Communities and Local Government Committee. Reforming Local Authority Needs Assessment. Paper 1 Simplifying the Needs Assessment Formula

Transcription

1 LG FUTURES Communities and Local Government Committee Reforming Local Authority Needs Assessment Paper 1 Simplifying the Needs Assessment Formula October 2017 FINANCE WITH VISION LG Futures Ltd., Marlowe House, Watling Street, Hockliffe, Leighton Buzzard LU7 9LS T Incorporated in England and Wales under registration number: VAT registration number:

2 Contents Contents 2 1. Key Points Context: an overview of the current needs assessment formula... 5 Table 1: Relative Needs Formula Breakdown by service block 2013/ Methodology Results 11 Social Services for Older People Table 2: Impact of excluding indicators Social Services for Older People Children s Services Children s Social Care Environmental, Protective and Cultural Services (District Level) Environmental, Protective and Cultural Services (County Level) Highway Maintenance Other needs assessment formulae Conclusions... 35

3 1. Key Points The Department for Communities and Local Government (DCLG) is currently undertaking a Fair Funding Review of local authority funding allocations. This review was originally initiated alongside plans for the introduction of 100% business rates retention, which was planned for 2019/20. Whilst indications from the government are that there will be movement towards 100% nationally, it is likely that this will not take place in the timeframe originally intended and that there will be no stepped change to the scheme as soon as 2019/20 (i.e. from 50% to 100% retention). The Fair Funding Review is likely to determine the future level of funding for each local authority in England. This funding could be delivered in form of 100% retained business rates (with growth incentives), in the form of central government grants, or a combination of the two as is currently the case. The Fair Funding Review involves an assessment of each local authority s relative needs; that is, the relative funding required to deliver a comparable level of local services as in other parts of England. Two principles underpinning the review are simplicity and transparency. The intention is to arrive at a measurement of relative needs that is less complex than the current system, and for councils and the public to be able to understand the resulting funding allocations. 1 This paper demonstrates a potential method for simplifying the needs assessment formula. It does this by quantifying the trade-off between a formula s simplicity (using fewer indicators) and its precision (in terms of the ability to replicate DCLG s original formula). No judgement is made as to whether the original formula represents a fair assessment of local authorities actual need. It simply assumed that, where the trade-off between a formula s simplicity and precision is sufficiently small, policymakers may prefer a less complex formula in the interests of transparency. The analysis takes the existing Relative Needs Formula (RNF) as its starting point, which consists of over 120 socio-economic, demographic and other indicators. The formula is used to estimate each local authority s relative spending requirements for 16 services. While the paper focuses on the current RNF, the same simplification method could be applied to any future needs assessment formula. An outline of the methodology is as follows. For each service s formula, we drop the indicator whose exclusion would cause the smallest change in local authorities assessed needs per head. The remaining indicators are then reweighted so as to minimise the overall change in needs per head. This step is then repeated, dropping 1 DCLG and Local Government Association, Discussion 1: The Progress of the Fair Funding Review, from the Business Rates Retention Regional Consultation Event, FINANCE WITH VISION 3

4 L the next indicator (in addition to the first) whose exclusion would have the smallest incremental impact on needs per head. The remaining indicators are again reweighted. This process continues until all indicators have been excluded from the formula. Each time an indicator is removed from the formula, we measure the resulting impact on local authorities needs. Impact is defined as the absolute (±) percentage change in each authority s assessed needs per head, relative to the original formula. This is used to illustrate the trade-off between (i) improving each formula s simplicity and (ii) the formula s ability to replicate the original pattern of assessed needs. The optimal trade-off between a formula s simplicity and precision is a policy choice beyond the scope of this paper. Selecting an arbitrary cut-off point, where half of local authorities experience an impact on their assessed needs of within ±1.0%, it was found that the following simplifications could be achieved: Social Services for Older People: reduced from 6 to 4 indicators Children s Social Care: reduced from 9 to 6 indicators District-level EPCS: reduced from 9 to 3 indicators County-level EPCS: reduced from 8 to 4 indicators Highway Maintenance: reduced from 8 to 5 indicators In addition to reducing the number of indicators used, the paper identifies two further ways in which the needs assessment formula could be simplified. First, it would be possible to simplify the structure of the formulae for some services, reducing the complexity of the calculations while having only a limited effect on the pattern of assessed needs. This is the case with Social Services for Older People and Children s Social Care. Second, the needs assessment model could be simplified by removing or consolidating formulae for individual services that account for a small share of total assessed needs, particularly where the results of these formulae are highly correlated with those of other services. Overall, this paper shows that a simplified needs assessment formula, based on a smaller number of indicators, could achieve outcomes that approximated those of a more complex system. It demonstrates a workable methodology that could be used to reduce the complexity of any future needs assessment formula, while minimising the impact on local authorities assessed levels of relative need, and making this trade-off clear to policymakers. FINANCE WITH VISION 4

