s and 2005. One important difference across samples is that when we study social care contacts during secondary school, children in the sample would have been mostly affected by the earlier phase of Sure Start (i.e. SSLPs) that operated differently to later SSCCs. By contrast, when we study impacts on social care interactions during KS2, we use a sample of children born later who were affected by SSCCs. More detail on sample selection can be found in Appendix A.2.

Moreover, this reflects a more general limitation across all our datasets: we are limited in our ability to study differences in impact between the early phase of Sure Start (SSLPs) and the later phase (SSCCs). As outlined in the history of Sure Start, the programme operated differently during these two phases, which is a potentially important dimension of heterogeneity (Carneiro et al., 2024a).  

5. Results

Absences from school, school suspensions, offending and contacts with children’s social care range from events that most children would expect to experience at some point (school absences) to rarer interventions by schools, the youth justice system or children’s social services. We present results in terms of increasing severity of outcomes,14 focusing first on behavioural issues within a school setting (absences and suspensions), then turning to the effect of Sure Start on youth offending and experiences with children’s social care.

Overall, we find rises in absences and suspensions during secondary school amongst children who lived closer to a Sure Start centre under the age of 5, and we see no aggregate reduction in low-level misdemeanours (cautions). At the same time, however, we find substantial reductions in children exposed to Sure Start committing offences leading to serious sentencing. It appears that at least some of this shift comes from an increase in poor behaviour among those who would not otherwise have engaged in serious criminal behaviour. This suggests that for some children, Sure Start actually worsened behaviour. However, for others, there were large benefits, decreasing the most serious forms of misbehaviour. There is limited evidence for Sure Start affecting the numbers of children interacting with children’s social care, but we find reductions in the average amount of time children spend as CLA. This could be consistent with Sure Start shifting children ‘upstream’, away from the most intensive interventions.

5.1 Suspensions and absences from school

Beginning with less severe forms of poor behaviour, we start by looking at the effect of Sure Start on exclusions from school. As these are the only ones for which we are capable of detecting effects, we focus on the approximately 98% of exclusions that are temporary ‘suspensions’, and are related to poor behaviour in school (rather than to misdemeanours outside school, as we capture in the crime data).

Figure 5.1 shows the effect of living near a Sure Start centre on suspensions from school at ages 11 and 16, reflecting the end of primary and secondary school, respectively. It shows that at the end of primary school, there was no significant effect on suspensions from school. However, at age 16, exposure to Sure Start generated significant increases in the number of children suspended from school. The left chart shows that suspension rates rose by 10%, or about 0.5 percentage points, for those exposed to Sure Start.

Figure 5.1. Effect of living near a Sure Start centre on probability of being suspended and number of sessions suspended for, at ages 11 and 16

Figure 5.1. Effect of living near a Sure Start centre on probability of being suspended and number of sessions suspended for, at ages 11 and 16

Note: Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Pre-Sure Start levels for suspension rates (left chart) are 1.2% at age 11 and 5% at age 16. Pre-Sure Start levels for sessions missed are 0.11 per year at age 11 and 0.387 at age 16. This means that the average student at age 16 missed 0.387 half-days of school due to suspension from school.

Source: ONS.

The right chart shows what this meant for the amount of schooling missed: exposure to Sure Start led to 25% more time in school missed through suspension. While this increase is statistically significant, it equates to only an extra 0.005 days of school missed per student, because the baseline level of school missed through suspension is so low. Nevertheless, it appears school-based misbehaviour in secondary school became more common for young people who lived closer to a Sure Start centre. This could reflect that for some children there was a shift towards worse behaviour because of Sure Start, or that there was a change in the type of bad behaviour children engage in towards worse behaviour in schools.

To get another sense of what Sure Start did for behaviour-related school outcomes, Figure 5.2 shows the effect of Sure Start on both authorised and unauthorised absences from school, at ages 11 and 16. It shows that, at age 11, there was a small but significant drop in the number of days of school missed for authorised reasons, with no change for unauthorised reasons. The decrease in authorised absences perhaps reflects fewer illness episodes, consistent with findings in other

Figure 5.2. Effect of living near a Sure Start centre on absences from school at ages 11 and 16

Figure 5.2. Effect of living near a Sure Start centre on absences from school at ages 11 and 16

Note: Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Pre-Sure Start levels for authorised absence rates were 4.1% of sessions missed at age 11 and 4.9% at age 16. Pre-Sure Start levels for unauthorised absences were 0.5% of sessions missed at age 11 and 2.2% at age 16.

Source: ONS.

work on Sure Start of improved health for affected children (Cattan et al., 2022). However, at age 16, the effect reverses, and children exposed to Sure Start were absent from school more, for both authorised and unauthorised reasons. As a percentage of the baseline, the effect is similar on both types of absences, but the total number of authorised absences is around 2.5 times larger than unauthorised absences, so the increase due to Sure Start was larger for authorised absences.

These increases suggest that Sure Start’s negative effect on behaviour at the end of secondary school extends beyond just suspensions, as the number of children missing school for unauthorised reasons also increased. The explanation for the increase in authorised absences is less clear, as Cattan et al. (2022) – using a different methodology, looking at the density of centres rather than the effect of being close to one specific centre – find a reduction in hospitalisations for teenagers, suggesting that this result is unlikely to be driven by worse health. The increase in authorised absences might therefore also reflect changes in behaviour among children.

A key reason why suspensions and absences from school are of concern is because of their effect on academic performance – a wide range of work has shown that missing school can damage grades (Goodman, 2014; Aucejo and Romano, 2016; Cattan et al., 2023). However, previous work on Sure Start has shown that, on aggregate, exposure to Sure Start significantly improved overall academic performance up to age 16. This suggests that while these increases in absences and suspensions from school likely had a negative effect on the children affected, they were not enough in this case to completely mitigate the positive effect of Sure Start on academic outcomes through other channels.

5.2 Youth offending

In this section, we focus on the effects of Sure Start on more serious behavioural problems, which we consider by estimating the effect on interactions with the youth justice system. Figure 5.3 reports the effect of exposure to Sure Start on criminal offending, broken down by severity of offence. Results are reported as a percentage of baseline levels of each outcome.

Figure 5.3. Effect of living near a Sure Start centre on the probability of receiving each type of sentence up to age 16

Figure 5.3. Effect of living near a Sure Start centre on the probability of receiving each type of sentence up to age 16

Note: Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Pre-Sure Start levels for total offending rates were 10.5%. Caution rates were 9.4%, conviction rates were 3.7% and custody rates were 0.5%. Results reflect percentage change in the number of children with each type of offence. Custodial sentences are also included within convictions.

Source: ONS.

Overall, Sure Start only generated a very small reduction in total offending rates of young people. The number of children with some criminal history by age 16 was only 2.5% lower due to exposure to Sure Start. This difference is not statistically significant. However, this masks significant shifts in the severity of the offences young people commit.

We find that access to Sure Start had no significant effect on the share of young people receiving a police caution before the age of 16. Instead, it seems any reduction in offending is concentrated among offences ending in more serious punishment. We find significant decreases both in court convictions and in custodial sentences. Having access to a Sure Start centre within 2.5 km reduced the percentage of young people who received a conviction by 16 by over 10%. Receiving a conviction before the age of 16 is relatively rare: among those children born just before the opening of centres we consider, just under 4% received a conviction in this time period – this decrease is therefore equivalent to 2,900 fewer children receiving a criminal conviction nationwide.

The effect of Sure Start was largest on the probability of children receiving a custodial sentence before the age of 16, which was reduced by a fifth by exposure to a Sure Start centre. Prior to the opening of a centre, 0.5% of young people had received a custodial sentence – a drop of a fifth in custodial sentences nationwide would mean approximately 550 fewer young people with a custodial sentence by age 16. However, given the severity of the punishment and of the offences that cause them, the large drop caused by Sure Start represents a major positive effect.

Taken together, these results suggest that Sure Start did not have a particularly large effect on the overall probability of having some criminal history by the age of 16. However, it changed the composition of those with a criminal record, and it significantly reduced the severity of offences, with a shift away from convictions and custodial sentences.

Impacts on different types of offences

Figure 5.4 breaks down the effect of Sure Start on different types of crime. We separately consider the six most common categories of offence: theft is the most common (with around 53 thefts per thousand young people born in the years before an SSLP opened), followed by battery, drug offences, violent crime, weapons offences and criminal damage. Collectively, these make up more than half of all convictions of young people. For each offence, cautions make up around half of all offences.

Figure 5.4. Effect of living near a Sure Start centre on the number of cautions and convictions for each type of crime per young person up to age 16

Figure 5.4. Effect of living near a Sure Start centre on the number of cautions and convictions for each type of crime per young person up to age 16

Note: Black error bars represent 95% confidence intervals. Prior to Sure Start, the average young person had 0.053 theft cautions and 0.046 theft convictions, 0.021 battery cautions and 0.020 battery convictions, 0.008 drug cautions and 0.009 drug convictions, 0.009 violent cautions and 0.007 violent convictions, 0.004 weapons cautions and 0.005 weapons convictions, 0.006 damage cautions and 0.004 damage convictions. These categories make up more than half of all cautions. Young people with zero convictions are included.

