Introduction
Austerity policies, retrenchment, and decentralisation have affected the quality and availability of public services in inequitable ways over the past decade (Webb and Bywaters, Reference Webb and Bywaters2018; Hernandez, Reference Hernandez2021). Children’s services in England devote resources to supporting families of children who may develop health or developmental impairments and safeguard children who may be at risk of maltreatment. They are also responsible for arranging care placements and care-leaving support for children who are not able to live with their families of origin.
There has been renewed international interest in studies that view children’s services through a public health and inequalities lens; these studies show that children living in more deprived communities are far more likely to: require additional child welfare support; experience abuse or neglect; become subject to child protection investigations; and be placed in care (Sethi et al., Reference Sethi, Bellis, Hughes, Gilbert, Mitis and Galea2013; Bywaters et al., Reference Bywaters, Bunting, Davidson, Hanratty, Mason, McCartan and Steils2016; Doidge et al., Reference Doidge, Higgins, Delfabbro, Edwards, Vassallo, Toumbourou and Segal2017; Bywaters et al., Reference Bywaters, Brady, Bunting, Daniel, Featherstone, Jones, Morris, Scourfield, Sparks and Webb2018; Webb et al., Reference Webb, Bywaters, Scourfield, McCartan, Bunting, Davidson and Morris2020). As adults, they are at higher risk of mortality and other adverse outcomes than the general population (Gypen et al., Reference Gypen, Vanderfaeillie, De Maeyer, Belenger and Van Holen2017; Murray et al., Reference Murray, Lacey, Maughan and Sacker2020; van IJzendoorn et al., 2020; Jackisch et al., Reference Jackisch, Ploubidis and Gondek2021). There is also evidence of an ‘inverse care law’ for children’s services in England similar to that found in medical care (Tudor Hart, Reference Tudor Hart1971), an inverse relationship between intervention rates and population needs (Bywaters et al., Reference Bywaters, Brady, Bunting, Daniel, Featherstone, Jones, Morris, Scourfield, Sparks and Webb2018; Webb et al., Reference Webb, Bywaters, Scourfield, McCartan, Bunting, Davidson and Morris2020).
Such studies highlight the potential of proactively addressing demand- and supply-side determinants of intervention (Bywaters et al., Reference Bywaters, Brady, Bunting, Daniel, Featherstone, Jones, Morris, Scourfield, Sparks and Webb2018; Hood et al., Reference Hood, Goldacre, Gorin and Bywaters2020a; Webb et al., Reference Webb, Bywaters, Scourfield, McCartan, Bunting, Davidson and Morris2020). To prevent the escalation of need or risk, it is argued, policies are needed to: remedy structural inequalities; alleviate, and as far as possible eliminate, root causes of child abuse and neglect, principal among them being poverty; and develop community-led infrastructures of family support (Featherstone et al., Reference Featherstone, Gupta, Morris and White2018b). This reflects wider calls to invest in preventive services to improve the support offered to children and families, deinstitutionalise children, and reduce health inequalities (Goldman et al., Reference Goldman, Bakermans-Kranenburg, Bradford, Christopoulos, Ken, Cuthbert, Duchinsky, Fox, Grigoras, Gunnar and Ibrahim2020; van IJzendoorn et al., 2020).
Literature review
This section introduces literature on the relationship between incidence of child welfare interventions, poverty, and the funding of children’s services in England. We present existing evidence on the relationship between poverty, public services expenditure, and children’s services quality before discussing some political implications of inspections of children’s services quality.
Socioeconomic determinants of incidence
Existing literature has primarily focused on inequalities in child maltreatment incidence – typically approximated by children’s services interventions and substantiation of abuse – rather than the quality of services that prevent maltreatment (Hood et al., Reference Hood, Goldacre, Grant and Jones2016a). Lower intervention rates may not necessarily reflect prevented abuse or neglect: they could equally reflect more ‘missed’ maltreatment, especially in poor quality services. They might also represent changes in local and national policy that shift thresholds for intervention, effectively rationing services (Hood et al., Reference Hood, Goldacre, Grant and Jones2016a; Devaney, Reference Devaney2019; Hood et al., Reference Hood, Goldacre, Gorin and Bywaters2020a). The same can be true of higher intervention rates. Poor quality services may operate on unreasonably low levels of risk tolerance; this may lead to worse social and health outcomes for children than if they had not been subject to state intervention (Featherstone et al., Reference Featherstone, Gupta, Morris and Warner2018a; Murray et al., Reference Murray, Lacey, Maughan and Sacker2020; Jackisch et al., Reference Jackisch, Ploubidis and Gondek2021). It is important to understand both incidence and quality in child welfare.