5 2. Context: an overview of the current needs assessment formula 2.1. DCLG is currently undertaking a Fair Funding Review of local authority funding allocations. This review was originally initiated alongside plans for the introduction of 100% business rates retention. Whilst indications from the government are that there will be movement towards 100% nationally, it is likely that this will not take place in the timeframe originally intended and that there will be no stepped change to the scheme as soon as 2019/20 (i.e. from 50% to 100% retention) Potentially, this funding could be delivered in the form of 100% retained business rates. The funding review would determine each local authority s baseline funding level. At the outset of the scheme, each authority s baseline funding level and its business rate income would be equalised, either through a tariff (a deduction, should its business rate income exceed its funding baseline) or a top up (an additional payment, should its business rate income fall short of its funding baseline). In future years, authorities would retain any growth in their business rate receipts, providing incentives for local authorities to support growth in the business tax base To a large extent, the review of local government funding is independent of a 100% business rates retention scheme. In simple terms, the review of local funding will determine how much local authorities should receive in future, while decisions regarding business rates will determine how the funding is delivered. This could be delivered in the form of 100% retained business rates, as central government grants, or a combination of the two, as is presently the case This research identifies ways in which the current funding formula could be simplified. The current formula is highly complex and poorly understood by even experienced local government officials. Simplifying the formula would greatly enhance its transparency and accessibility to local authority practitioners, elected officials and members of the public The research considers two aspects of the funding formula: The underlying needs assessment formula. This is the formula that is used to estimate the relative funding requirement for each local authority in England. Simplifying the underlying needs assessment is the subject of this paper (Paper 1). The wider funding model. Currently, funding is not simply allocated in proportion to each local authority s assessed needs. There is also a deduction in funding to reflect 2 Full details of how the 100% retained business rates scheme would work have not yet been announced. The description provided here is inferred by LG Futures from government working papers, consultations and the operation of the current 50% business rates retention scheme. FINANCE WITH VISION 5

6 L their assumed ability to fund services through council tax revenue. Simplifying this wider funding model is the subject of the second paper in this series (Paper 2) This paper focuses on the underlying needs assessment formula. Specifically, it identifies potential simplifications to the current Relative Needs Formula (RNF) The RNF was last updated in 2013/14, at the outset of the current Business Rates Retention Scheme. The formula was used to allocate the majority of funding received by local authorities as part of the scheme, delivered in the form of central government grants and a share of 50% of locally retained business rates The RNF is intended to measure the relative cost of providing comparable services in each local authority in England. The formula is highly complex, combining over 120 socioeconomic, demographic, geographic, and other indicators. For each service, the general structure of the formula is as follows: A basic amount per client: The client group is specific to each service, and is typically a subset of the population. For example, for Children s Social Care, the client group is the projected number of residents ages 0 to 17. An additional amount per client: These reflect differences in spending pressures, using proxy indicators that include deprivation (such as the proportion of residents receiving benefits), socioeconomic variables (such as qualification levels, or the proportion of older people living alone) and geography (such as density and sparsity). Each indicator is assigned a weighting, based on its historical statistical relationship with expenditure or using judgement. The Area Cost Adjustment (ACA): This adjusts each authority s assessed needs to reflect the costs of providing services in different parts of England, due to variations in local wage and salary costs The three components above are multiplied by the number of clients, giving the total relative needs for each authority. This is a purely relative measure, as opposed to a monetary ( terms) amount for each authority The RNF estimates needs for 16 separate services. These are listed in the table below. At the England level, each service is assigned a control total, which determines its relative importance when estimating local authorities relative needs. Social Services for Older People has the largest control total, accounting for 19.1% of total assessed needs, followed by district-level Environmental, Protective and Cultural Services (EPCS) (15.1%) and Social Services for Younger Adults (12.9%). FINANCE WITH VISION 6

7 Table 1: Relative Needs Formula Breakdown by service block 2013/14 Service Block Sub -block Control total Youth & Community 1.2% Children s Services Central Education Functions 4.9% Children s Social Care 9.7% Adult Social Services Older People 19.1% Younger Adults 12.9% Police 11.7% Fire & Rescue 4.3% Highways & Maintenance 2.9% District-level EPCS 15.1% County-level EPCS 9.0% Concessionary Travel 1.7% Environmental, Protective & Cultural Services (EPCS) Flood Defence 0.1% Continuing EA Levies 0.0% Coast Protection 0.0% Fixed Costs 0.2% Capital Financing 7.3% Total 100.0% The paper demonstrates that a simplified needs assessment formula, based on a smaller number of indicators, could achieve outcomes that approximated those of the more complex RNF. The analysis takes the current RNF as its starting point, rather than developing a new formula from a zero base. No assumption is made as to whether the original RNF reflects a fair assessment of local authority s relative needs. The analysis illustrates the trade-off between simplifying the formula and its ability to replicate the original outcomes. It demonstrates a methodology that could be applied to a future needs assessment formula, to assess whether reducing its complexity would have a material impact on its ability to explain differences in authorities needs. FINANCE WITH VISION 7

8 L 3. Methodology (i) Scope of the analysis 3.1. The analysis considered those services where the needs assessment formula appeared to offer scope for simplification, based on the number of indicators used. The following eight services were considered: Children s Services Youth & Community Children s Services Local Authority Central Education Functions Children s Services Children s Social Care Social Services for Older People Social Services for Younger Adults District-level EPCS County-level EPCS Highway Maintenance 3.2. The needs assessment formulae for emergency services (Police and Fire & Rescue) were outside of the scope of the project. Of the remaining services, these either consisted of only a single indicator (Concessionary Travel, Capital Financing and Fixed Costs) or accounted for a very small share of total needs (Flood Defence, Continuing Environmental Agency Levies, and Coast Protection) The analysis also considered the allocation formula for the Supporting People grant. This large grant, worth 1.6 billion in 2013/14, accounted for the majority of the Tailored Grants component of the Formula Funding model (shown in the diagram above). These grants were part of Formula Funding, but were allocated using their own bespoke criteria, rather than the Relative Needs Formula. (ii) Method used for simplifying the formulae 3.4. For each service, the following steps were used to simplify the needs assessment formula: The first step was to remove the effect of the Area Cost Adjustment from each authority s current level of assessed need. It was assumed that the ACA would be added back on to any simplified formula; i.e. that it would not be excluded. The second step was to express each authority s total needs as an amount per head. Where possible, we used the same client group as was used in the original formula. For example, for Children s Social Care, total assessed need per resident aged 0 to 17 was used. The third step was to assign a weight to each indicator in the simplified model. The indicators were weighted so as to minimise the difference in local authorities FINANCE WITH VISION 8