Source: ONS.

Our results show that the drop in convictions was driven by a decline in thefts and drug offences. The biggest drop in total number came from theft convictions, which fell by nine per thousand young people. Meanwhile, convictions for battery, weapons offences, other violent offences and criminal damage offences do not seem to have been affected by Sure Start.

In contrast, we find that for several categories of offence, namely violent offences and criminal damage, cautions actually increased, though drug cautions fell as a result of access to Sure Start. This suggests that there were some young people for whom cautions became more common, perhaps aligning with other evidence of increased misbehaviour from data on suspensions from school.

Impacts on youth offending at different ages

Figure 5.5 shows that this shift is also linked to a change in the ages at which offences take place. It depicts the effect of living near a Sure Start centre on the percentage of children receiving either a caution or a conviction before each age. While there was no effect on the probability of receiving a caution by age 16, living near a Sure Start centre does seem to have increased the probability of receiving a caution at earlier ages.

Figure 5.5. Effect of living near a Sure Start centre on the percentage of young people with cautions and convictions received before each age

Fig 5.5

Note: Black error bars represent 95% confidence intervals. Approximate pre-Sure Start cumulative caution rates are 0.3% at 10, 0.9% at 11, 2.0% at 12, 3.6% at 13, 5.8% at 14, 7.8% at 15 and 9.4% at 16. Approximate pre-Sure Start cumulative conviction rates are <0.1% at 10, 0.1% at 11, 0.3% at 12, 0.8% at 13, 1.6% at 14, 2.6% at 15 and 3.7% at 16.

Source: ONS.

This reflects that for some young people, Sure Start shifted the age of a first offence to a younger age, reflecting poorer behaviour at younger ages. Because of Sure Start, fewer young people committed serious offences at ages 14, 15 or 16, but more young people started committing less serious offences at ages 11 and 12. This also suggests that there were some children for whom behaviour worsened – if the shift from convictions to cautions was just within the same group of children, we would not expect them to start interacting with the justice system earlier. Therefore, it appears that in addition to reducing serious criminal behaviour among some young people, there were others for whom Sure Start generated worse behaviour, though of less severity and at younger ages.

5.3 Children’s social care

We now turn to look at the impacts of access to a Sure Start centre on children’s contacts with the social care system during school. Figure 5.6 shows the effect on living close to a Sure Start centre on the likelihood of different types of interactions with the social care system: probability of any referral to social services, any CIN episode and any period as a CLA. The green bars show social care interactions during KS2 between ages 7 and 11; the yellow bars show interactions during secondary school between ages 12 and 16. These experiences range from one of the lighter touch interactions within the social care system (a referral) to the most intensive (removal into care).

Figure 5.6. Impact of Sure Start on referrals, CIN and CLA, and duration as CIN and CLA between ages 7 and 11, and between ages 12 and 16

Figure 5.6. Impact of Sure Start on referrals, CIN and CLA, and duration as CIN and CLA between ages 7 and 11, and between ages 12 and 16

Note: Outcomes between ages 7 and 11 use sample of cohorts of children born between 2002 and 2009 for referrals and CIN, and between 1999 and 2009 for CLA. Outcomes between ages 12 and 16 use sample of cohorts of children born between 1997 and 2005 for referrals and CIN, and between 1994 and 2005 for CLA. Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Duration as CIN and CLA is among CIN and CLA, respectively. Duration CLA refers to average duration of most recent period of care.

Source: ONS.

Across both the KS2 and secondary school samples, we find no significant impact of Sure Start on the probability of these social care contacts. For impacts on CIN and CLA, there is a relatively high degree of uncertainty around our estimates (as indicated by the wide confidence intervals), meaning that if the true effect is small, we would be underpowered to detect an effect.

The final two points in Figure 5.6 depict impacts of Sure Start on one proxy of intensity of social care intervention: average duration as a CIN and average duration as a CLA between ages 7 and 11 and ages 12 and 16. While there is no impact on the duration children spend as a CIN, there is a sizeable, statistically significant reduction in the time spent in care of 13% between ages 7 and 11, and a smaller reduction between ages 12 and 16 (5%).

Figures 5.7 and 5.8 provide supplementary evidence on the intensive margin: whether access to Sure Start had an impact on the needs for which children are referred or receive services as a CIN or CLA. Notwithstanding the limitations of these need categorisations detailed in Section 4.2, this offers some indication of whether Sure Start affected the primary reason for interacting with social care. Primary need on the horizontal axis is ordered according to the hierarchy used by social services, that is, for a child presenting with multiple needs, abuse or neglect is treated as highest priority, then child disability, and so on. Breaking down social care contacts by need, we exacerbate the issue around sample size and power and, overall, we find very limited evidence that Sure Start had an impact on the composition of CIN or CLA. This is except for Figure 5.8, which suggests an increase in referrals and CIN due to family in acute stress during adolescence (ages 12–16). This can be related to factors such as parental mental health crises or child misbehaviour that the family struggles to cope with. The latter could be consistent with the rises in child misbehaviour we see for suspensions and cautions; however, we cannot know this with confidence, and it is subject to the overall limitations of these need categories as outlined in Section 4.2.

Figure 5.7. Impact of Sure Start on referrals, CIN and CLA between ages 7 and 11, by need

Figure 5.7. Impact of Sure Start on referrals, CIN and CLA between ages 7 and 11, by need

Note: Uses sample of cohorts of children born between 2002 and 2009 for referrals and CIN, and between 1999 and 2009 for CLA. Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size.

Source: ONS.

Figure 5.8. Impact of Sure Start on referrals, CIN and CLA between ages 12 and 16, by need

Figure 5.8. Impact of Sure Start on referrals, CIN and CLA between ages 12 and 16, by need

Note: Uses sample of cohorts of children born between 1997 and 2005 for referrals and CIN, and between 1994 and 2005 for CLA. Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size.

Source: ONS.

Given that we find falls in the average duration spent looked after, in Figure 5.9 we break down the impact of Sure Start on duration in care by primary reason for CLA period of care for ages 7–11 (green bars) and 12–16 (yellow bars). At both sets of ages, reduction in CLA length is driven by the more common needs: at ages 7–11, this is abuse or neglect episodes, while at ages 12–15 reductions are driven by episodes due to family in acute stress and family dysfunction, which are more common at these older ages.

Figure 5.9. Impact of Sure Start on duration as CLA at ages 7–11 and 12–16, by need

Figure 5.9. Impact of Sure Start on duration as CLA at ages 7–11 and 12–16, by need

Note: Outcomes for ages 7–11 use sample of cohorts of children born between 2002 and 2009 for referrals and CIN, and between 1999 and 2009 for CLA. Outcomes for ages 12–16 use sample of cohorts of children born between 1997 and 2005 for referrals and CIN, and between 1994 and 2005 for CLA. Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Duration as CLA is among CLA and refers to average duration of most recent period of care.

Source: ONS.

Taken together, we find little robust evidence that access to Sure Start had a significant impact on the probability of a referral to social services, receiving support from social services as a CIN or being in the care of the local authority (CLA) when children are in KS2 and secondary school. We also do not find any changes to the types of needs children presented with during school age as a result of Sure Start. Yet, across both age groups, we see reductions in the average duration children spend in care, suggesting that Sure Start did not divert children away from social services entirely, but may have reduced the need for longer durations in care. This might be expected given that Sure Start was not an intervention designed to prevent interventions by social services, rather it was aimed at supporting families and improving the family environment.

Box 5.1. Sure Start Local Programmes versus Children’s Centres

For the majority of our analysis, data availability means that we are unable to test for differential impacts between SSLPs and SSCCs, which previous work suggests delivered different benefits. In particular, the early phase of Sure Start (i.e. SSLPs) has been shown to be more effective for improving educational outcomes than SSCCs (Carneiro et al., 2024a). 

To study the impact of Sure Start on CLA between ages 7 and 11, however, we rely on a set of cohorts – those born between 1999 and 2009 – who would have been exposed to both SSLPs and SSCCs, allowing us to break down effects by exposure to SSLPs compared with SSCCs. Although not statistically significant, we find children exposed to SSCCs experienced a drop in likelihood of being a CLA (around 12%), in comparison to very little impact on children who had access to SSLPs. Meanwhile, the reduction in duration in care appears to be driven by SSLPs (14% reduction) as opposed to SSCCs. We are cautious to over-interpret given the wide band of uncertainty around these estimates; however, this could indicate that SSLPs, which provided more intensive and more targeted provision, did not divert children in higher-needs families away from becoming looked after but may have helped reduce the severity of their needs, resulting in less time spent in care. Meanwhile, SSCCs, which delivered less specialised services for a more general population, may have offered early detection opportunities that reduced the likelihood of being in care at later ages; this would be consistent with Anderberg and Olympiou (2023), and is discussed in greater detail in Chapter 7.

 

6. Who experienced the biggest effect of Sure Start?

6.1 Effect by household income

A key goal of Sure Start was to mitigate the effect of poverty on children, and the rollout of centres focused first on areas with the highest levels of deprivation, as well as areas with high rates of low birthweights and teenage pregnancy (Department for Education and Employment, 1999). Therefore, it is important to understand whether Sure Start reduced disadvantage gaps across a range of outcomes.