Cuts to welfare benefits beginning in 2010 have disproportionately affected families with children (Tucker, Reference Tucker2017; Joseph Rowntree Foundation, 2020). A growing body of literature has identified causal relationships between income, poverty, and child abuse and neglect (Bywaters et al., Reference Bywaters, Bunting, Davidson, Hanratty, Mason, McCartan and Steils2016). This relationship creates a ‘social gradient’ in the incidence of child welfare interventions, whereby poorer families are increasingly likely to have children living in out-of-home care (Bywaters et al., Reference Bywaters, Brady, Bunting, Daniel, Featherstone, Jones, Morris, Scourfield, Sparks and Webb2018). While there are differences among countries, this is a global phenomenon observable in the United Kingdom (Bywaters et al., Reference Bywaters, Brady, Bunting, Daniel, Featherstone, Jones, Morris, Scourfield, Sparks and Webb2018; Webb et al., Reference Webb, Bywaters, Scourfield, McCartan, Bunting, Davidson and Morris2020), the USA (Drake and Jonson-Reid, Reference Drake, Jonson-Reid, Korbin and Krugman2014; Eckenrode et al., Reference Eckenrode, Smith, McCarthy and Dineen2014), Canada (Esposito et al., Reference Esposito, Chabot, Rothwell, Trocmé and Delaye2017), Australia (Doidge et al., Reference Doidge, Higgins, Delfabbro, Edwards, Vassallo, Toumbourou and Segal2017), and Aotearoa New Zealand (Keddell et al., Reference Keddell, Davie and Barson2019). Studies in England by Hood et al. (Reference Hood, Goldacre, Gorin and Bywaters2020a) and Webb et al. (Reference Webb, Bywaters, Scourfield, McCartan, Bunting, Davidson and Morris2020) have shown that these inequalities are embedded in the child welfare system at the stage of referral and in ‘child in need’ designations under Section 17 of the Children Act 1989. The social gradient typically gets steeper as the severity of state intervention approaches care entry.
Where poverty creates demand for services, the supply to meet demand depends on central funding. Funding of services has become less proportionate to underlying levels of need under austerity (Webb and Bywaters, Reference Webb and Bywaters2018). Early help services have been heavily and inequitably defunded in the 2010 decade, with the greatest reductions being over 50 per cent per child between 2010 and 2015 (Webb and Bywaters, Reference Webb and Bywaters2018; Action for Children, 2020; YMCA, 2020). These services typically include some combination of ‘early intervention’ and ‘family support’ (Frost et al., Reference Frost, Abbott and Race2015). The former might include specific interventions – for example, parenting programmes such as ‘Triple-P’ (Sanders, Reference Sanders2008) – and the latter might include community development social work (Jack and Gill, Reference Jack and Gill2010).
Hood et al. (Reference Hood, Gorin, Goldacre, Muleya and Bywaters2020b) found that early help services had shifted towards more complex forms of ‘late-early’ intervention as a response to funding cuts, diminishing the capacity of local authorities to resolve more universal needs. A shift away from open-door provision has been reported nationally (Action for Children, 2020). Recent research has identified that cuts to these services are causally associated with increases in the rates of children in need and sixteen to seventeen year olds in care (Bennett et al., Reference Bennett, Webb, Mason, Schlüter, Fahy, Alexiou, Wickham, Barr and Taylor-Robinson2021; Webb, Reference Webb2021). While a £14million investment in ‘Family Hubs’ (Department for Education, 2021a) has been proposed, this funding, if split evenly, would equate to only 0.58 per cent of the average spending that each authority has cut from their family support expenditure between 2010 and 2019 (authors’ analysis, Department for Education, 2021b).
Poverty, ‘early help’, and failure demand
The concept of ‘failure demand’ in systems-thinking approaches to public services is of increasing relevance to children’s services (Seddon, Reference Seddon2008; Munro, Reference Munro2010; Hood, Reference Hood2015). Failure demand is defined as ‘demand caused by a failure to do something or do something right for the customer’ (Seddon and Brand, Reference Seddon and Brand2008: 8). In relation to children’s services, Hood (Reference Hood2015: 10) writes: ‘issues that are not resolved straightaway keep reappearing and cumulatively start to overload the system’s ability to cope’. This type of demand is theorised to arise from a failure to ‘intervene quickly and efficiently at the point that families start to experience problems’ (Hood, Reference Hood2015: 10); a limited capacity to respond to a wide variety of needs (Wastell, Reference Wastell2011; Wastell and White, Reference Wastell and White2014; Hood, Reference Hood2015); the arrangement of socio-technical systems (Broadhurst et al., Reference Broadhurst, Hall, Wastell, White and Pithouse2010a; Gibson and O’Donovan, Reference Gibson and O’Donovan2013); a defensive approach towards both family and institutional risk (Munro, Reference Munro2010; Hood, Reference Hood2015); and a technocratic focus on key performance metrics and proceduralism (Broadhurst et al., Reference Broadhurst, Wastell, White, Hall, Peckover, Thompson, Pithouse and Davey2010b; Wastell and White, Reference Wastell and White2014; Hood, Reference Hood2015; Hood et al., Reference Hood, Grant, Jones and Goldacre2016b). Even when not explicitly stated, failure demand is often implicated in what has been called the ‘care crisis’ in the child welfare system. This is neatly illustrated in the title to the All Party Parliamentary Group for Children and the National Children’s Bureau’s (2018) inquiry report on the subject: ‘Storing Up Trouble’.