9 assessed needs per head relative to the original formula. This was done using linear regression analysis, in which each indicator was weighted so as to best explain the local authorities original needs per head. The fourth step was to exclude a single indicator from the formula. The indicator to be dropped was the one whose exclusion would have the smallest impact on the formula s outcomes. In other words, the approach was to exclude the indicator which would result in the smallest overall change in local authorities assessed needs, relative to the original formula. The remaining indicators were re-weighted so as to minimise the difference between local authorities new and original needs per head. The statistical method used to select the indicator to be dropped at each step is described in more detail below (at para 3.5). This final step was then repeated, selecting the next indicator to be excluded from the model. This was the indicator whose exclusion, in addition to the first indicator, would have the smallest marginal effect on the formula s outcomes, relative to the original formula. The remaining indicators were again reweighted. This process continued, until all indicators had been excluded from the formula A statistical method was used to select the order in which indicators were excluded from the model. As described above, at each step, we dropped the indicator whose exclusion would cause the smallest overall change in local authorities assessed needs per head. In technical terms, this was the indicator whose exclusion had the least impact on the model s goodness-of-fit. The goodness-of-fit (or R squared) measures how much of the original variance in local authorities needs per head can be explained by the indicators that remained in the model. This ranges from 0% to 100%. Each time an indicator was dropped, it was not entered back in. This avoided the need to consider every potential combination of indicators. Each time an indicator was dropped, the remaining indicators were reweighted using a linear regression To illustrate, assume that the original indicators were A, B and C. To determine which of these indicators would be dropped first, three models were considered: one excluding A, one excluding B, and one excluding C. Assume the goodness-of-fit for these simplified models were 90%, 80% and 70%, respectively. In this case, indicator A would have been excluded, as this resulted in the simplified model with the highest goodness-of-fit (90%). The process would then be repeated, to determine which of the remaining indicators, B or C, would be dropped next. 3 An ordinary least squares (OLS) regression was used to reweight the remaining indicators. The dependent variable was local authorities original needs per head, and the independent variables were the indicators that had not yet been dropped from the formula. OLS regression minimises the sum of squared residuals; in this case, the residuals were the difference between each authority s needs per head under the simplified formula and the original formula. Other criteria could have been used to re-weight the remaining indicators, but the sum of squared residuals had the advantage of being computationally simple. FINANCE WITH VISION 9

10 L (iii) Summarising the impact of simplifying the formulae 3.7. The section above explains the rules used to determine the order in which indicators were dropped from the simplified model. At each step, we dropped the indicator that made the smallest contribution to the model s ability to explain the original pattern of assessed needs per head In order to illustrate the trade-off between simplicity and precision, the impact of each simplified formula was measured. This is defined as the absolute (±) percentage change in local authorities needs per head, relative to DCLG s original value. Dropping each indicator simplifies the formula, but results in assessed needs that deviate from the value as originally estimated by DCLG. As described above, we take this original formula as a given, and make no judgment as to whether it is a fair and accurate estimate of local authorities relative needs There are at least three reasons why dropping an indicator may have a limited impact on assessed needs, potentially offering a way to simplify the existing formulae without having a major impact on the current pattern of need: Where DCLG assigns weights to indicators based on judgement, some indicators could have been assigned relatively little weight compared to others. Where weightings are based on the statistical relationship between the indicator and past expenditure, the relationship may be statistically significant (i.e. unlikely to be due to chance), but not practically significant in terms of the magnitude of the relationship. An indicator could be highly correlated with other indicators in the formula, meaning that the role it plays in determining assessed need is partially redundant in the presence of other indicators The following sections describe the results of the analysis. FINANCE WITH VISION 10

11 4. Results Social Services for Older People 4.1. DCLG s original formula for Older People s RNF is based on the following six indicators: 4 Population sparsity among older residents (SparsityOlderAdults) Older adults living alone (LivingAlone) Older adults receiving Attendance Allowance (AttendanceAllowance) Older adults living in rented accommodation (Renting) The ratio of older residents aged 90+ (OldAgeRatio) Older adults receiving certain income and employment related benefits (BenefitsIncomeOlder1) 4.2. The following table measures the impact of excluding indicators from the formula. The impact is measured in terms of the absolute (±) percentage change in local authorities assessed needs per head, relative to the original needs assessment formula. The table shows the cumulative impact of progressively removing indicators from the formula. The top row of the table shows the impact when all of the original indicators are included in the formula; the bottom row shows the impact when all the indicators have been excluded. Table 2: Impact of excluding indicators Social Services for Older People Indicators dropped from the model [None dropped] No. of indicators in model 6 Impact Absolute (±) change from original needs per head 90th Median Maximum percentile 0.7% 1.8% 3.5% Goodness of fit (R squared) 99.8% SparsityOlderAdults 5 0.7% 1.9% 3.2% 99.7% LivingAlone 4 0.8% 1.9% 7.3% 99.6% AttendanceAllowance 3 2.1% 4.5% 9.6% 98.3% Renting 2 2.7% 5.5% 20.3% 96.9% OldAgeRatio 1 3.0% 8.5% 27.2% 93.8% BenefitsIncomeOIder1 [All dropped] % 33.5% 74.6% 0.0% Dependent variable: 5 Units of RNF divided by projected household and supported residents aged 65 and over. 4 Unless otherwise stated, all indicators in this report are proportions of the relevant client group; for example, as a proportion of the relevant population aged 65 and over in the case of Social Services for Older People. 5 The dependent variable is the variable that is being explained by the indicators in the linear regression models. It is calculated by LG Futures. Unless otherwise stated, the denominator is the same as the client group used in DCLG s original needs assessment formula. FINANCE WITH VISION 11