Past work has already shown that Sure Start improved the academic outcomes of children from low-income backgrounds by much more than those from higher-income backgrounds (Carneiro et al., 2024a), and that it played a larger role in reducing hospitalisations in more-deprived areas than in less-deprived areas (Cattan et al., 2022).

Exposures to the criminal justice and social care systems, as well as suspensions and absences, are more common among those from low-income backgrounds. We analyse whether Sure Start had differential effects for children from low-income backgrounds by breaking down outcomes by FSM eligibility at age 5, a proxy for household income.15 Children from approximately the 15% lowest-income households are eligible for FSM.

In general, we do not find statistically significant differences in outcomes between those who were and were not eligible for FSM at age 5, across each of criminal offending, behaviour in school and children’s social care. Some of this is due to the smaller size of the population of young people who were eligible for FSM at age 5, leading to wider confidence intervals, so we cannot rule out differential effects by household income.

Figures A.5–A.8 of the Appendix show the effects on the outcomes considered in this report by household income. While the effects on absences and suspensions are not statistically significantly different across family incomes, there is some suggestion that the increase in both is driven by those not eligible for FSM. Earlier findings show that performance in school improved most among those who were eligible for FSM, while here we show that there might have been some less-disadvantaged children for whom Sure Start was less beneficial.  

It appears possible that young people from lower-income households saw smaller reductions in convictions and custodial sentences, and larger increases in cautions. While these differences are not statistically significant, it appears unlikely that the larger benefits for lower-income children found in academic outcomes also existed in criminal outcomes.

Meanwhile, estimates for the effect on social care use suggest that children from more-advantaged backgrounds may have experienced bigger reductions in duration in care between ages 7 and 11, though the differences are not statistically significant. Results for adolescents are similar across more- and less-disadvantaged young people.

6.2 By gender

Figures A.9 and A.10 of the Appendix show the differences by gender. For absences and suspensions, we largely see no significant differences in the effect of Sure Start by gender – the increases at age 16 seem to apply to both boys and girls.

Historically, youth offending rates are much higher for boys than for girls: by the age of 16, boys are twice as likely to have received a police caution, three times as likely to have received a conviction in court, and nine times as likely to have received a custodial sentence. In that context, it is perhaps not surprising that the effect of Sure Start on the total number of convictions and custodial sentences is larger for boys than girls, as it represents a reduction from a much higher baseline. In percentage terms, however, the difference in the effect of convictions and custodial sentences by gender is not statistically significant, though it does appear to be larger for girls.16

In contrast to offending and school suspensions, most types of social care activity do not differ significantly by gender. There is some evidence that, at older ages (ages 12–16), girls experienced increases in referrals to social care and in the likelihood of being a CIN due to exposure to Sure Start; however, these differences are not significantly different to results for boys.

6.3 By ethnicity

Similarly to gender, there are big differences by ethnicity in baseline levels of interaction with the criminal justice system. For those cohorts born just before the introduction of Sure Start, young people from Asian backgrounds (Chinese, Indian, Pakistani and Bangladeshi) were three times less likely to receive a police caution than White or Black adolescents before the age of 16. They were also half as likely to receive a court conviction as White young people, and a third as likely as Black young people. These gaps expand further for custodial sentences, with Asian students again half as likely as White students to have a custodial sentence by age 16, but five times less likely than Black students.

In percentage terms, there are few statistically significant differences in the effect of Sure Start by ethnicity, though it does appear that Asian young people experienced significantly larger reductions in convictions and custodial sentences than White young people (see Figure A.11 of the Appendix). This perhaps aligns with evidence that Sure Start had larger benefits for non-White students in other outcomes, such as grades in school (Carneiro et al., 2024a).

These differences by ethnicity do not appear in our analysis of absences and suspensions, where the differences between effect sizes across ethnicities are not statistically significant. However, due to small sample sizes, confidence intervals are much larger, so we again cannot rule out differential effects by ethnicity. For results on social care, as social care interactions are less common, low sample size means we are only able to split children by White and non-White ethnic backgrounds. We do not find significant differences in the effects of Sure Start on social care use between these two groups of children.

7. Discussion

We find that children with greater access to Sure Start experienced moderate increases school suspensions and small rises in missed school days in secondary school. In terms of more severe misbehaviour, Sure Start resulted in significantly fewer serious criminal convictions and far fewer custodial sentences, but did not seem to have any aggregate effect on less serious offences (cautions). Pooling offences of different severity together, we do not observe a statistically significant effect of Sure Start on criminal behaviour. The lack of change in cautions and fall in more severe offences may partially reflect a change in the composition of offences, shifting some children towards less severe forms of criminal misbehaviour. However, the fact that there was a shift towards these misdemeanours taking place at younger ages suggests that, for some children, there were increases in cautions, which came from children who would otherwise have not engaged in any offending behaviour. Table 7.1 summarises these results.

Table 7.1. Summary of effects of Sure Start on school attendance, youth offending and social care contacts

Outcomes

Effect of Sure Start

Ages

Birth cohorts

Sure Start phase

School attendance:

 

 

 

 

Suspensions

No significant effect at age 11

Moderate increase at age 16

11 and 16

1994–2007

SSLP and SSCC

Absences

Slight decline in authorised at age 11

Small increase in authorised and unauthorised at 16

11 and 16

1994–2007

SSLP and SSCC

Youth offending:

 

 

 

 

Cautions

Small rises in cautions for ages 10–14

Largely for violent and damage cautions

11–16

1994–2001

SSLP

Convictions and custodial sentences

Moderate to large falls in convictions and custodial sentences

Concentrated amongst theft, drug and weapon offences, and between ages of 14 and 16

11–16

1994–2001

SSLP

Social care contacts:

 

 

 

 

Referrals and CIN

No significant effect

7–11

2002–2009

SSCC

 

No significant effect

12–16

1997–2005

SSLP

CLA

Moderate fall in duration CLA

7–11

1999–2009

SSLP and SSCC

 

Small fall in duration CLA

12–16

1994–2005

SSLP

Note: This table reports effects that are statistically significant at the 95% level.

Reductions in youth offending would be consistent with Sure Start improving children’s socio-emotional development and reducing hospitalisations for mental health disorders during adolescence, as found in Cattan et al. (2022) and in Carneiro et al. (2024a), likely to be facilitated by improvements in parenting and the home environment.

At the same time, the rise in absences and suspensions suggests that, for some children, behaviour worsened, while shifts in cautions suggest that this misbehaviour may have extended beyond the school environment. The fact that Sure Start affected behaviour differently for different groups of children would be consistent with previous evaluations of the programme, which found worsening of behaviour amongst some children, and improvements in behaviour for others (such as those who were and were not raised by teenage mothers; Belsky and Melhuish, 2007). The wider literature also provides evidence that group-based childcare, which Sure Start provided to some extent, can improve behaviour for some children but can result in elevated behavioural issues in others (e.g. Baker et al., 2008; Fort et al., 2016). Another candidate explanation is that the effects run through improving links between young people, whereby children become more sociable and strengthen peer relationships, in turn engaging in more low-level, collaborative misbehaviour such as fighting or graffiti, which results in, for instance, higher police cautions for violence or criminal damage.

On the one hand, we find that exposure to Sure Start did not affect the likelihood of coming into contact with children’s social care, or the level of social care intervention experienced between the ages of 7–11 and 12–16. On the other hand, it appears that Sure Start reduced the intensity of social care contact at the more severe end, by reducing the average length of time in care across for children both in KS2 and during secondary school. While there are fewer impacts on these outcomes than others, as we discuss in Chapter 8, even small impacts on social care usage are important for the public purse, due to the high cost of provision of children’s social care services.

The fact that Sure Start had a limited impact on the likelihood of encountering social services during KS2 and secondary school may be expected given the design of the programme. As a relatively light-touch, area-based intervention primarily focused on low-income families, Sure Start might have had less scope to change the circumstances of families with deeper and more complex needs.17 Indeed, there is mixed evidence that even more intensive early years programmes (such as the Family Nurse Partnership) can shift patterns of children’s social care use: for instance, Robling et al. (2021) find that the Family Nurse Partnership in the UK did not reduce child maltreatment, whereas Kliem and Sandner (2021) find that a similar programme in Germany, ProKind, reduced abusive parenting practices. Interestingly, the Robling et al. (2021) evaluation of the Family Nurse Partnership did find that the programme reduced the amount of time children spent in care, mirroring our results.

Alternatively, it could be that Sure Start both increased detection and decreased need for social services support, and that these patterns offset each other at the ages we study. This is plausible as studies estimating prevalence of child maltreatment suggest that underlying need for social services is higher than their reach (e.g. Gilbert et al., 2009).

A final insight into potential mechanisms comes from a related study, Anderberg and Olympiou (2023), who find that local authorities with greater coverage of SSCCs experienced elevated rates of entry into care (CLA) amongst children who were eligible to attend the centres and (smaller) reductions in flows into care at older ages. The study is described in greater detail in Box 7.1. This pattern of findings suggests that Sure Start led to increased identification of children at risk of harm (rise in care entry for under-fives) and, in addition, may have brought forward in time social care interactions that would have happened at later ages (falls in care entry at ages 7–9). Our results do not find a significant change in the stock of children in care but do imply falls in time spent looked after. This would be consistent with Sure Start changing the timing of contacts with children’s social care and with early detection resulting in children spending less time in care, but not shifting the likelihood of interacting with social services at later ages.

