Consultation outcome

Annex: Children and Young People’s Services Formula - Detailed response to Question 40

Updated 20 November 2025

1. A summary of the government response to Question 40 on the design of the Children and Young People’s Services Formula (from now on, the children’s formula) is set out in Chapter 12. A number of respondents called for greater transparency on the detail of the formula. This Annex provides a detailed description of the formula and outlines changes we are planning to make in light of the consultation responses.

2. The Fair Funding Review 2.0 sought views on a new formula to inform the distribution of funding to local government for children and family services. The Children and Young People’s Services formula will form part of the Fair Funding Assessment and will be used to allocate grants for children and family services, subject to future decisions.

3. The current funding formula is no longer fit for purpose. It was developed in 2013-14, using even older data and it no longer reflects current levels of need, deprivation, or even the numbers of children, in different parts of the country. In its place we will use the new children’s formula.

4. The children’s formula was completed by LG Futures with academics from the University of Huddersfield and the University of Plymouth in 2020. The methodological approach has been peer reviewed and approved by Professor Anne Vignoles in 2025. Department for Education analysts have updated the 2020 model using the latest available data and an interim version was used to distribute the 2025-26 Children’s Social Care Prevention Grant.

5. In the following response, we outline:

• How the analytical model behind the children’s formula works.
• The substantive criticisms we received to the children’s formula, the changes we have made to the formula in response, and why we chose not to make changes in some areas.
• The next steps, including the implementation of the formula and plans to improve transparency.

The children’s formula analytical model

How the model works

6. The formula works by assessing the relative need of a council compared to others by estimating the likelihood of each child in a local authority interacting with children’s social care services, by either being a child in need, in care, or having ceased being in care. It does this by examining which child and neighbourhood characteristics are most predictive of a child interacting with the children’s social care system.

7. There are eight primary factors that are assessed for their interaction with a child’s probability of entering care.^{[footnote 1]} For example, a child in poor health is more likely to be in care than other children, and an older child is more likely to have ceased being in care than a younger child. So all else being equal, councils with a higher proportion of 14-15 year-old children in poor health will have a higher share of need compared to councils with fewer.

8. The interactions with the system – being a child in need, in care, or having ceased being in care – are weighted to reflect national spend on each of those categories.^{[footnote 2]} Around half of national spend is on children in need, so children in need reflect around half of the council’s need. A council that has more children with a higher likelihood of being children in need will receive a higher assessment of need.

9. Taken together, a council that has more children with characteristics predictive of interacting with the children’s social care system (such as being male, being more deprived, or living in overcrowded housing) will have a higher assessment of need. Likewise, a council which has more children that are more likely to interact with the more expensive parts of the children’s social care system (such as being a child in need or being in care) will have a higher assessment of need.

10. Whilst many consultation respondents disagreed with some of the specific data variables used, during the Department for Education’s engagement with stakeholders most confirmed they were supportive of the overarching principles of the model.

The model’s prediction of need

11. This type of analytical model enables a more robust prediction of relative need:

12. The model uses granular, child-level data to make predictions about need at the level of each individual. The use of child-level data makes the formula more robust in its ability to identify need at local, neighbourhood levels, rather than models which use geographically aggregated data. This was supported by the Institute for Fiscal Studies, who stated that ‘the use of… child-level data in the children and young people’s service formula means that these formulas are likely to be more robust than the other formulas, which have been developed using council-level data.’^{[footnote 3]}

13. The metrics the children’s formula uses to assess whether a child will need support include child age, gender, and eligibility for free school meals. Each metric within the model was only included if it improved the model’s ability to predict need.

14. The children’s formula is designed to be independent of local authority practice. The formula does not take into account local authorities’ historic spending, or how many children they have in care. It aims to estimate underlying levels of need, and not how different local authorities respond to that need. This means that local authorities which have prioritised preventative services and have fewer children in care are not penalised financially.

15. The children’s formula uses the most up to date data, including the most recent data from the National Pupil Database available at the time of modelling, and the 2025 Income Deprivation Affecting Children Index. This ensures that the formula reflects the most up-to-date statistics on child population sizes and evolving local demographics, including deprivation levels, providing a more accurate assessment of need for children and family services.

16. Following careful consideration of the consultation responses to question 40, we are confident that children’s formula offers the most accurate, up-to-date and reliable means of assessing local authorities’ relative need for services.

Response to substantive issues raised

17. A number of specific points were made about the children’s formula and a range of suggestions made for how it could be improved. We considered each proposed change and assessed whether it would improve the accuracy of the formula – only including those where that is the case. These are set out in the sections below.

18. Many of those who disagreed either mentioned directly or referenced analysis from a report by the National Children’s Bureau and Public Alchemy, commissioned by London Councils, which raised a number of concerns with the children’s formula and underpinning model. We have sought to address these concerns, as well as others raised through the consultation, in our response.