12 4.3. The impact of simplifying the formula is summarised at the England level by values that correspond to the median, 90 th percentile and maximum. These can be interpreted as follows: Half of all local authorities would experience an impact equal to or less than the median; 90% of local authorities would experience an impact equal to or less than the 90 th percentile; and All local authorities would experience an impact equal to or less than the maximum For example, dropping SparsityOlderAdults and LivingAlone, and re-weighting the remaining four indicators, would have a median absolute impact of ±0.8% (shown in the third row of the table above). This means that half of local authorities would see their needs per head change by ±0.8% or less. Similarly, 90% of authorities would change by ±1.9% or less, while the maximum change for any authority would be ±7.3% The table above includes the goodness-of-fit of each simplified model. The baseline model, which contains all six original indicators, explains 99.8% of the variation in DCLG s original model. 6 This is shown in the top row of the table. If the model was simplified to include only five indicators, dropping the sparsity indicator (SparsityOlderAdults) was found to result in the next-highest goodness-of-fit, of 99.7%. Further simplifying the model, having already dropped sparsity, it was found that excluding older adults living alone (LivingAlone) gave the next highest goodness-of-fit, of 99.6%. Impact by local authority characteristic 4.6. Simplifying the needs assessment formula would have different impacts on different types of authority. Here, we summarise the impact based on local authorities administrative type (e.g. county vs. shire district), its rural-urban characteristics, and its level of deprivation. Details on these groupings, and the definitions used, are presented in Annex 1. 6 It does not explain 100% of the variance because DCLG s original formula is not based on a linear combination of these indicators. Regression analysis assumes a linear combination of indicators, e.g. ax + by + cz, where X, Y and Z are indicators and a, b and c are constants. In cases where DCLG s original formula was also a linear combination of indicators, the simplified model would exactly replicated the original formula. This was not the case where the original formula included more complex interactions between variables, e.g. ax * (by + cz), as was the case for Older Adult s Social Care. FINANCE WITH VISION 12

13 L 4.7. Presenting the results for each simplified model, for every service, would require an excessive amount of detail. For each service, a single, simplified model was therefore selected in order to provide a more detailed description of the impact on different types of authorities. For the purpose of illustration, we selected the simplified model which excluded the greatest number of indicators, while having a median absolute impact on assessed needs per head of ±1.0% or less (after rounding). We emphasise that this rule is ultimately arbitrary for the purposes of presentation, and any other cut-off point could have been used For Social Services for Older People, two variables could be excluded from this formula, while keeping median impact within ±1.0% of original assessed needs per head. These were SparsityOlderAdults and LivingAlone The following charts illustrate the maximum, minimum and median impact for local authorities, based on (i) their administrative type, (ii) their rural-urban category, derived from Defra s Rural-Urban Classifications, and (iii) their levels of income and employment deprivation, based on the English Indices of Deprivation (see Annex 1 for more details). For each group of authorities, the circles illustrate the median impact. The impact for 90% of authorities falls within the dark blue band, while the impact for all authorities would fall within the light blue band (in other words, the light blue band includes the maximum and minimum impact on any authority) As can be seen, the impact on all authorities would range between +3.6% and -7.3%. The impact broken down by local authority characteristic would be as follows: Metropolitan districts would see the largest median increase in assessed needs per head (+0.6%), while London boroughs would see the largest median decrease (-0.8%). London boroughs would experience the widest range of impacts (from +0.7% to -7.3%), though 90% of boroughs would fall within a much narrower band than this (as illustrated by the dark blue band). The most urban group of authorities would experience the largest median decrease (of -0.3%), while the second-most urban group would see the largest increase (+0.4%). The most urban category would experience the widest impact on assessed need; though, for 90% of these authorities, the change would be around -2% or higher. The most deprived quarter of authorities would experience the largest positive median impact (0.5% increase) while the third most-deprived quarter would see the largest median decrease (-0.6%). The least deprived quarter of authorities would experience the largest range of impacts Annex 2 reports the median impact, by local authority characteristic, when fewer or more indicators were excluded from the formula. FINANCE WITH VISION 13