The prevalence of poverty (demand) and the lack of availability and diversity of early help (supply) can be hypothesised to be predictors of the quality of children’s services. Both factors are theorised to create failure demand, which can mean that the services provided to families to prevent maltreatment and abuse can be inappropriate, untimely, or, at worst, unavailable. As such, children’s services quality can be linked to national policy. Placing local services under quality inspection without attention to the impact of geographical inequalities in poverty and funding, then expecting them to have equal capabilities of achieving the same levels of quality, therefore becomes problematic. This is especially concerning when the consequence of failing to meet adequate standards can mean the removal of public ownership and dramatic reform of services that, if the underlying needs of their population and service were properly met through national policy, might not have been needed.
Socioeconomic determinants of children’s services quality in England
The Office for Standards in Education, Children’s Services, and Skills (Ofsted) is responsible for regulating the quality of local authority children’s services in England through inspections, which typically include a mixture of observational fieldwork, case audits, and analysis of key performance indicators assessed against an inspection framework. Ofsted inspections result in a published judgement or ‘grade’ for children’s services. Prior to 2013, these ranged through ‘inadequate’, ‘adequate’, ‘good’, and ‘outstanding’. Since the introduction of the Single Inspection of LA Children Services Framework (SIF) in 2013, the ‘adequate’ outcome category was replaced with the ‘requires improvement to be good’ category. While the validity of Ofsted inspection judgements and the extent to which they reflect good-quality outcomes for children remains a point of considerable contention (La Valle et al., Reference La Valle, Holmes, Gill, Brown, Hart and Barnard2016; Hood, Reference Hood2019), inspection judgements are the only national, publicly available and broadly comparable measure of local authority children’s services quality in England. Moreover, the results of Ofsted inspections have very real consequences for services and families, and for the provision of children’s services more generally (Jones, Reference Jones2019). For example, Hood et al. (Reference Hood, Grant, Jones and Goldacre2016b) found that in the year following an ‘inadequate’ Ofsted outcome services tended to make greater use of more intrusive and resource intensive child protection interventions.
A limited number of studies address the association between Ofsted judgements and either socioeconomic deprivation, a principal demand factor, or, at the level of supply, service funding. The National Audit Office (NAO) in two separate reports on children’s social care found no association between children’s services expenditure and service quality (National Audit Office, 2016, 2019). Ofsted later claimed that ‘inadequacy is not a function of size, deprivation or funding, but of the quality of leadership and management’. Prompted by the publication of statistics that reported deprived local authorities with positive inspections had around 21 per cent higher expenditure per child than those with ‘inadequate’ or ‘requires improvement’ outcomes (Bywaters et al., Reference Bywaters, Webb and Sparks2017), Ofsted included an admission that there may be a link between deprivation and children’s services inspection outcomes but rejected any link between spending and outcomes (Ofsted, 2017). Analysing inspections from two different Ofsted inspection frameworks, Wilkins and Antonopoulou (Reference Wilkins and Antonopoulou2020a, 2020b) reached similar conclusions: they found an association between deprivation and quality but not between spending and quality.
These studies used either total expenditure per child or ‘social work expenditure’ as predictors of inspection outcomes. The NAO’s 2016 report used an undefined aggregate of children’s services spending, a subset of safeguarding expenditure, which usually accounts for less than a third of all expenditure on average. Their 2019 report, and the research by Wilkins and Antonopoulou (Reference Wilkins and Antonopoulou2020a, 2020b), used total expenditure per child. No study has assessed the link between spending and quality of services while distinguishing between “preventative” and “acute” spend. This is potentially important, given the strong evidence-base for the effectiveness of early intervention in reducing child maltreatment (Yousafzai, Reference Yousafzai2020) and the specific role that early help and family support spending plays in reducing ‘failure demand’ (Hood, Reference Hood2015). The studies also only consider inspections from only one inspection framework at a time. While this side-steps problems caused by the lack of independence of observations within judgements of the same local authority and inspection framework over time, it curtails the power of models to identify multiple or small-to-moderate effects (Button et al., Reference Button, Ioannidis, Mokrysz, Nosek, Flint, Robinson and Munafò2013). To our knowledge, studies using analytical approaches such as multilevel models that can utilise larger samples while incorporating hierarchical dependence have not been published. Such models are necessary for accurately assessing predictors of Ofsted outcome in pooled inspection data as changes in framework have often been accompanied by changes in the inspectorate’s propensity to rate services as ‘Good’ or ‘Outstanding’. For instance, under the SIF, 34.9 per cent of inspection outcomes were ‘Good’ (32.9 per cent) or ‘Outstanding’ (2 per cent) (author’s analysis). Under the Inspecting Local Authority Children’s Services (ILACS) framework 56.1 per cent of outcomes were ‘Good’ (43.9 per cent) or ‘Outstanding’ (12.2 per cent) (author’s analysis).
The political consequences of children’s services quality
Kim and Warner (Reference Kim and Warner2021) report that local government responses to fiscal austerity have opened up publicly-controlled services to marketisation. Given the uneven nature of the cuts to children’s services, this creates a more fundamental concern about geographical inequalities in the relationship between citizens and the state along the lines of socioeconomic class. Since 2015, children’s services deemed to be of ‘inadequate’ quality can be subject to ‘take over’ (Stevenson, Reference Stevenson2015; Jones, Reference Jones2019). This policy reform was explained by a government spokesperson in 2015 as ‘a formalised academy-style system … any local authority judged as inadequate by Ofsted has to show significant improvement within six months or be taken over’ (Stevenson, Reference Stevenson2015).