Box 7.1. Impact of Sure Start Children’s Centres on local authority rates of entry into care

A related study (Anderberg and Olympiou, 2023) looks at the impact Sure Start had on rates of entry into care and provides a valuable complement to this work. The authors leverage variation in the expansion of Sure Start, as well as variation in centre closures, to study the impact of local Sure Start coverage on local authority-level rates of entry into care (CLA). Their focus is on the impact of SSCCs between 2004 and 2017. This differs from our study as they do not look at the impact of SSLPs and include the period of centre closures from 2010 onwards. Crucially, by using area-level data, the authors are able to study impacts for children under the age of 5 (i.e. during the time when children would have attended Sure Start centres).

The study finds that areas with greater access to SSCCs experienced elevated rates of entry into care (CLA) for children aged 4 and under. Beyond the age that children attended centres, there is suggestive evidence that areas exposed to Sure Start experienced drops in entry to care at ages 7–9. Translating this into a cumulative effect, the authors find that due to SSCCs, children were about 20% more likely to have entered care by age 5, with most of this effect driven by entry amongst infants (under the age of 1).

Our most comparable results – those using a similar set of cohorts affected by SSCCs, as detailed in Box 5.1 – are consistent with this: although insignificant, we find Sure Start is associated with a fall in the stockof children in care between ages 7 and 11.

 

Finally, it is interesting to consider for whom Sure Start was beneficial, compared with which children may have experienced unintended consequences. It is less plausible that the children who saw the strong improvements in academic outcomes identified in previous evaluations were the same children who missed more days of school or committed more low-level offences, suggesting that Sure Start had different effects on different groups of children. This would be consistent with other evaluations of early years childcare that find different impacts for children from different backgrounds (e.g. Havnes and Mogstad, 2011; Baker, Gruber and Milligan, 2019). These studies often find positive impacts for children at the bottom of the income distribution and negative effects for children from more advantaged backgrounds; however, given that we do not find heterogeneous treatment effects along traditional socio-economic or demographic lines, such heterogeneity may be more complex to characterise than we can measure in the administrative data. This is an area for future research.

8. Cost–benefit analysis

Most of the outcomes we consider in this report entail significant costs, both to the government and to society more widely. In this chapter, we outline how the effect of Sure Start on contacts with the criminal justice system and children’s social care affected those costs. Included in these calculations are the government savings from reductions in use of the youth justice system, savings made by potential victims from the prevention of crime, and government savings from reductions in the use of the children’s social care system. We do not account for possible long-run earning changes from the effect of Sure Start on absences and suspensions from school, to avoiding double-counting returns captured within academic performance, the benefits of which are outlined in Carneiro et al. (2024a). For the same reason, we do not include long-run earnings increases that might be expected from reduced contact with the criminal justice or children’s social care system. Given that we focus on children’s social care usage and do not observe child maltreatment directly, we also cannot quantify the extent to which child maltreatment is averted, and therefore cannot realise any potential savings beyond social care costs, such as health-care costs or labour market productivity costs, that arise from child maltreatment (Conti et al., 2021).18 Box 8.1 details further assumptions we make in the cost–benefit analysis.

Box 8.1. Assumptions underpinning cost–benefit analysis

The effect of living near a Sure Start centre is not the same as the effect of using Sure Start. We should expect proximity to a centre to increase the likelihood of attendance, but also that some who lived nearby would not necessarily attend or benefit, and that some who lived further away would attend and benefit. Therefore, we follow Carneiro et al. (2024a) in translating the effect of living nearby to the effect of attendance using data from the ECCE reports (Goff et al., 2013; Maisey et al., 2013; Smith et al., 2014).

As a starting point, we assume that the effect of Sure Start on children’s social care and crime affects only those who actually used Sure Start services and does not spill over to other children in the area. To do this, we use estimates from Carneiro et al. (2024a) on how the effect of living near a centre translates to the effect of actually using the centre. In their central estimate, they find that the effect of using a Sure Start centre is 2.9 times greater than the effect of living near it. For example, if living near a centre reduced the probability of interacting with the social care system by 1%, we would infer that actually attending the centre reduced the probability by 2.9%.

 

8.1 Costs of Sure Start

We estimate the costs of Sure Start both as a total, and on a per-person basis of the total usership of Sure Start. To do so, we require information on the total government expenditure on the Sure Start programme (publicly available) and information on usership, which was not systematically collected. As a final step, we also need to identify usership for children living within 2.5 km of a centre between ages 0 and 4 to align with our measure of treatment.

To estimate usership, we use data from the ECCE reports, which include information on the total number of users of the average centre, as well as information on where those users lived. Using data on where children lived at a young age, we estimate that by the time the rollout of Sure Start ended in 2010, around 82% of children born in the last four years lived within 2.5 km of a centre, and thus would be counted as living nearby in this report. Of these, we estimate that around 87% of them lived near a Sure Start centre for their whole childhood, the treatment measure we use. This means that, on average, for each of the four years prior to 2010, around 385,000 children born each year lived within 2.5 km of a centre for their whole childhood.

Meanwhile, estimates from Carneiro et al. (2024a) using the ECCE reports suggest that around 380,000 children in each birth cohort at this time actually used a Sure Start centre. It is worth noting that these are not necessarily the same children: many children who lived near a centre did not actually use it, and many who lived further away from a centre did use Sure Start services.

By 2010, as Sure Start expanded further, the total cost had risen to £2.52 billion (in 2023 prices). This means that by the peak of Sure Start in 2010, the average spend across all users of Sure Start was approximately £1,300.

8.2 Savings to government

In estimating the impact of Sure Start on savings (costs averted) to the government, we first consider the crime channel. Evidence from the National Audit Office (2011) gives a best estimate of the cost to the public purse of a range of different crimes, depending on criminal justice outcomes. They estimate that the average cost of processing a criminal conviction of a young person included – in 2023 prices – £650 of policing costs, between £1,800 and £3,500 of costs per court appearance, depending on the type of crime, £1,900 per offender per year for the costs of offender management teams, and £6,400 per month for custodial sentences.19 Youth cautions are not reported as having a significant cost, so we do not include them in the cost–benefit analysis. Due to a lack of evidence on the effect of youth crime on later crime, we also do not estimate the effect on the cost of crime at later ages, though we might expect that to be an additional saving from the policy.

Combining these costs with the effect sizes in this report for the impact of Sure Start on the number of offenders, the number of convictions broken down by type and the number of custodial sentences, we estimate that for each individual who attended a Sure Start centre, the Ministry of Justice saved £677 through the justice system.20 Of this, £509 comes from the reduction in the number of custodial sentences, by far the single biggest saving to the criminal justice system from the effects of Sure Start.

Scaling these effects up to the total population in each birth cohort who lived near a Sure Start centre gives total approximate savings to the government of £257 million by 2010, as Sure Start expanded to more areas.

While the processes involved with responding to crime are costly, crime also entails wider costs to society. To capture these social costs averted, we use data from Heeks et al. (2018), which provides estimates of the economic and social cost of crime. This report estimates the effect of anticipation, property lost, emotional harm and lost output in response to crimes that have a specific victim. Of these crimes, the only category on which Sure Start had significant effects is theft, the social cost of which ranges from £800 for theft from a vehicle to £7,700 for robbery. Using this, we estimate that for each individual who lived near a Sure Start centre throughout childhood, all other individuals received £18 of benefit in the form of reduced thefts. Across all individuals who used a centre, this translates to £7 million of additional returns for those attending a centre in 2004, rising to £20 million by 2010.

For children’s social care, the key financial cost to the government is the cost of children looked after. In 2016 (the median year that children in our analysis sample experienced the social care system), the government spent £4.9 billion on children looked after (in 2023 prices). On average, 69,500 children were in care each day, meaning a cost of each day in care of £194. On average, this report finds that living near a Sure Start centre reduced the average duration in care by 113 days for children aged 7–11, and by 41 days for children aged 12–16. Adjusting for those who actually used Sure Start services, this means that for a birth cohort attending Sure Start at its peak in 2010, we should expect total savings of £234 million due to the effect of Sure Start on children’s social care usage.

8.3 Balancing costs and benefits

The total cost of Sure Start reached £2.52 billion (in 2022–23 prices) by its peak in 2010. Table 8.1 outlines the total averted costs, both to the government and to members of the public. Combining our results with 2010 attendance patterns, we should expect that about 19% of the cost of Sure Start was saved by the government through reductions in youth offending and time in social care. These savings are close to evenly split between savings to the justice system and to the social care system.

Table 8.1. Estimated costs and benefits of Sure Start per cohort in 2010

 

Costs

Total upfront cost

£2,520 million

Costs averted – justice system

£257 million

Costs averted – social care system

£234 million

Total costs net of averted

£2,029 million

Benefits – reduced costs to victims

£20 million

Note: 2022–23 prices.