14 Chart 1: Impact of excluding selected indicators from Older People s formula By administrative type 6% 3.6% 4% 3.1% 2% 0% -2% -4% -6% -1.9% 0.1% 0 0.6% -1.8% 0.7% 0-0.8% 1.4% -2.9% 0.0% 90% of authorities Maximum 0 Median Minimum Indicators excluded: SparsityOlderAdults LivingAlone -8% Unitary Metropolitan District -7.3% London Borough County By rural / urban category 6% 4% 3.6% 3.1% 2% 1.0% 1.4% 0% ----o o~ % % -0.3% 0.0% -2% -2.0% -1.7% -4% -2.9% -6% -8% -7.3% Major Urban Urban with City Urban with Predominantly and Town Significant Rural Rural By deprivation level 6% 4% 2% 0% -2% -4% 3.6% 3.4% 1.8% 1.6% 0.5% 0.0% ---0~------o:, ~o~-- o -0.6% -0.1% -2.1% -2.9% -2.5% -6% -8% -7.3% Most deprived Second most Third most Least deprived 25% deprived 25% deprived 25% 25% FINANCE WITH VISION 14

15 Simplifying the formula structure In addition to simplifying the Social Services for Older People formula by using fewer indicators, the structure of the formula would also be greatly simplified. The following formula is based on the spreadsheet calculations currently used to determine each authority s assessed needs per head (rounded to three decimal places): Formula 1: Original Relative Needs Formula (per head) for Older People ( OldAgeRatio , AttendanceAllowance Renting LivingAlone BenefitsIncomeOlder1, ) LowIncomeAdjustment SparsityAdjustment ACA Where: LowIncomeAdjustment 0.120, BenefitsIncomeOlder = (1, ) ACA And: SparsityOlderAdults SparsityAdjustment = ( ) In contrast, the following simplified formula is one in which two indicators had been dropped (LivingAlone and SparsityOlderAdults), the impacts of which were described above: Formula 2: Simplified Relative Needs Formula (per head) for Older People (, BenefitsIncomeOlder OldAgeRatio AttendanceAllowance Renting) ACA This second formula is less complex both because it contains fewer indicators and because it is based on a simple linear structure, in which the impact of each indicator is added together, unlike the original formula in which some of the terms were multiplied together. Note that the Area Cost Adjustment (ACA) has been added back onto the simplified formula. FINANCE WITH VISION 15

16 L Children s Services Children s Social Care For Children s Services, DCLG s needs assessment formula has three sub-blocks: Social Care, Youth & Community, and Local Authority Central Education Functions. For Social Care, the original formula is made up of nine indicators: People in mixed ethnic groups (EthnicityMixed) People in other ethnic groups (EthnicityOther) People aged whose highest qualification attained was Level 1 or 2 (Qualification12) People aged whose highest qualification attained was Level 4 or 5 (Qualification45) People aged receiving certain income and employment related benefits (BenefitsIncome18to64) Females aged looking after home and/or family (FemalesHomeFamily) Children without good health (ChildPoorHealth) Children in black ethnic groups (ChildEthnicityBlack) Children in out-of-work families receiving Child Tax Credit (ChildTaxCredit) The impact of progressively excluding these indicators from the needs assessment formula is presented in the table below. FINANCE WITH VISION 16

17 Table 3: Impact of excluding indicators Children s Social Care Impact Goodness Indicators dropped from Absolute (±) change from original needs per head of fit the model 90th (R Median Maximum squared) percentile [None dropped] 9 0.5% 1.2% 3.3% >99.9% EthnicityMixed 8 0.5% 1.4% 3.2% 99.9% EthnicityOther 7 0.6% 1.7% 3.8% 99.9% Qualification % 2.6% 4.4% 99.8% Qualification % 2.5% 4.2% 99.8% BenefitsIncome18to % 3.3% 4.5% 99.7% FemalesHomeFamily 3 1.7% 3.6% 9.3% 99.4% ChildPoorHealth 2 2.4% 5.2% 18.7% 98.9% ChildEthnicityBlack 1 5.2% 13.4% 21.4% 90.3% No. of indicators in model ChildTaxCredit % 61.6% 124.4% 0.0% [All dropped] Dependent variable: Units of RNF divided by projected population aged Up to three indicators could be dropped while keeping the median impact within ±1.0% of the original assessed needs per head. These were the two ethnicity indicators (EthnicityMixed and EthnicityOther) and a qualification indicator (Qualification12). After excluding these three indicators, the simplified model would be able to explain 99.8% of the variance in the local authorities original assessed needs per head As with Social Services for Older People, these simplified formulae would also make the structure of the formula less complex. At present, some of the indicators are multiplied together as part of the Foster Cost Adjustment, rather than being a simple linear combination of indicators. The original and simplified formulae are compared below (shown to three decimal places). Formula 3: Original Relative Needs Formula (per head) for Children s Social Care ( ChildPoorHealth BenefitsIncome18to ChildTaxCredit ChildEthnicityBlack, ) FosterCostAdjustment ACA Where: FostCostAdjustment = (( EthnicityOther EthnicityMixed Qualification Qualification FemalesHomeFamily, 5.102)/ ) + 0.8) FINANCE WITH VISION 17

18 L Formula 4: Simplified Relative Needs Formula (per head) for Children s Social Care (, ChildPoorHealth BenefitsIncome18to ChildTaxCredit ChildEthnicityBlack Qualifications FemalesHomeFamily) ACA The simplified formula uses fewer indicators, and can be expressed as a single linear equation, eliminating the need for a separate Foster Cost Adjustment and the associated, additional calculations. Impact by local authority characteristic Moving to a simplified formula which excluded EthnicityMixed, EthnicityOther and Qualification12, the impact on all authorities would range between +4.4% and -4.3%. This is broken down by local authority characteristic below. The changes result in the largest median increase for counties, predominantly rural authorities, and least deprived quarter of authorities Annex 2 reports the median impact, by local authority characteristic, when fewer or more indicators were excluded from the needs assessment formula. FINANCE WITH VISION 18