To date, nine local authority children’s services have been removed from public control and replaced with independent trusts, with a tenth service poised to be taken over; these account for approximately 6.5 per cent of all children’s services in England covering approximately 8 per cent of the child population (Turner, Reference Turner2020a). Almost half of these local authorities - Doncaster, Sunderland, Sandwell, and Birmingham - are in the most deprived 20 per cent of all local authorities in England according to the Indices of Multiple Deprivation 2019 (Ministry of Housing, Communities and Local Government, 2019a). The remaining local authorities - Slough, Northamptonshire, Kingston upon Thames (now merged with Richmond upon Thames), and West Sussex - all reported severe fiscal pressures and risk of insolvency as a result of austerity policies and central government underfunding prior to conversion to independent children’s trusts (Rutter, Reference Rutter2018; Vise, Reference Vise2019; Powling, Reference Powling2020; Robson and Manning, Reference Robson and Manning2020; Turner, Reference Turner2020b). If poverty and service funding are significant determinants of children’s services quality and a consequence of government policy and factors outside local authority control, such a policy may be inappropriate.
Data
Outcomes from 415 Ofsted inspections of local authority children’s services in England over eight fiscal years, between 1st April 2011 and 31st March 2019 (Ofsted, 2020), were linked to local authority data on children’s services expenditure (Department for Education, 2021b) over the same period, and to Index of Multiple Deprivation scores (Ministry of Housing, Communities and Local Government, 2019a). These data are nested within 152 upper-tier local authorities that provide children’s services across England. Re-inspections over the study period (N = 31) were excluded from the analysis. Four inspections of the City of London and the Isles of Scilly were removed; these outlier local authorities are commonly excluded in local area studies due to their small resident populations. Inspections from the short-lived Targeted Looked After Children (TLAC) framework for children in care, which resulted in only five published outcomes, were also excluded. One authority had no spending data available to match to Safeguarding and Looked After Children (SLAC) inspections due to a local authority merger, and was excluded. The final number of inspections included was 374.
‘Overall judgements’ or ‘overall effectiveness’ ratings from Ofsted inspections are based on four-tiers of outcomes: ‘inadequate’, ‘requires improvement to be good’, ‘good’, and ‘outstanding’. Our sample included seventy-four ‘inadequate’ overall judgements (19.8 per cent), 166 ‘requires improvement’ overall judgements (44.4 per cent), 121 ‘good’ overall judgements (32.4 per cent), and thirteen ‘outstanding’ judgements (3.5 per cent). Four inspection frameworks were used by Ofsted between 2009 and 2019, excluding the TLAC. The first was the Safeguarding and Looked After Children Inspections (SLAC) framework (August 2009–August 2012); followed by the Child Protection Inspections (CPI) framework (July 2012–August 2013); which became the Single Inspection of Local Authority Children Services Framework (SIF) (February 2014–August 2019). The current framework is the Inspecting Local Authority Children’s Services (ILACS) framework (from March 2018).
Table 1 shows that the SLAC framework had no comparable ‘overall judgement/effectiveness’ inspection outcome. To create some parity of measurement over the study period, we included only safeguarding overall effectiveness judgements for SLAC and excluded the Looked After Children services overall judgements. For SIF and ILACS, we included only the overall judgements for the services and not the judgements for every domain. We conducted further statistical tests using each inspection framework’s domains to assess the extent to which overall judgements corresponded with domain judgements.
Table 1 Ofsted inspection frameworks and domains for children’s services
We created two categories of children’s services expenditure: ‘preventative’ and ‘safeguarding’, following strategies for delineating spending in administrative data (Webb and Bywaters, Reference Webb and Bywaters2018). The preventative category included spending on children’s centres, family support services, services for young people, and other children’s and families’ services not directly related to child protection social work or children in care. Safeguarding expenditure incorporated all spending associated with child protection social work: child protection investigations, plans, and safeguarding boards. While a more detailed breakdown of spending categories is possible, greater specificity introduces inconsistency across local authorities and over time. Expenditure was adjusted for inflation using the Office for National Statistics’ (ONS) GDP deflator. Per capita spend was derived using ONS population estimates for individuals under eighteen years old. Expenditure was scaled to £100’s per child and centred at mean values. Summary statistics are provided in table 2.
Table 2 Summary statistics for variables in inspection data
The Indices of Multiple Deprivation (IMD) are a relative measure of area-based multidimensional deprivation, released approximately every five years by the Ministry of Housing, Communities, and Local Government (MHCLG) (MHCLG, 2019b). The index is constructed from seven weighted domains of deprivation: income deprivation (22.5 per cent); employment deprivation (22.5 per cent); education, skills, and training deprivation (13.5 per cent); health deprivation and disability (13.5 per cent); crime (9.3 per cent); barriers to housing and services (9.3 per cent); and living environment deprivation (9.3 per cent) (MHCLG, 2019a). The latest IMD scores, released in 2019, were calculated using data from 2015/16, and were therefore considered the most suitable ‘mid-point’ measure of local deprivation for this analysis. As IMD scores have no straightforward interpretation these were transformed into deciles, where decile ten represents the most deprived ten per cent of all local authorities and decile one the least deprived ten per cent.