Table 8.1 also shows that the effects outlined in this report did not bring large benefits to other members of society. The only direct benefit, through reducing the costs of crime experienced by victims, only brought around £20 million of benefits by 2010, less than 1% of the total upfront cost of the programme.

The translation of the effects of living nearby into a cost–benefit analysis rely on data about proximity to centres and take-up from 2010 and 2011 respectively, as the best data on usership we have come from the ECCE reports. Therefore, translating these effects to other years relies on the assumption that effects were consistent over time.

Evidence from Carneiro et al. (2024a) suggests that this was not always the case: centres that opened earlier, as SSLPs, had larger effects on educational outcomes than those that opened later as SSCCs, though earlier opening centres continued to have positive effects going forward, so the overall effect of Sure Start was still positive in all years.

In this report, the largest effects are spread unevenly across early and later centres – the crime results come from earlier centres, while the biggest effects on social care come from children who attended Sure Start closer to 2010. This gives reason to believe the effects on the outcomes we consider here may be more evenly spread, but we do not have the data to say for certain. Therefore, while this is our best estimate of the overall financial benefits of Sure Start via the outcomes we consider, it is a very rough estimate of the overall effect in 2010.

The savings and financial benefits in this report are not the only ones that stem from Sure Start. Cattan et al. (2022) outline the long-run savings to the government from improved health outcomes for children who used Sure Start. Carneiro et al. (2024a) find savings from a reduced need for education, health and care plans among those who used Sure Start, as well as large expected increases in earnings arising from improved performance in GCSEs. Across each report, different methodologies and periods of data have been used to estimate the costs and benefits of the policy, and so they are not directly comparable, and cannot simply be added to each other. Future work will lay out possible estimates of the total long-run costs and benefits of Sure Start, taking all of these factors into account.

9. Conclusion

The rollout of Sure Start centres across England from 1999 to 2010 targeted improvements in a wide range of outcomes for children in the short run, including school preparedness and health outcomes. Not only was Sure Start effective at achieving these aims, but, as prior studies have shown, it also delivered improvements to children’s socio-emotional development and the home environment. At the same time, the persistent effects of Sure Start on health and academic outcomes throughout childhood suggest that benefits stemmed not just from the support children received at Sure Start centres, but also from how additional support for the whole household made a difference in the long run.

Given these multi-dimensions of impact, it is natural to consider how Sure Start affected children’s behaviour at school, outcomes in the criminal justice sphere, and their contacts with children’s social care. Changes in socio-emotional development could translate into changes in children’s behaviour, as might changes to the home environment, while centre-based provision may have had an impact on children’s peer interactions and provided staff with opportunities to refer children to children’s social care.

For youth offending, Sure Start had different impacts for offences of different severity. Sure Start significantly reduced serious criminal behaviour, with children exposed to Sure Start receiving 20% fewer custodial sentences by age 16 than similar children who were not exposed. Yet police cautions did not change for young people who lived near centres as children, and even increased for violence and criminal damage. These rises in less severe forms of misbehaviour also extended in schools, with secondary school pupils more likely to miss days of school, and more likely to be suspended from school.

Sure Start’s effect on the children’s social care system was not as large as that on the criminal justice system, with no significant effects of Sure Start on the probability of referral to social care services, a CIN episode, or a period as a CLA. At the same time, it appears that Sure Start reduced the intensity of social care contact at the more severe end, by reducing the average length of time in care. Less time spent in care, as well as having likely impacts on children and families, is also significant from a fiscal perspective as the annual financial cost of CLA is very high.

Overall, it seems that Sure Start was significant in shaping the behaviour of those who lived near a centre – but with different effects for different children. The upshot of this was a reduction in more severe criminal behaviour but increases in less severe bad behaviour for some children, both within and outside school. Sure Start’s effect on children’s social care is perhaps less surprising: one might expect that the task of diverting a household away from contact with the social care system requires a more intensive approach than Sure Start offered, but improvements in family functioning may have contributed to children spending less time in care.

These effects of Sure Start have a significant impact on the overall cost of Sure Start. Across savings in the justice and children’s social care systems, we should expect almost one-fifth of the total upfront cost of the policy to have been recouped. Much of this is due to the very high cost to the government of youth custody, and of CLA. Reducing the number of young people in custody, and reducing the amount of time young people spend in care, has significant benefits to the Exchequer.  

This report builds on past work showing that Sure Start generated significant reductions in child hospitalisations and improvements in academic performance, by demonstrating that the benefits extended to reduction in serious crimes. These effects stretched out over a long time frame, highlighting the potential of early intervention to produce long-lasting benefits. However, it is important to weigh up Sure Start’s benefits against potential unintended consequences. Large-scale interventions may improve some outcomes, for some groups of children, but do not preclude negative effects. Meanwhile, fewer significant impacts on children’s social care show that while early interventions can deliver a lot, they are not a panacea; higher-intensity interventions are likely necessary to help families facing the most severe difficulties.

Appendix

A.1    Difference-in-differences specification

The difference-in-differences method estimates the effect of a Sure Start centre opening (in our case, the treatment) by calculating the difference between the change in outcomes over time in areas where Sure Start centres opened and the change in outcomes over time in areas where no centres opened.21 This relies on the assumption that there is no other reason why the outcomes in areas with Sure Start centres might have changed in a different way from those in other areas.

The most common reason why this might not be true is if the pre-treatment trends in areas with and without centre openings are not parallel. For example, if youth conviction rates were falling quickly in areas where Sure Start centres opened even before the centres opened, but were not changing in areas where Sure Start centres did not open, then our difference-in-differences method would pick up that difference as part of the effect of Sure Start. This could plausibly be happening, as Sure Start’s rollout was specifically targeted to areas with higher deprivation, higher rates of teenage pregnancy and higher levels of low birthweight. Areas with these features might have been changing differently from other areas, perhaps because they were catching up with areas with better outcomes.

To account for this, we control for deprivation of each LSOA, to ensure we are not capturing the effects driven by deprivation. Additionally, we calculate the pre-treatment trends in youth crime rates by Middle Statistical Output Area (MSOA), a larger geographical area, and control for these trends. This gives more emphasis to the comparison in the differences for those treated areas that have more similar trends to untreated areas, and vice versa. This resembles the synthetic difference-in-differences method proposed by Arkhangelsky et al. (2021), but using the method for aggregating staggered difference-in-differences proposed in Callaway and Sant’Anna (2021). With the latter estimator, controls are implemented using propensity score matching of pre-treatment values only.

Figure A.1 shows that with these controls, there are parallel pre-treatment trends between treated and untreated LSOAs in conviction rates. The chart shows the difference between those with and without Sure Start centres in the difference in trends between year t 1 and each year. The fact that there is no trend in the values before year 0 (when the Sure Start centre opened) indicates that conviction rates for those too old for Sure Start were not changing in different ways in the two types of area, once adjusting for area characteristics. Figure A.2 shows that the same is true for the effect on custodial sentences age 16. These parallel trends hold for all our headline results.

Figure A.1. Effect of Sure Start on probability of having had a conviction by age 16, by treatment length

Figure A.1. Effect of Sure Start on probability of having had a conviction by age 16, by treatment length

Note: Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Children born after a Sure Start centre opened near them are displayed as having five or six years of Sure Start access, even though some of these years are before their birth.

Source: ONS. Authors’ calculations using data from the NPD.

Figure A.2. Effect of Sure Start on probability of having had a custodial sentence by age 16, by treatment length

Figure A.2. Effect of Sure Start on probability of having had a custodial sentence by age 16, by treatment length

Note: Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Children born after a Sure Start centre opened near them are displayed as having five or six years of Sure Start access, even though some of these years are before their birth.

Source: ONS. Authors’ calculations using data from the NPD.

As in Carneiro et al. (2024a), these event studies demonstrate interesting dynamics with regards to the effect of a Sure Start centre opening on children who were aged between 0 and 4 when it opened. In both charts, the estimate at ‘0’ represents the effect on those who were aged 4 when a Sure Start centre first opened near them. Meanwhile, the estimate at ‘4’ represents the effect on those who were born the year the centre opened, and years beyond that represent the effect on those who were born in the years after a centre opened. Therefore, those who are fully treated are those from ‘4’ onwards, while those who are partially treated by Sure Start are those who were aged between 1 and 4 when a Sure Start centre opened, and so had between zero and three years of Sure Start access on the chart.

Figures A.1 and A.2 show that the effects were much smaller for partially treated children than for fully treated children. Being partially exposed to Sure Start only reduced conviction rates by 2%, while being fully exposed reduced conviction rates by 12%. The drop-off is similar for custodial sentences, where partial exposure reduced custodial sentencing rates by 4% (though this effect is not statistically significant), while full exposure reduced sentencing rates by 24%.

A.2    Children’s social care data

Cleaning the CIN and CLA censuses

The CIN census is derived from the administrative records of local authority social services. It is an episode-based dataset, where each episode represents an open case following a referral to social services. In theory, each child should only have one episode open at a time, but as existing cases are updated with new information, often a new record is generated. This can lead to duplicate records, records with missing information (especially closure dates) and episodes that overlap. Generally, the most recent record of an episode is most likely to be complete and up to date (Emmott, Jay and Woodman, 2019). The preparation of these data involves dropping duplicate records, consolidating information from overlapping episodes and imputing missing closure dates. For instance, many episodes in the final year of data 2021/2022 have missing closure dates because the episode is ongoing. Episodes with missing start dates or with implausible episode dates (a very small minority) are dropped.