19 Chart 2: Impact of excluding selected indicators from Children s Social Care formula By administrative type 5% 4.4% 4.0% 4% 3.2% 3.1% 3% 2% 1% 0.0% 0 0.8% 0% -0.1% 0-0.4% -1% -2% -1.2% -0.1% 0-0.6% % of authorities Maximum 0 Median Minimum Indicators excluded: EthnicityMixed EthnicityOther -3% Qualification12-3.0% -4% -3.1% -5% -4.3% Unitary Metropolitan London County District Borough By rural / urban category 5% 4% 3% 2% 1% 0% -1% -2% -3% -4% -5% 4.4% 4.0% 3.1% 0.5% 2.3% -0.4% 1.1% -2.4% -3.1% -4.3% Major Urban Urban with City Urban with Predominantly and Town Significant Rural Rural By deprivation level 5% 4.4% 4% 3.5% 2.8% 3% 1.9% 2% 1% 0 1.1% 0.2% 0% -0.1% 0-0.5% -1% -2% -3% -2.4% -2.4% -2.0% -4% -5% -4.3% Most deprived Second most Third most Least deprived 25% deprived 25% deprived 25% 25% FINANCE WITH VISION 19

20 L Environmental, Protective and Cultural Services (District Level) Environmental, Protective and Cultural Services (EPCS) encompasses a wide range of services provided by local government. At the district level, these include council tax collection, planning, building regulations, economic development, parking, museums and galleries, recreation, and waste collection At the district level, the original EPCS formula is comprised of nine indicators: Older adults receiving certain income and employment related benefits (BenefitsIncomeOlder2) Unemployment related benefit claimants (BenefitsUnemployed) Incapacity Benefit and Severe Disablement Allowance (BenefitsDisability) Country of birth of residents (CountryBirth) Day visitors (DayVisitors) Net in-commuters (InCommuters) People receiving certain income and employment related benefits (BenefitsIncomeAll) Population sparsity (Sparsity) Population density (Density) This EPCS formula appears to offer the greatest scope for simplification. It would be possible to exclude six of the original nine indicators while keeping the median impact within ±1.0% of original needs per head. These indicators include benefits for older adults (BenefitsincomeOlder2), unemployment-related benefits (BenefitsUnemployed), disability-related benefits (BenefitsDisability), residents country of birth (CountryBirth), as well as the number of day visitors (DayVisitors) and net incommuters (InCommuters) Unlike for Social Services for Older People and Children s Social Care, it would not be possible to simplify the structure of the formula (i.e. in addition to reducing the number of indicators used). This also applies to all the other services discussed below. FINANCE WITH VISION 20

21 Table 4: Impact of excluding indicators District-level EPCS Impact Goodness Indicators dropped from Absolute (±) change from original needs per head of fit the model 90th (R Median Maximum squared) percentile [None dropped] 9 0.0% 0.0% 0.0% 100.0% BenefitsIncomeOIder % 0.3% 0.7% >99.9% BenefitsUnemployed 7 0.1% 0.3% 0.9% >99.9% BenefitsDisability 6 0.2% 0.6% 1.2% >99.9% CountryBirth 5 0.3% 0.8% 2.9% 99.9% DayVisitors 4 0.4% 1.1% 3.0% 99.8% InCommuters 3 0.6% 2.1% 16.0% 97.7% BenefitsIncomeAll 2 3.5% 8.2% 14.9% 90.0% Sparsity 1 5.3% 13.3% 26.2% 75.2% No. of indicators in model Density 0 6.7% 19.1% 56.8% 0.0% [All dropped] Dependent variable: Units of RNF divided by projected population, all ages. Analysis excludes City of London as an outlier Some technical points on the analysis of the EPCS formula are as follows (these apply to both the district- and county-level formulae): In the original EPCS formulae, the indicators are expressed both as rates (for example, population density), and as counts (for example, the number of commuters). In the simplified models presented here, all indicators are first converted to a rate per person. This is purely presentational and does not affect the resulting pattern of need. The City of London was excluded from the analysis as an outlier. This authority has very high assessed needs per head, presumably due to its large number of commuters and day visitors per resident. This meant that it had disproportionate influence over the final indicator weights. FINANCE WITH VISION 21

22 L Impact by local authority characteristic The impact of excluding the first six indicators on authorities with different characteristics is shown below. For all authorities, the impact would range between +4.8% to -16.0%. The reduction of 16% for one of the London boroughs appears to be something of an outlier, with the majority of authorities experiencing much smaller variations (as a group, London boroughs also see the largest median increase, of 1.7%) Annex 2 reports the median impact, by local authority characteristic, when fewer or more indicators were excluded from the formula. FINANCE WITH VISION 22

23 Chart 3: Impact of excluding selected indicators from EPCS (district-level) formula By administrative type Maximum 10% 4.8% 5% 3.0% 2.5% 90% of 0 Median 0.9% authorities 0.2% 0 1.7% % 0% -0.4% -5% -3.0% -3.1% Minimum -5.4% Indicators excluded: -10% BenefitsIncomeOlder2 BenefitsUnemployed -15% BenefitsDisability -20% CountryBirth -16.0% DayVisitors InCommuters Unitary Metropolitan London Shire District District Borough By rural / urban category 10% 5% 0% -5% -10% -15% -20% --- o 4.8% 3.0% 1.3% 1.6% 0.1% % % -0.2% -1.7% -2.4% -5.4% -16.0% Major Urban Urban with City Urban with Predominantly and Town Significant Rural Rural By deprivation level 10% 5% 0% 4.8% 3.8% 4.3% 1.8% 0.0% 0.1% ~ % -0.2% -5% -10% -4.7% -5.4% -5.0% -15% -20% -16.0% Most deprived Second most Third most Least deprived 25% deprived 25% deprived 25% 25% FINANCE WITH VISION 23