Methods
Multilevel logistic regression models were estimated to predict whether inspections resulted in ‘good’ or ‘outstanding’ Ofsted judgements of overall child welfare services quality, using IMD score, preventative expenditure, and safeguarding expenditure as predictors. As mentioned in the literature review, the structure of pooled inspection data can lead to biased estimates when observations are not truly independent (Robson and Pevalin, Reference Robson and Pevalin2015) – for example, when there are multiple inspections of a single authority or where certain years or inspection frameworks are associated with very different outcomes on aggregate. Local authority and year variables were entered into the model as random effects to adjust for this bias. Year was chosen as a way to capture variation associated with both time and inspection framework, which are intrinsically linked. A null model including only local authority and year random effects was estimated to test, using a Likelihood Ratio Test, whether the inclusion of expenditure per child and IMD decile significantly improved model fit. Model fit was assessed using predictive accuracy, and predicted probabilities were calculated for the effects of expenditure per child and IMD decile and plotted with 95 per cent confidence intervals to contextualise effect sizes.
To assess the extent to which findings based on overall judgements (SIF, ILACS) and safeguarding judgements (SLAC) were applicable to other domains of inspection and across frameworks, we calculated Cronbach’s alpha and Gutmann’s lambda-6 values for all SLAC, SIF, and ILACS inspection outcomes as well as for each framework’s set of outcome domains. Since CPI inspections include no specific domains and only fifty inspections, they were excluded from these analyses. We then estimated a single-factor categorical Confirmatory Factor Analysis for each of the SIF and ILACS areas of inspection and calculated the Spearman rank correlation coefficient for the two inspection outcomes that made up the SLAC framework to assess domain and overall judgement communality.
Finally, we tested whether estimates of the effect of deprivation decile varied notably between different domains of multiple deprivation. Separate models using deciles from each of nine multiple deprivation domains were estimated. Deprivation decile effect estimates with 95 per cent confidence intervals were compared with results from the main model which used deciles derived from the composite IMD 2019 index. All analyses were conducted in R version 4.0.4. For reproducibility, details of all packages used, further information on model specification, and a copy of data and code are available from the following repository: https://github.com/cjrwebb/osd-repo
Findings
The association between expenditure, deprivation, and positive inspection outcomes
Model results for a multilevel logistic regression model predicting ‘good’ or ‘outstanding’ Ofsted inspection outcomes are shown in table 3. The model had significantly better accuracy (77.5 per cent) than the no-information rate (64.2 per cent) and was a better fit to the underlying data than a simpler comparison model that included only local authority and year fixed effects (LRT = 19.54, p<0.001). There was evidence of an association between expenditure on preventative services per child and the likelihood of a positive inspection judgement; and between IMD decile and the likelihood of a positive inspection judgement. All else being equal, an increase of £100 per child on preventative services (around 0.7 standard deviations) was associated with a 1.69 times increase in the odds of a ‘good’ or ‘outstanding’ inspection outcome (B = 0.525, 95 per cent CI: 0.241, 0.808). A one-decile increase in IMD score (higher deprivation) was associated with a 16 per cent decrease in the odds of a ‘Good’ or ‘Outstanding’ inspection outcome (B = −0.173, 95 per cent CI: −0.287, −0.058). In contrast, there was no statistically significant association between safeguarding services expenditure per child and the likelihood of a positive inspection outcome (B = −0.028, 95 per cent CI: −0.422, 0.368). Local authority membership (
$\sigma $
= 0.802) and year of inspection (
$\sigma $
= 0.852) random effects indicated large differences in the likelihood of a positive inspection outcome depending on local authority and year, illustrating the need to adjust for the hierarchical structure of inspection data in order to accurately assess the effects of expenditure and deprivation on the quality of child welfare services when using pooled inspection data over time.
Table 3 Multilevel logistic regression model predicting the likelihood of a children’s services inspection resulting in a ‘good’ or ‘outstanding’ outcome by expenditure on preventative services per child, expenditure on safeguarding per child, and Indices of Multiple Deprivation decile
B = Model estimate (log odds), S.E. = standard error, O.R. = odds ratio, CI = Parametric confidence interval,
${\sigma ^2}\;$
= Variance,
$\sigma $
= Standard deviation, NIR = No-Information Rate
Applicability of findings on inspection outcomes across domains of children’s services
Table 4 shows the results of Cronbach’s alpha and Gutmann’s lambda-6 tests of internal reliability, as well as results showing communality across inspection outcome domains using Confirmatory Factor Analysis for frameworks where the number of domains is greater than two and Spearman’s rank correlation coefficients where there are only two domains. We found that the twelve domains of assessment across frameworks had very good internal reliability (
$\alpha $
= 0.87,
$\lambda $
= 0.95), with the weakest internal reliability within the two SLAC assessment outcomes (
$\alpha $
= 0.77,
$\lambda $
= 0.66). This was unsurprising given their limited number. A Spearman’s rho correlation confirmed that there are some significant discrepancies between SLAC safeguarding and SLAC looked-after children judgements (
$\rho $
= 0.678, p<0.01). By contrast, SIF and ILACS areas of judgement had high communality, shown by the good fit statistics of a one-factor CFA (SIF: TLI = 1, SRMR = 0.041,
${\chi ^2}$
= 18.07, p = 0.054; ILACS: TLI = 0.999, SRMR = 0.027,
${\chi ^2}$
= 2.81, p = 0.246). While this indicates that the findings for overall inspection outcomes across all frameworks are likely applicable to all domains of inspection and, therefore, the quality of all aspects of child welfare services, the SIF and ILACS CFA factor loadings and only moderate SLAC inspection correlation (
$\rho $
= 0.68) suggests that the findings may be slightly less applicable in the context of quality of services for children in care and care leavers (SIF:
${\rm{\Gamma }}$
= 0.787, ILACS:
${\rm{\Gamma }}$
= 0.899), and services for adoption and post-adoption support (SIF:
${\rm{\Gamma }}$
= 0.818).