The CLA census is generally more complete but comes with the limitation that only the most recent period of care within a recording year is reported. This will miss care episodes for children who cycle in and out of being CLA within a reporting year. This would not have an impact on estimates of Sure Start’s effect on average duration under the assumption that Sure Start did not have a differential effect on shorter spells occurring earlier in the recording year.

A final point to note is that the first year of the CIN census (2008–09) covers a six-month period, and early years of data collection had missing returns from some local authorities.

Sample selection for children’s social care

The full CLA and CIN censuses are available from the mid-2000s onwards (2006 and 2009, respectively). This means we only pick up earlier cohorts of children born in the 1990s at later ages. For example, children born in 1994 appear in the CIN census between ages 14 and 16. Meanwhile, later-born cohorts are not yet old enough to observe up until age 16: those born in 2008, for instance, are observed from age 5 until age 14. Studying the impact of Sure Start on social care contacts at different ages therefore draws on different sets of birth cohorts: later-born cohorts to study impacts on social care at younger ages and earlier-born cohorts for social care contacts at older ages.

In addition, at a given point in time, it is uncommon for children to interact with social services (although referrals are more common), making it challenging to detect small impacts of Sure Start on social care contacts. To improve our ability to detect effects, we therefore pool contacts with social care across groups of ages and use all cohorts available at the relevant ages to maximise sample size.

This results in two samples for analysis. We study social care contacts during KS2 (ages 7–11) using children born from the early 2000s to 2009 and we study contacts with social care during secondary school (ages 12–16) using children born between the mid-1990s and 2005. The exact cohorts used differ slightly between whether we are studying outcomes from the CIN census (referrals and CIN status) or from the CLA census (children in care). The combinations of ages and cohorts studied using the CIN census are shown in Figure A.3 in Appendix A.4; analysis using the CLA census uses additional cohorts as it is longer running than the CIN census.

The characteristics of these two samples differ due to the different cohorts used but also due to inclusion of different areas (LSOAs). Due to the staggered nature of Sure Start rollout across areas and birth cohorts (and the method we use to account for this), the analysis samples only include areas (LSOAs) whose earliest-born cohorts have not yet had access to Sure Start. Because the KS2 sample relies on children born during the later stages of Sure Start’s rollout, it excludes areas that had access to Sure Start earlier, which tended to be more disadvantaged. These differences in sample are shown in Table A.1 to aid interpretation of results. As discussed in the main body of the report, another key difference between these samples is exposure to SSLPs versus SSCCs.

Table A.1. Analysis samples and characteristics for social care outcomes

 

Social care contacts measured at:

KS2 
(ages 7–11)

Secondary school 
(ages 12–16)

Birth cohorts

2002 to 2009

1994 to 2005

N LSOAs

14,954

32,199

Share FSM

0.06

0.15

Share white

0.9

0.83

Share English additional language

0.07

0.11

Share 30% most deprived areas

0.08

0.3

Share middle 40% areas

0.43

0.4

Share 30% most affluent areas

0.49

0.3

Average population density

0.87

1.91

Average childcare coverage

0.82

0.33

Note: Area deprivation based on Index of Multiple Deprivation. Childcare coverage proxied by local area take-up rates for free-entitlement amongst 3 year olds.

Source: ONS.

A.3    ‘First stage’ estimates and the cost–benefit analysis

To compute the effect of living near a Sure Start centre on attending one, we collect data from two sources. The first of these is the Evaluation of Children’s Centres in England (ECCE), and the second is the Millennial Cohort Study (MCS). ECCE provides data from a survey of centres’ registration data conducted in 2011 and 2012 (Goff et al., 2013; Maisey et al., 2013; Smith et al., 2014), while the MCS provides data from a survey of parents in 2004, including questions on usership of Sure Start.

When using ECCE to approximate the effect of living near a Sure Start centre on usership, we collect information on the number of children registered at the average centre (Smith et al., 2014), survey answers on what proportion of those registered ever used Sure Start (Maisey et al., 2013), and how far registered users lived from the Sure Start centre (Goff et al., 2013). From the NPD, we also use data on the number of children in each cohort who did and did not live within 2.5 km of a Sure Start centre.

Using these sources, we first compute the proportion of children aged 04 in England who used Sure Start services. Data from surveys of all Sure Start centres show that the average centre had 770 children registered at it in 2012. Multiplying across the number of centres, this means that over 2.5 million children aged 04, or about 90% of this age group, were registered at a Sure Start centre. However, in many cases, children and families were automatically registered, such as when registration was closely linked to universal health visitor programmes, and in some cases they would never have had any contact with the Sure Start centre. The three ECCE reports referenced in the previous paragraph provide two different interpretations of usage: the first comes from data from centres, which report that 55% of the 04 age group had used the centre in the last year. This may be an underestimate of total usage, as some may have used it more than a year ago, perhaps to benefit from the services targeted at newborn children, and then not used it again. Meanwhile, the survey of registered users reports that only 11% of them had never used the centre, implying that approximately 89% of the registered population and 80% of the total age 0–4 population used a Sure Start centre at some point. This should be thought of as an upper bound on total usage, as it is based on a survey of registered users that did not have a 100% response rate, and we expect those who never used the centre to have been much less likely to respond to the survey. Therefore, our best estimate is that the proportion of children aged 04 using Sure Start in 2012 was between 55% and 80%.

Using these bounds, data from the NPD on the proportion of children who lived within 2.5 km of a Sure Start centre and the proportion of Sure Start users who lived within 2.5 km – which Goff et al. (2013) estimate to be approximately 90% – we now estimate the impact of living within 2.5 km (as opposed to further away) on the probability of using Sure Start services. To do this, we use data on where children lived, from the last cohort we have detailed data on (those born in 2006, whose exposure to Sure Start would have ended in 2010). There were about 563,000 children born in 2006 who appear in our KS2 sample. From our lower and upper bounds, we estimate that between 310,000 and 450,000 of them attended a Sure Start centre. If 90% of attendees lived within 2.5 km of a centre, this means that, in the upper-bound case, around 405,000 children lived near a centre and attended, and about 45,000 did not live nearby but attended anyway. Of the 563,000 children born in 2006 who appear in our KS2 sample, around 465,000 lived near a centre and around 98,000 did not (based on our calculations in the NPD). This implies that about 87% of those who lived nearby attended and that about 46% of those who did not live nearby attended. This, in turn, implies that living near a Sure Start centre made children 41 percentage points more likely to attend. In the lower-bound case, the equivalent figure is 28 percentage points.

A.4    Supplementary figures

Figure A.3. Sample selection for social care outcomes

Note: Grey areas indicate the age at which we observe cohorts born in each year in the CIN census. The red box indicates the set of cohorts we use to study referrals and CIN between ages 7 and 11; the cohorts used to study referrals and CIN between ages 12 and 16 are indicated in the green box.

Figure A.4. Effect of living near a Sure Start centre on the number of sentences of each type per young person up to age 16

Figure A.4. Effect of living near a Sure Start centre on the number of sentences of each type per young person up to age 16

Note: Black error bars represent 95% confidence intervals. Pre-Sure Start levels for offences per young person were 0.32. There were 0.14 cautions per young person at baseline, 0.18 convictions and 0.02 custodial sentences. Results reflect percentage change in the number of offences per child due to Sure Start. Custodial sentences are also included within convictions. Those receiving no sentences are included.

Source: ONS.

Figure A.5. Effect of living near a Sure Start centre on absence and suspension rates at age 16, by FSM eligibility

Figure A.5. Effect of living near a Sure Start centre on absence and suspension rates at age 16, by FSM eligibility

Note: Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Prior to Sure Start, the percentage of school missed for any reason at 16 was 11% for children who were eligible for FSM at age 5, and 6% for non-eligible children. Baseline rates of suspension or suspension from school at age 16 for FSM-eligible children were 10%, but were 4% for those not eligible for FSM at age 5.

Source: ONS.

Figure A.6. Effect of living near a Sure Start centre on the probability of each offence up to age 16 by FSM eligibility

Figure A.6. Effect of living near a Sure Start centre on the probability of each offence up to age 16 by FSM eligibility

Note: Black error bars represent 95% confidence intervals. Prior to Sure Start, caution rates were 16% for FSM-eligible young people and 7% for non-eligible young people. Conviction rates were 8% for those eligible and 2% for those ineligible. Custody rates were 1.1% for eligible children and 0.3% for non-eligible children. Results reflect percentage change in the number of children with each type of offence. Custodial sentences are also included within convictions.

Source: ONS.

Figure A.7. Effect of living near a Sure Start centre on contacts with children’s social care between ages 7 and 11, by FSM eligibility

Figure A.7. Effect of living near a Sure Start centre on contacts with children’s social care between ages 7 and 11, by FSM eligibility

Notes: Outcomes between ages 7 and 11 use sample of cohorts of children born between 2002 and 2009 for referrals and CIN, and between 1999 and 2009 for CLA. Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Duration as CIN and CLA is among CIN and CLA, respectively. Duration CLA refers to average duration of most recent period of care.