24 Environmental, Protective and Cultural Services (County Level) The county-level EPCS formula is also intended to reflect cost pressures for a wide range of local services. These include libraries, refuse disposal, magistrates and coroners courts, and civil defence The county-level needs assessment formula uses the same indicators as at the district level, with the exception of older adults receiving benefits (BenefitIncomeOlder2). The indicators are assigned different weightings than in the district-level formula. The following table identifies the impact of excluding each of these indicators. Table 5: Impact of excluding indicators County-level EPCS Indicators dropped from the model [None dropped] No. of indicators in model 8 Impact Absolute (±) change from original needs per head 90th Median Maximum percentile 0.0% 0.0% 0.0% Goodness of fit (R squared) 100.0% BenefitsDisability 7 0.2% 0.6% 1.3% >99.9% BenefitsUnemployed 6 0.4% 0.9% 2.2% 99.9% Sparsity 5 0.6% 1.5% 4.3% 99.7% CountryBirth 4 1.0% 2.2% 4.7% 99.3% DayVisitors 3 1.1% 3.0% 7.0% 98.7% Density 2 3.3% 6.9% 23.2% 86.6% BenefitsIncomeAll 1 8.0% 17.1% 26.9% 60.1% InCommuters [All dropped] 0 9.5% 20.6% 56.7% 0.0% Dependent variable: Units of RNF divided by projected population, all ages. 7 Analysis excludes City of London as an outlier Four indicators could be dropped from the county-level EPCS formula while keeping the median impact within ±1.0% of original assessed needs per head. These are residents receiving disability-related benefits (BenefitsDisability), those receiving unemploymentrelated benefits (BenefitsUnemployed), population sparsity (Sparsity) and residents country of birth (CountryBirth). 7 In DCLG s original county-level EPCS formula, the assessed needs of London authorities are reduced by 19% to reflect public transport support for buses and civil emergency contingency functions, which in London are assumed by other authorities. This adjustment is excluded from the analysis above, and so would presumably be applied to the final simplified formula. FINANCE WITH VISION 24

25 L Impact by local authority characteristic The impact of excluding these indicators on local authorities assessed needs per head would range from +3.7% to 4.7%. The largest median increase in assessed needs would be for authorities categorised as urban with city and town (+0.7%), with the largest decrease for predominantly rural authorities (-1.1%) Annex 2 reports the median impact, by local authority characteristic, when fewer or more indicators were excluded from the formula. FINANCE WITH VISION 25

26 Chart 4: Impact of excluding selected indicators from EPCS (county-level) formula By administrative type Maximum 5% 4% 3.7% 3% 2.4% 90% of 1.9% authorities 2% 1.4% 0 Median 1% 0.1% 0% % 0-0.3% 0-0.3% -1% Minimum -2% Indicators excluded: -1.7% BenefitsDisability -3% -2.8% BenefitsUnemployed -4% Sparsity -5% -4.5% CountryBirth -4.7% -6% Unitary Metropolitan London County District Borough By rural / urban category 5% 4% 3% 2% 1% 0% -1% -2% -3% -4% -5% -6% 2.4% 3.7% 0.2% 0 0.7% 1.8% -2.2% -1.9% -0.1% 0.5% -4.5% -4.7% t 'o..,. 0.0% 0.2% 0.1% ---10,, 0 0.6% _ 0-1.1% Major Urban Urban with City Urban with Predominantly and Town Significant Rural Rural By deprivation level 5% 4% 3% 2% 1% 0% -1% -2% -3% -4% -5% -6% 2.9% 3.1% 3.7% 2.2% -3.6% -3.5% -4.5% -4.7% Most deprived Second most Third most Least deprived 25% deprived 25% deprived 25% 25% FINANCE WITH VISION 26

27 Highway Maintenance While most of the RNF formulae calculate needs per person, the Highway Maintenance formula calculates needs per kilometre of road (weighted by road type). Needs per kilometre are calculated using the following eight indicators: Days with snow lying (SnowDays) Domestic visitor-nights per km of road (DomesticVisitorsKM) Foreign visitor-nights per km of road (ForeignVisitorsKM) Predicted gritting days (GrittingDays) Traffic flow, all motor vehicles (TrafficFlowAll) Commuters per km of road (CommutersKM) Residents per km of road (ResidentsKM) Traffic flow of heavy goods vehicles, buses and coaches (TrafficFlowHeavy) The impact of excluding each of these indicators is presented in the table below. Table 6: Impact of excluding indicators Highway Maintenance Impact Goodness Indicators dropped from Absolute (±) change from original needs per KM of fit the model 90th (R Median Maximum squared) percentile [None dropped] 8 0.0% 0.0% 0.0% 100.0% SnowDays 7 0.2% 0.4% 1.2% >99.9% DomesticVisitorsKM 6 0.2% 0.6% 2.6% >99.9% ForeignVisitorsKM 5 0.3% 0.9% 5.1% 99.9% GrittingDays 4 1.2% 3.4% 6.3% 99.8% TrafficFlowAll 3 1.4% 4.1% 9.3% 99.7% CommutersKM 2 3.2% 7.7% 32.8% 93.0% ResidentsKM 1 7.5% 22.1% 50.0% 82.4% No. of indicators in model TrafficFlowHeavy % 70.7% 143.0% 0.0% [All dropped] Dependent variable: Units of RNF divided by weighted road length (kilometres). Analysis excludes City of London as an outlier The first three of these indicators could be excluded while keeping the median impact within ±1.0% of original needs per kilometre. Excluding these indicators (SnowDays, DomesticVisitorsKM and ForeignVisitorsKM) would have a relatively small impact on assessed needs per head, with a median impact of ±0.3%. This simplified model would still be able to explain 99.9% of the variance in the original assessed needs per head. FINANCE WITH VISION 27