Transparency

19. Many of those who disagreed with question 40 sought more transparency of the data used and drivers of need within the model.

20. We have consulted on reforming children and young people’s services funding in two phases. Following the first phase, where we consulted on the high-level principles underpinning the children’s formula, we published a suite of supporting documents. These documents, published at Children and young people’s services formula review in February 2025, aimed to strengthen stakeholder understanding of the formula and provide additional clarity. This included the final report by the original authors of the children’s formula and an independent academic peer review.

21. The government is being as transparent as possible in publishing the formula’s data inputs, which would allow local authorities to see what is driving their need-shares. However, we are not able to publish some elements of that data due to our obligations under data protection legislation, specifically those data from the National Pupil Database.

22. Some respondents raised concerns about transparency in relation to the complexity of the formula, commenting that it makes it difficult to fully understand how the formula works. We acknowledge that the formula is complex – however the inclusion of additional variables is supported by robust testing, which demonstrates that each individual variable improves the model’s predictive power. This results in a model that more accurately directs funding to areas with the greatest need. This issue was considered by Professor Anna Vignoles, published earlier this year, in her peer review of the formula where she concluded that the underpinning methodology of the formula is sound. Given this complexity, the Department for Education will run analyst led sessions before the provisional Local Government Finance Settlement is published to help relevant leads in local authorities understand the formula and improve transparency and understanding of how we are allocating funding.

Recommended variable changes

23. A substantial number of respondents who disagreed with the children’s formula made suggestions of data and variables that should be included or taken out of the model. All the variables in the children’s formula have been included specifically because they strengthen the assessment of relative need.

24. The Department for Education have reviewed each of the recommended changes to variables and underpinning datasets submitted through the consultation.

Local authority-level variables

25. We acknowledge that some known indicators of need for children’s social care services, such as domestic abuse and mental health are not included in the children’s formula. This is because these indicators are only available at the local authority level (or police/NHS equivalent level). Analysis by LG Futures concluded that the inclusion of local authority level data does not improve the model’s overall ability to accurately predict need for services. To preserve the accuracy of the model, we therefore chose to only use variables recorded at the Lower layer Super Output Area level, which offer greater granularity and consistency in terms of data collection. This strength of detail and nuance at the neighbourhood level ensures the model better identifies pockets of need than it would if we used local authority-level aggregate data.

26. Formulas developed using individual child-level data are more robust than other formulas, which have been developed using local authority-level data. This is because, any data derived from local authority-level service usage will be affected by the level of funding available to different councils at the time (the data were collected) and by local authorities’ policy and practice. As well as not properly capturing levels of underlying need consistently across the country, models based on local authority-level data could also incentivise specific local policies as these would affect funding allocations.

Housing costs and deprivation

27. Beyond suggestions of local authority-level data that could be used in the model, one of the main concerns cited was that deprivation metrics used within the children’s formula don’t account for the impact of housing costs. The updated 2025 Income Deprivation Affecting Children Index dataset, which we will use for the children’s formula, which reflects housing costs. This will ensure that relative needs shares will reflect the impact of high housing costs on families in certain areas.

28\ The updated deprivation data is very highly correlated with the low parental qualifications variable data in the model. In line with modelling best practice (and adhered to in the LG Futures report), it has been removed from the model, with testing showing this improves the model’s ability to predict need and reduces its complexity.

Free School Meals and Universal Credit

29. Many of the same respondents also queried the Free School Meals element of the model – the main concern raised was that there is likely a significant undercounting of those accessing Free School Meals. Respondents suggested that children living in households that are in receipt of Universal Credit would be a better data source – and would align with changes to Free School Meals eligibility. This was explored in the LG Futures report and subsequent peer review.

30. School funding is affected by the number of children eligible for Free School Meals and we know that schools make considerable efforts to identify eligibility, even where children will not take up the actual provision of Free School Meals.

31. With regards Universal Credit data – all benefits claimants are now included in the new 2025 Income Deprivation Affecting Children Index, including those claiming Universal Credit. We are therefore confident that Free School Meals data, combined with the Universal Credit data captured in our income deprivation measure, are the most robust and accurate measures currently available to contribute to the assessment of need. As with all variables, we are committed to regularly testing the model as and when new data are published and available and updating it when that would improve accuracy.