Source: ONS.

Figure A.8. Effect of living near a Sure Start centre on contacts with children’s social care between ages 12 and 16, by FSM eligibility

Figure A.8. Effect of living near a Sure Start centre on contacts with children’s social care between ages 12 and 16, by FSM eligibility

Note: Outcomes between ages 12 and 16 use sample of cohorts of children born between 1997 and 2005 for referrals and CIN, and between 1994 and 2005 for CLA. Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Duration as CIN and CLA is among CIN and CLA, respectively. Duration CLA refers to average duration of most recent period of care.

Source: ONS.

Figure A.9. Effect of living near a Sure Start centre on the number of convictions for each type of crime per young person up to age 16, by gender

Figure A.9. Effect of living near a Sure Start centre on the number of convictions for each type of crime per young person up to age 16, by gender

Note: Black error bars represent 95% confidence intervals. Pre-Sure Start levels for caution rates were 6% for girls and 12% for boys. Conviction rates were 2% for girls and 6% for boys, while custody rates were 0.1% for girls and 0.9% for boys. Results reflect percentage change in the number of children with each type of offence. Custodial sentences are also included within convictions.

Source: ONS.

Figure A.10. Effect of living near a Sure Start centre on social care contacts between ages 12 and 16, by gender

Figure A.10. Effect of living near a Sure Start centre on social care contacts between ages 12 and 16, by gender

Note: Outcomes between ages 12 and 16 use sample of cohorts of children born between 1997 and 2005 for referrals and CIN, and between 1994 and 2005 for CLA. Black error bars represent 95% confidence intervals. Effects are expressed relative to their baseline means as a proportionate effect size. Duration as CIN and CLA is among CIN and CLA, respectively. Duration CLA refers to average duration of most recent period of care.

Source: ONS.

Figure A.11. Effect of living near a Sure Start centre on the number of convictions for each type of crime per young person up to age 16, by ethnicity

Figure A.11. Effect of living near a Sure Start centre on the number of convictions for each type of crime per young person up to age 16, by ethnicity

Note: Black error bars represent 95% confidence intervals. Pre-Sure Start levels for caution rates were 10% for White young people, 4% for Asian young people and 9% for Black young people. Conviction rates were 4% for White young people, 2% for Asian young people and 6% for Black young people. Custody rates were 0.5% for White young people, 0.3% for Asian young people and 1.1% for Black young people. Results reflect percentage change in the number of children with each type of offence. Custodial sentences are also included within convictions.

Source: ONS.

References

Anderberg, D. and Olympiou, C., 2023. Children’s social care and early intervention policy: evidence from Sure Start. Economica, 90 (359), 953–77, https://doi.org/10.1111/ecca.12473

Anders, J., Barr, A. C. and Smith, A. A., 2023. The effect of early childhood education on adult criminality: evidence from the 1960s through 1990s. American Economic Journal: Economic Policy, 15 (1), 37–69, https://doi.org/10.1257/pol.20200660

Arkhangelsky, D., Athey, S., Hirshberg, D. A., Imbens, G. W. and Wager, S., 2021. Synthetic difference-in-differences. American Economic Review, 111 (12), 4088–118, https://doi.org/10.1257/aer.20190159

Aucejo, E. M. and Romano, T. F., 2016. Assessing the effect of school days and absences on test score performance. Economics of Education Review, 55(C), 70–87, https://doi.org/10.1016/j.econedurev.2016.08.007

Baker, M., Gruber, J. and Milligan, K., 2008. Universal child care, maternal labor supply, and family well-being. Journal of Political Economy, 116, 709–45, https://doi.org/10.1086/591908

Baker, M., Gruber, J. and Milligan, K., 2019. The long-run impacts of a universal child care program. American Economic Journal: Economic Policy, 11 (3), 1–26, https://doi.org/10.1257/pol.20170603

Belsky, J. and Melhuish, E., 2007. Chapter Eight: Impact of Sure Start Local Programmes on children and families. In Belsky, J., Barnes, J. and Melhuish, E. (eds), The National Evaluation of Sure Start: Does Area-Based Early Intervention Work?, 1st edn, Bristol University Press, 133–54, https://doi.org/10.51952/9781847423009

Blattman, C., Jamison, J. C. and Sheridan, M., 2017. Reducing crime and violence: experimental evidence from cognitive behavioral therapy in Liberia. American Economic Review, 107 (4), 1165–206, https://www.aeaweb.org/articles?id=10.1257/aer.20150503

Callaway, B. and Sant’Anna, P. H. C., 2021. Difference-in-differences with multiple time periods. Journal of Econometrics, 225, 200–30, https://doi.org/10.1016/j.jeconom.2020.12.001

Carneiro, P., Cattan, S. and Ridpath, N, 2024a. The short- and medium-term impacts of Sure Start on educational outcomes. IFS Report 307, https://ifs.org.uk/publications/short-and-medium-term-impacts-sure-start-educational-outcomes

Carneiro, P., Cruz-Aguayo, Y., Salvati, F. and Schady, N, 2024b. Disruption in the classroom: experimental evidence from Ecuador, https://www.ucl.ac.uk/~uctppca/DisruptiveKids_050124.pdf

Carrell, S. E. and Hoekstra, M. L., 2010. Externalities in the classroom: how children exposed to domestic violence affect everyone’s kids. American Economic Journal: Applied Economics, 2 (1), 211–28, https://doi.org/10.1257/app.2.1.211

Cattan, S., Conti, G., Ginja, R., Farquharson, C. and Pecher, M., 2022. The health impacts of universal early childhood interventions: evidence from Sure Start. IFS Working Paper 21/43, https://ifs.org.uk/publications/health-impacts-universal-early-childhood-interventions-evidence-sure-start

Cattan, S., Kamhöfer, D. A., Karlsson, M. and Nilsson, T., 2023. The long-term effects of student absence: evidence from Sweden. The Economic Journal, 133 (650), 888–903, https://doi.org/10.1093/ej/ueac078

Cattan, S., Fitzsimmons, E., Goodman, A., Phiminster, A., Ploubidis, G. B. and Wertz, J., 2024. Early childhood inequalities—IFS Deaton Review of Inequalities, Oxford Open Economics, 3, i711–i740, https://doi.org/10.1093/ooec/odad072

Conti, G., Pizzo, E., Morris, S. and Melnychuk, M., 2021. The economic costs of child maltreatment in UK. Health Economics, 30, 3087–105, https://onlinelibrary.wiley.com/doi/10.1002/hec.4409

Dearden, L., Gilbert, R., Harron, K., McGrath-Lone, L. and Nasim, B., 2016. Changes in first entry to out-of-home care from 1992 to 2012 among children in England. Child Abuse & Neglect, 51, 163–71, https://doi.org/10.1016/j.chiabu.2015.10.020

Department for Education, 2009. Children Looked After in England (including adoption and care leavers) year ending 31 March 2009, https://webarchive.nationalarchives.gov.uk/ukgwa/20130403134936/ http://www.education.gov.uk/rsgateway/DB/SFR/s000878/index.shtml

Department for Education, 2015. Characteristics of children in need: 2014 to 2015, https://assets.publishing.service.gov.uk/media/5a7f11e6e5274a2e87db39bd/SFR41-2015_Text.pdf

Department for Education, 2022. Drivers of activity in children’s social care, https://assets.publishing.service.gov.uk/media/62961ec9d3bf7f036ddfe7ce/Drivers_of_Activity_in_Children_s_Social_Care.pdf

Department for Education, 2024a. Pupil attendance in schools, https://explore-education-statistics.service.gov.uk/data-tables/pupil-attendance-in-schools

Department for Education, 2024b. Suspensions and permanent exclusions, https://explore-education-statistics.service.gov.uk/data-tables/suspensions-and-permanent-exclusions-in-england

Department for Education and Employment, 1999. Sure Start: a guide for trailblazers (available at https://eric.ed.gov/?id=ED433092)

Emmott, E. H., Jay, M. A. and Woodman, J., 2019. Cohort profile: Children in Need Census (CIN) records of children referred for social care support in England, BMJ Open, 9, e023771, https://doi.org/10.1136/bmjopen-2018-023771

Fort, M., Ichino, A. and Zanella, G., 2016. Cognitive and noncognitive costs of daycare 0–2 for girls. Institute for the Study of Labor (IZA) Discussion Paper 9756, https://docs.iza.org/dp9756.pdf

Gilbert, R., Kemp, A., Thoburn, J., Sidebotham, P., Radford, L., Glaser, D. and MacMillan, H. L., 2009. Recognising and responding to child maltreatment. The Lancet, 373 (9658), 167–80, https://doi.org/10.1016/s0140-6736(08)61707-9

Goff, J. et al., 2013. Evaluation of Children’s Centres in England (ECCE): Strand 3: delivery of family services by Children’s Centres. Department for Education, Research Report DFE-RR297, https://assets.publishing.service.gov.uk/media/5a7cc90fe5274a2f304f00f4/DFE-RR297.pdf