Table 4 Internal consistency and communality across Ofsted children’s services inspection frameworks and areas of judgement
$\alpha $
= Cronbach’s alpha,
$\lambda $
= Gutmann’s lambda-6,
$\rho $
= Spearman’s rank correlation,
${\chi ^2}$
= Chi-squared goodness-of-fit statistic;
${\rm{\Gamma }}$
= Unstandardised factor loading, S.E. = Standard error.
1 Full inspection data was used, which included additional SLAC and ILACS inspections that were excluded from the main analysis due to inspection dates without expenditure data.
2 Loading fixed to 1 to address very high correlation with indicator variable.
Contextualisation of effect sizes
Effect sizes were contextualised by plotting the predicted probabilities for changes in each predictor between their minimum and maximum observed values, while holding all other predictors and random effects constant at their mean value. Figure 1 shows the effect of expenditure on preventative services per child on the probability of an inspection resulting in a ‘good’ or ‘outstanding’ judgement; Figure 2 shows the effect of safeguarding expenditure per child on the probability of a positive inspection outcome; and Figure 3 shows the effect of IMD decile on the likelihood of a positive inspection outcome.
Figure 1. Predicted probability of a ‘good’ or ‘outstanding’ inspection outcome by amount of spending on preventative services per child, adjusted for levels of deprivation and safeguarding spending. Dashed lines indicate 95 per cent CI upper and lower bounds.
Figure 2. Predicted probability of a ‘good’ or ‘outstanding’ inspection outcome by amount of spending on safeguarding services per child, adjusted for levels of deprivation and preventative spending. Dashed lines indicate 95 per cent CI upper and lower bounds.
Figure 3. Predicted probability of a ‘good’ or ‘outstanding’ inspection outcome by IMD decile, adjusted for levels of preventative and safeguarding spending. Dashed lines indicate 95 per cent CI upper and lower bounds.
The effects are substantial; all else being equal, the predicted probability of a local authority in the 25th percentile of preventative spending (£227 per child) receiving a positive inspection outcome is around 23 per cent (95 per cent CI: 13 per cent, 37 per cent) whereas the equivalent predicted probability for a local authority in the 75th percentile of spending (£385 per child) is approximately 40 per cent (95 per cent CI: 26 per cent, 56 per cent). A local authority in the least deprived 10 per cent of all local authorities had a 53 per cent probability of their children’s services receiving a positive inspection outcome (95 per cent CI: 34 per cent, 72 per cent), at mean levels of expenditure, whereas a local authority in the most deprived 10 per cent with identical levels of expenditure had a predicted probability of only 19 per cent (95 per cent CI: 10 per cent, 35 per cent).
Sensitivity to Indices of Multiple Deprivation domain
Table 5 and Figure 4 show the extent to which the effect of deprivation differs depending on multiple deprivation domain used in the model in contrast to the composite Indices of Multiple Deprivation score. Most of the deciles of domains that shared very high (r > 0.8) correlation with the composite score deciles (Income, Employment, Health, Crime, and Income Deprivation Affecting Children; Appendix Table 1), with the exception of the Income Deprivation Affecting Older People Index, resulted in only slightly weaker estimations of the effect of one-unit increases in deprivation – between an 11.3 per cent and a 14 per cent decrease in the odds of a positive Ofsted judgement. Education, Skills and Training deprivation decile was strongly correlated with the composite IMD score decile (r = 0.767), but resulted in a slightly stronger estimated effect, a 17.8 per cent decrease in the odds of a positive Ofsted judgement. Barriers to housing and services domain decile was negatively correlated with the IMD composite score (r = −0.321) and had a negligible association with Ofsted outcome (95 per cent confidence interval range: 0.947–1.160). This was also the case for living environment deprivation decile (95 per cent confidence interval range: 0.888–1.101), which was weakly correlated with composite score decile (r = 0.347). Predicted probabilities for each domain’s deciles can be found in Appendix Figure 1.
Table 5 Comparison of IMD Domain effect sizes with 95% confidence intervals
Figure 4. Comparison of IMD Domain effect sizes with 95 per cent confidence intervals.