Education, Health and Care Plans

32. A number of responses recommended including data on Education, Health and Care Plans collected by local authorities as a variable to account for representation of children with Special Education Needs and Disabilities requiring children’s services. This was explored in the LG Futures report. While disability is a recognised driver of demand for children’s services, Education, Health and Care Plan rates are not an entirely objective measure. They can be influenced by levels of parental advocacy, local assessment practices and thresholds for support – meaning that Education, Health and Care Plan rates may not accurately reflect levels of underlying need. Including Education, Health and Care Plan rates as a variable within the formula could therefore create a risk of perverse financial incentives, if higher Education, Health and Care Plan rates are linked to increased need shares and funding allocations. Although Education, Health and Care Plan data was considered for inclusion in the formula, it was ultimately excluded due to recognised data variations in recording and reporting practices.

33. The “Proportion of Children in Poor Health” metric (from now on ‘child health metric’) is a more reliable proxy for child disability than the number of children with an Education, Health and Care Plan. Whilst the child health metric is self-reported, large-scale responses to a standardised question on perceived levels of health is a credible and robust data source. Furthermore, testing confirmed it improved the children’s model’s ability to predict need. The Department for Education have therefore concluded that the “Proportion of Children in Poor Health” metric should be included as a more robust and reliable proxy for child disability.

Overcrowding

34. A number of respondents recommended an alternative metric be used to assess overcrowding beyond the “Proportion of overcrowded households” data. Following further testing and analysis, the Department for Education agree with this and have replaced the original metric with “Proportion of households with dependent children which are overcrowded”. This change reflects the fact that overcrowding is a more relevant indicator of need for children’s social care services when specifically considering households with dependent children.

Unaccompanied Asylum-Seeking Children

35. Some respondents for areas with high numbers of Unaccompanied Asylum-Seeking Children called for data on Unaccompanied Asylum-Seeking Children to be included in the formula model. While the formula cannot reliably predict the costs of Unaccompanied Asylum-Seeking Children to local authorities, the characteristics of any those children who are within the care system will have been accounted for. Additionally, the Home Office provides an additional top-up grant to local authorities to support their costs in caring for Unaccompanied Asylum-Seeking Children. As such the Department for Education are content additional data to account for Unaccompanied Asylum-Seeking Children specifically is not required.

Crime rate

36. Respondents also suggested that crime data be included in the model to reflect need in high-crime rate areas. LG Futures and academics who developed the children’s formula assessed a range of other variables for potential inclusion. In evaluating crime data for potential inclusion, the authors found that much of this data was only available at local authority level and lacked the granularity required to improve the model’s ability to predict need. Moreover, crime statistics at lower levels of geography are collected and reported by individual police forces, each with its own recording practices – this introduces considerable variability and inconsistency across areas, making it difficult to accurately compare crime rate data. As a result, crime datasets were excluded as they failed to enhance the model and were unable to capture localised pockets of increased deprivation, criminal activity and other known key drivers of need.

Funding

37. The remaining substantive theme from consultation respondents was around the level of funding for children’s services and concern as to whether the children’s formula inadvertently discourages investment in preventative services (i.e. that those with fewer children in care as a result of investment in preventative services might receive a lower share of available funding). The children’s formula does not penalise local authorities which choose to invest in early intervention, as it does not predict need based on local authority spending and/or decision-making (or that of other agencies such as the court service).^{[footnote 4]} This means that the number of children categorised as Children Looked After or Children in Need by a local authority, or the past spending patterns that result from these categorisations, have no influence on their funding allocations. Instead, the formula works by assessing anticipated future demand for services based on child and neighbourhood characteristics which are established, well-evidenced predictors of need for children’s social care services.

Next steps

38. The Department for Education will hold a series of transparency webinars with the sector before the provisional Local Government Finance Settlement to improve understanding of the formula and the drivers of change in need-share.

39. The Department for Education will modify the children’s formula to respond to consultation feedback to use the 2025 Income Deprivation Affecting Children Index which will reflect housing costs, to remove the parental qualifications metric, and to change our overcrowding metric to “Proportion of overcrowded households with dependent children”. Following those changes, the Department for Education believe that the children’s formula will produce need-shares which accurately reflect need across the country.

40. We will use this final version of the formula as part of the upcoming multi-year Local Government Finance Settlement, subject to final consultation on the provisional Settlement.

1. These are: Sex of child (categorised as male or female); Age of child; Eligibility for free school meals (FSM) on date of the census; Socio-economic deprivation level in child’s LSOA (as measured by the IDACI); Proportion of children in child’s LSOA with poor health; Proportion of overcrowded households in child’s LSOA; Population density (measured in persons per km2) in child’s LSOA; Travel time from LSOA centroid to nearest town centre (mins). ↩

2. The resource weightings for interactions are: Child in Need - 49.8%; Child Looked After - 39.4%; Care Leaver - 10.8%. ↩

3. Response to ‘Fair Funding Review 2.0’ consultation by MHCLG. ↩

4. Children and Young People’s Services Formula Review (p.11). ↩