Goodman, J., 2014. Flaking out: student absences and snow days as disruptions of instructional time. NBER Working Paper 20221, https://doi.org/10.3386/w20221

Gray-Lobe, G., Pathak, P. A. and Walters, C. R., 2023. The long-term effects of universal preschool in Boston. Quarterly Journal of Economics, 138, 363–411, https://doi.org/10.1093/qje/qjac036

Havnes, T. and Mogstad, M., 2011. No child left behind: subsidized child care and children’s long-run outcomes. American Economic Journal: Economic Policy, 3 (2), 97–129, https://doi.org/10.1257/pol.3.2.97

Heckman, J. J., Stixrud, J. and Urzua, S., 2006. The effects of cognitive and noncognitive abilities on labor market outcomes and social behavior. Journal of Labour Economics, 24, 411–82, https://doi.org/10.1086/504455

Heeks, M., Reed, S., Tafsiri, M. and Prince, S., 2018. The economic and social costs of crime: second edition, Home Office, Research Report 99, https://assets.publishing.service.gov.uk/media/ 5b684f22e5274a14f45342c9/the-economic-and-social-costs-of-crime-horr99.pdf

HM Treasury, 1998. Comprehensive Spending Review 1998, https://www.gov.uk/government/publications/comprehensive-spending-review-1998

Kliem, S. and Sandner, M., 2021. Prenatal and infancy home visiting in Germany: 7-year outcomes of a randomized trial. Pediatrics, 148 (2), https://doi.org/10.1542/peds.2020-049610

Lavy, V., Paserman, M. D. and Schlosser, A., 2012. Inside the black box of ability peer effects: evidence from variation in the proportion of low achievers in the classroom. The Economic Journal, 122, 208–37, https://doi.org/10.1111/j.1468-0297.2011.02463.x

Lindo, J. M., Schaller, J. and Hansen, B., 2018. Caution! Men not at work: gender-specific labor market conditions and child maltreatment. Journal of Public Economics, 163, 77–98, https://doi.org/10.1016/j.jpubeco.2018.04.007

MacAlister, J., 2022. The independent review of children’s social care, https://webarchive.nationalarchives.gov.uk/ukgwa/20230308122535mp_/https:/childrenssocialcare.independent-review.uk/wp-content/uploads/2022/05/The-independent-review-of-childrens-social-care-Final-report.pdf

Maisey, R., Speight, S. and Haywood, S., with Hall, J., Sammons, P., Hussey, D., Goff, J., Evangelou, M. and Sylva, K., 2013. Evaluation of Children’s Centres in England (ECCE): Strand 2: baseline survey of families using Children’s Centres in the most disadvantaged areas. Department for Education, Research Report DFE-RR260, https://assets.publishing.service.gov.uk/media/5a7a3a8140f0b66a2fc00d2e/DFE-RR260.pdf

Mc Grath-Lone, L., Harron, K., Dearden, L. and Gilbert, R., 2020. Exploring placement stability for children in out-of-home care in England: a sequence analysis of longitudinal administrative data. Child Abuse & Neglect, 109, 104689, https://doi.org/10.1016/j.chiabu.2020.104689

Meadows, P. et al., 2011. National Evaluation of Sure Start Local Programmes: an economic perspective. Department for Education, Research Report DFE-RR073, https://www.gov.uk/government/ publications/national-evaluation-of-sure-start-local-programmes-an-economic-perspective

Ministry of Justice, 2024. Youth Justice Statistics: 2022 to 2023, https://www.gov.uk/government/statistics/youth-justice-statistics-2022-to-2023

National Audit Office, 2011. The cost of a cohort of young offenders to the criminal justice system, https://www.nao.org.uk/wp-content/uploads/2010/12/1011663_technical_paper.pdf

Preston, O., Godar, R., Lefevre, M., Boddy, J. and Firmin, C., 2021. Considerations in the use of local and national data for evaluating innovation in children’s social care (Version 1). University of Sussex. https://hdl.handle.net/10779/uos.23481599.v1

Robling, M. et al., 2021. The Family Nurse Partnership to reduce maltreatment and improve child health and development in young children: the BB:2 6 routine data-linkage follow-up to earlier RCT. Public Health Research, 9 (2), https://doi.org/10.3310/phr09020

Sandner, M., Thomsen, S. L. and González, L., 2024. Preventing child maltreatment: beneficial side effects of public childcare. The Economic Journal, forthcoming, https://academic.oup.com/ej/advance-article-abstract/doi/10.1093/ej/ueae070/7717371

Smith, G., Field, K., Smith, T., Noble, S., Smith, T. and Plunkett, E., 2014. Evaluation of Children’s Centres in England (ECCE): the extent to which centres ‘reach’ eligible families, their neighbourhood characteristics and levels of use.

Department for Education, Research Report DFE-RR358, https://assets.publishing.service.gov.uk/media/5a7ee6a240f0b6230268c3fd/RR358_-_FINALrevisedReachReport07042014.pdf

Viding, E. and McCrory, E., 2024. Individuals as active co-creators of their environments: implications for prevention of inequalities—IFS Deaton Review of Inequalities, Oxford Open Economics, 3, i749–i753, https://doi.org/10.1093/ooec/odad071

Webb, C. J. R. and Bywaters, P., 2018. Austerity, rationing and inequity: trends in children’s and young peoples’ services expenditure in England between 2010 and 2015. Local Government Studies, 44, 391–415, https://www.tandfonline.com/doi/abs/10.1080/03003930.2018.1430028

 

Endnotes

  1. 1

    These studies build on earlier evaluations – for example, National Evaluation of Sure Start and Evaluation of Children’s Centres in England – demonstrating the short-run benefits of Sure Start, which was associated with improved higher-quality home environments, improved parenting and better maternal mental health.

  2. 2

    A future report will bring together findings from each of the evaluations of Sure Start on health, educational attainment, school absence and exclusions, crime, and children’s social care outcomes, and will conduct a comprehensive cost–benefit analysis of the programme.

  3. 3

    SSLPs had much larger budgets per centre than SSCCs, though both were responsible for similar numbers of children. This continued even after all centres rebranded as SSCCs, as former SSLP budgets were ring-fenced. This meant they spent more on outreach to parents, and on teacher input into learning development, among other things (National Audit Office, 2009). SSLPs also had less restrictive offerings, allowing for more input from parents in the local community as to which services should be offered.

  4. 4

    See Cattan et al. (2022) for a more detailed description of the rollout.

  5. 5

    Rates of youth offending are much lower today than they were among cohorts born in the middle of the 1990s (Ministry of Justice, 2024). Absence and exclusion rates have remained relatively stable, though have increased sharply since the COVID-19 pandemic (Department for Education, 2024a, b). In contrast, some forms of social services activity have risen over time, for instance children in out-of-home care (Dearden et al., 2016).

  6. 6

    Suspensions are very rare in primary school – even by the end of primary school, at age 11, only 1% of students are suspended over an academic year.

  7. 7

    See https://osf.io/6ecrz/wiki/home/.

  8. 8

    Some children may also receive an immediate provision of services (this can be with or without also receiving a needs assessment), for instance, cases where a child urgently requires protection and enters into the care of the local authority immediately

  9. 9

    See Department for Education (2015).

  10. 10

    In practice, there is substantial overlap between categories; children often have multiple needs but only the most severe reason is recorded as the primary need. 

  11. 11

    Some studies suggest mixed effects arising from maternal employment. Lindo, Schaller and Hansen (2018), for instance, find increases in child maltreatment with rises in female employment, in part due to more time spent between children and fathers.

  12. 12

    There are around 33,000 LSOAs in England and the average LSOA has a population of around 1,500 residents.

  13. 13

    Theft and drug offences are those crimes that fall in their respectively named Home Office offence groups. Weapons offences are those in the offence group titled ‘Possession of weapons’, and violent crimes are those in the offence group titled ‘Violence against the person’. 

  14. 14

    This is not a perfect classification, as within each broad group of outcomes, interventions vary in intensity. For instance, a referral to children’s social care is more common than a school suspension.

  15. 15

    Previous work has shown that Sure Start did not have an impact on parental employment, so it is unlikely that the programme affected children’s eligibility for FSM directly.

  16. 16

    See Appendix A.4 for figures.

  17. 17

    Another possibility is that high-need families were less engaged with Sure Start services. As Carneiro et al. (2024a) demonstrate, in the earlier years of the programme, many resources were directed towards outreach and engaging families who were less likely to use services; therefore, this seems less likely, although it could play a role in the later phase of Sure Start.

  18. 18

    In any case, as many of the costs of maltreatment run through health channels, we also do not want to double-count the health benefits of Sure Start, which are accounted for in the cost–benefit analysis conducted in Cattan et al. (2022).

  19. 19

    The average length of a custodial sentence for young people (excluding life sentences – though these are very rare) under age 16 is 11 months.

  20. 20

    This figure is discounted to account for the fact that most of the savings happen when the individual is almost 16, while the last expenditure on Sure Start happens 11 years earlier. We follow the HM Treasury approach specified in the ‘Green Book’, discounting all future benefits at an annual rate of 3.5%, from age 5 onwards.

  21. 21

    This appendix draws on Appendix A1 of Carneiro et al. (2024), using examples from results in this report.