Discussion
Implications for child welfare
This study suggests that both greater expenditure on preventative services and lower levels of deprivation are associated with higher likelihoods of a positive quality inspection outcomes. The findings challenge existing literature and dominant policy narratives in England from the NAO and Ofsted concerning the association between supply- and demand-side determinants of child welfare interventions and children’s services quality. Greater funding of services that address universal or emerging needs is associated with increased likelihood of positive quality assessments (Bywaters et al., Reference Bywaters, Brady, Bunting, Daniel, Featherstone, Jones, Morris, Scourfield, Sparks and Webb2018; Yousafzai, Reference Yousafzai2020). In contrast, higher spending on activities related to child protection investigations and plans was not associated with better odds of a positive inspection.
The rising numbers of children in the care system in Britain has been described as a ‘crisis’ (House of Commons Library, 2018). Globally, rates of child maltreatment pose enormous moral and public health concerns (Sethi et al., Reference Sethi, Bellis, Hughes, Gilbert, Mitis and Galea2013). Mass separation of children from their birth families, and institutionalisation in the care system (Goldman et al., Reference Goldman, Bakermans-Kranenburg, Bradford, Christopoulos, Ken, Cuthbert, Duchinsky, Fox, Grigoras, Gunnar and Ibrahim2020; van IJzendoorn et al., 2020), are neither sustainable solutions nor do they do justice to the rights of children and the rights of parents when such outcomes could have been avoided with the provision of adequate state support. High-quality children’s services are essential and access to them should be equitable. Coordinated preventive and early help-focused approaches to social care services are needed to ensure the quality of services to address child maltreatment and reduce health inequalities (Sethi et al., Reference Sethi, Bellis, Hughes, Gilbert, Mitis and Galea2013; Goldman et al., Reference Goldman, Bakermans-Kranenburg, Bradford, Christopoulos, Ken, Cuthbert, Duchinsky, Fox, Grigoras, Gunnar and Ibrahim2020; Yousafzai, Reference Yousafzai2020). However, these services are not a replacement for a welfare system that adequately supports families’ material needs, ensuring they have a decent income, secure housing in good repair, and access to employment. For children’s services to be ‘good’ they need to be embedded within a society that provides the necessary infrastructure for family life. Sustained inattention to poverty in public policy, and budgets that continue to create shortfalls in local authority finances, create conditions where adequately funding preventative services becomes impossible. This is likely to both increase the prevalance of risks to children’s health and wellbeing and reduce the capacity of local children’s services to respond to such risks (Hood, Reference Hood2015; Bywaters et al., Reference Bywaters, Brady, Bunting, Daniel, Featherstone, Jones, Morris, Scourfield, Sparks and Webb2018).
The lack of association between the ‘Barriers to Housing and Services’ deprivation domain deciles and Ofsted outcomes may raise doubts about the potential of greater service provision. It should be clarified that this measure largely captures physical distance-related barriers to access of universal services (primary schools, GPs surgeries) and amenities (post offices and supermarkets) rather than ‘access’ in a more holistic sense. The conflation of housing quality with these physical distance measures may also be unhelpful. Similarly, the Living Environment deprivation domain includes not only housing condition but air quality and road traffic incidence rates. Notwithstanding their nomenclature, results from models using these domains may tentatively suggest that greater surveillance of children, either through proximity to services commonly engaged with the referral of children to social services or through greater public surveillence approximated by indicators of urbanisation, is not associated with more positive Ofsted outcomes.
Consequences for public services and policy
Some forms of additional ‘supply’ are associated with better service quality, whereas others are not; this may help explain previous findings of no association between public expenditure and quality. Greater investment in social work child protection, funded through safeguarding spending that has been relatively stable since 2010 (Webb and Bywaters, Reference Webb and Bywaters2018), appears to be no guarantee of quality if a great many of the families are living in poverty, and if the local authority does not have adequately resourced family support services. This may reflect ‘failure demand’ in children’s services (Munro, Reference Munro2010; Hood, Reference Hood2015; Hood et al., Reference Hood, Goldacre, Gorin and Bywaters2020a) – demand arising from a failure to address a need earlier in its emergence – when faced with diminished support options and socioeconomic adversity. The ‘revolving front-door’ of children’s social care, characterised in England by rates of re-referrals and repeat child protection plans that can escalate into care entry (Hood et al., Reference Hood, Goldacre, Grant and Jones2016a), highlights the very real consequences of failure demand.
This analysis shows an ‘inverse care law’ at work in children’s services quality, comparable to the inverse care law in access to medical services. The most deprived local communities with the greatest needs are least likely to have access to good quality children’s services. Reducing poverty directly may result in operational benefits for children’s services by addressing demand that could be managed through adequate income, housing, and employment. This can, in return, increase the availability of preventative support. However, research into child welfare inequalities has found that lower regional deprivation may adversely affect equity in intervention rates, particularly where income inequality is high (Webb et al., Reference Webb, Bywaters, Scourfield, McCartan, Bunting, Davidson and Morris2020). As such, policy responses must also carefully address the inequitable distribution of resources across society at multiple levels of geography (Bywaters et al., Reference Bywaters, Brady, Bunting, Daniel, Featherstone, Jones, Morris, Scourfield, Sparks and Webb2018; Webb and Bywaters, Reference Webb and Bywaters2018; Webb et al., Reference Webb, Bywaters, Scourfield, McCartan, Bunting, Davidson and Morris2020).