28 L Impact by local authority characteristic Excluding these three indicators would have an impact on authorities needs per head which ranged from +2.1% to -5.1%. The median impact would be zero or close to zero for all administrative types, rural-urban categories and deprivation groups. London boroughs would experience the full range of the impacts, though for 90% of boroughs the impact would be within ±0.9% of their original levels Annex 2 reports the median impact, by local authority characteristic, when fewer or more indicators were excluded from the needs assessment formula. FINANCE WITH VISION 28

29 Chart 5: Impact of excluding selected indicators from Highways Maintenance formula By administrative type Maximum 3% 2.1% 2% 1% 0.7% 90% of 0.7% 0.4% authorities 0 Median 0% --~O <Ot % 0.0% 0.0% 0.0% -1% Minimum -0.9% -2% -1.1% Indicators excluded: -3% -2.6% SnowDays DomesticVisitorsKM -4% ForeignVisitorsKM -5% -6% -5.1% Unitary Metropolitan London County District Borough By rural / urban category 3% 2% 1% 0% -1% -2% -3% -4% -5% -6% 2.1% 0.7% 0.3% 0.4% % 0-0.1% % o- 0.0% % Major Urban Urban with City Urban with Predominantly and Town Significant Rural Rural By deprivation level 3% 2% 1% 0% -1% -2% -3% -4% -5% -6% 2.1% -2.6% -0.9% -1.7% 0.7% 0.4% 0.7% % % ~0-0.0% % -2.6% -1.1% -1.0% -5.1% Most deprived Second most Third most Least deprived 25% deprived 25% deprived 25% 25% FINANCE WITH VISION 29

30 Other needs assessment formulae For a number of other needs assessment formulae, the trade-off between simplicity and the impact on assessed needs was relatively large. These formulae tended to be comprised of a smaller number of indicators, meaning that the removal of one or more indicators would have a large impact on needs per head These included the following elements of the Relative Needs Formula: Children s Services Youth & Community Children s Services Local Authority Central Education Functions Social Services for Younger Adults For each of these formulae, it was not possible to simplify the formulae without a median impact that was above our cut-off point of ±1.0%. For this reason, we do not provide a detailed breakdown of the range of impacts, based on local authority characteristics, for any one model. Children s Services Youth & Community Within Children s Services, the Youth & Community sub-block is made up of only two indicators: Secondary pupils in low-achieving ethnic groups, above a threshold (PupilEthnicityAbove) Children in out-of-work families receiving Child Tax Credit, above a threshold (ChildTaxCreditAbove) Excluding even one of these indicators results in a relatively large change in assessed needs per head, compared to the original allocations. The model in which the indicator for pupil ethnicity was dropped (PupilEthnicityAbove) would explain only 89.6% of the original variance in assessed need, as shown in the table below. The median impact of excluding this indicator would be ±4.1%. Table 7: Impact of excluding indicators Youth & Community Indicators dropped from the model No. of indicators in model Impact Absolute (±) change from original needs per head 90th Median Maximum percentile Goodness of fit (R squared) [None dropped] 2 0.0% 0.0% 0.0% 100.0% PupilEthnicityAbove 1 4.1% 10.3% 21.0% 89.6% ChildTaxCreditAbove [All dropped] % 36.3% 44.9% 0.0% Dependent variable: Units of RNF divided by projected population aged FINANCE WITH VISION 30

31 Children s Services Local Authority Central Education Functions Local Authority Central Education Functions is also comprised of two indicators: Children in out-of-work families receiving Child Tax Credit, above a threshold (ChildTaxCreditAbove) Population sparsity, at the ward level (SparsityWard) The original formula also includes a fixed amount of relative need per local authority. Before commencing the analysis, this amount was subtracted from each authority s assessed needs. In practice, this fixed amount per authority could be added back on to any simplified formula For this service, the simplified model was found to have a relatively large impact on assessed needs per head, even when both of the original indicators were used. In this case, the median impact was ±1.3%. The reason for this is that the original formula was calculated using two client groups: pupils aged 3-18, and resident pupils aged In our simplified formula, it was necessary to first express needs per head using only one of these client groups. It was found that using resident pupils gave the highest initial goodness-of-fit, though at 96.8% this was low compared to the services considered above. Table 8: Impact of excluding indicators Local Authority Central Education Functions Indicators dropped from the model No. of indicators in model Impact Absolute (±) change from original needs per head 90th Median Maximum percentile Goodness of fit (R squared) [None dropped] 2 1.3% 3.7% 11.8% 96.8% ChildTaxCreditAbove 1 3.9% 10.4% 23.2% 76.2% SparsityWard [All dropped] 0 6.8% 18.7% 34.7% 0.0% Dependent variable: Units of RNF divided by resident pupils aged FINANCE WITH VISION 31