Access to early help and support when a child’s health or development is at risk is a legal right in England; families and children not getting the support they need when they need it is a signifier of poor-quality services, according to Ofsted and to parents (Gupta and Blumhardt, Reference Gupta and Blumhardt2016; Ofsted, 2020). The fact that local variations in early help expenditure and levels of deprivation are strongly associated with Ofsted outcomes raises serious questions about how inspectors of children’s services contextualise and interrogate their own appraisals. This is apparent in the inspection reports themselves. In Hood et al.’s (Reference Hood, Nilsson and Habibi2019) analysis of sixty Ofsted reports, stratified by deprivation, rating, and rurality, contexts of funding and poverty were absent.
Lastly, there are important implications for social policy and the ‘take over’ of local services in England and Wales (Jones, Reference Jones2019; Kim and Warner, Reference Kim and Warner2021). Ofsted outcomes are not simply benign assessments of quality: ‘inadequate’ judgements can result in wholescale restructuring of local services and their governance (Stevenson, Reference Stevenson2015; Jones, Reference Jones2019; Hernandez, Reference Hernandez2021). If the quality of services is contingent firstly on pre-existing socioeconomic factors and national welfare policy, and, secondly, on sufficient funding for family support services, political decisions which defund services and exacerbate poverty are implicated in causing chains of events that can lead to restructuring of governance and delivery under a pretence of quality improvement (Jones, Reference Jones2019). Given the inequitable distribution of both poverty and funding, this risks creating a two-tier system of children’s services provision with differing forms of governance and public accountability in each: one for more affluent, better-resourced communities, and one for deprived, under-resourced communities.
Strengths and limitations
This study is the first, to our knowledge, to implement a methodological approach that allows for the assessment of predictors of inspection outcomes across multiple frameworks while adjusting for hierarchical dependence of errors. This enabled a greater sample size for the detection of smaller or multiple effects, which is critically important when considering the finite nature of local government data. This is the first study we know of that delineates between preventative and safeguarding expenditure, recognising that not all uses of funding are equal within complex systems like children’s services (Hood, Reference Hood2015). Further, we have attempted to show that the association between deprivation and Ofsted outcomes found in prior research studies is robust across multiple domains, with exceptions among living environment deprivation, barriers to housing and services, and income deprivation affecting older people.
A limitation of this study is its reliance on Ofsted inspection judgements as a measure of service quality. Others have noted Ofsted’s focus on procedure over the experiences and outcomes for children and families in contact with children’s services (Hood et al., Reference Hood, Grant, Jones and Goldacre2016b; La Valle et al., Reference La Valle, Holmes, Gill, Brown, Hart and Barnard2016). The contextually important effects nonetheless prompt further questions about whether the same trends exist for quality as assessed from a child or family perspective.
We remain unable to delineate between types of preventative spending over time and across local authorities using administrative data (Webb and Bywaters, Reference Webb and Bywaters2018). Children’s centres, early intervention programmes, family support, and other services, may be differently associated with service quality. Further, we cannot determine any causal direction in this association – while better funding and less deprivation may result in better Ofsted judgements, it may also be the case that better or worse Ofsted judgements result in different utilisation of funding or lower deprivation.
Conclusions
Our findings illustrate the size of socioeconomic determinants of quality in children’s services, challenging dominant and entrenched narratives in England that neither service funding nor deprivation are indicators of service quality (National Audit Office, 2016, 2019). They also add nuance to existing studies by introducing methods that can adjust for the bias associated with the incorporation of multiple frameworks. Policies that direct resources towards preventative spending and poverty alleviation may create overall benefits in quality across the children’s social care system.
Over the last decade, investment in preventative services has declined significantly and unevenly (Webb and Bywaters, Reference Webb and Bywaters2018; Action for Children, 2020; YMCA, 2020). The number of children living in families experiencing destitution is estimated to have increased by 75 per cent between 2015 and 2019 (Fitzpatrick et al., Reference Fitzpatrick, Bramley, Blenkinsopp, Wood, Sosenko, Littlewood, Johnsen, Watts, Treanor and McIntyre2020). Inequalities in child welfare interventions associated with deprivation have widened (Bennett et al., Reference Bennett, Mason, Schlüter, Wickham, Lai, Alexiou, Barr and Taylor-Robinson2020); this article highlights that the quality of child welfare services may also have suffered as a result. Those most in need of high-quality services to prevent child maltreatment are least likely to have access to them. Much might be learned from policies implemented in response to inequitable access to medical care (Tudor Hart, Reference Tudor Hart1971).
As long as ‘inadequate’ judgements can be used to justify the ‘take over’ of services, failure to acknowledge and address their socioeconomic determinants raises doubts about the appropriateness of any restructuring of public services. Policies that tackle the deep-rooted issues of failure demand and inequality, which so often characterise the child welfare system, are needed. This evidence suggests that investment in financial and material support for families, as well as in family support services, may be effective for addressing these issues and improving service quality.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S147474642200001X.
