Brighton & Hove: A Local Case Study

Applying the multilevel attainment model to a single local authority

1 Introduction

This case study applies the findings of the multilevel Attainment 8 model to Brighton & Hove, examining how the national model’s coefficients play out for a single local authority. Brighton & Hove is a useful test case because its secondary schools span a wide range of intake characteristics — from relatively affluent schools with low free school meals rates to schools serving communities with high levels of deprivation — all within one compact urban authority.

The analysis draws on the imputed full panel model (Analysis E), which uses nine fixed-effect predictors with random intercepts for year, Ofsted rating, and local authority nested within region. For full details on the model specification and the code used to fit the lmer() models, see the Model Experiments report.

NoteData note

The panel covers academic years 2021-22 to 2024-25. For 2024-25, four predictor variables (% low prior attainment, teacher retention, leadership pay %, and teacher sickness days) are carry-forward imputed from 2023-24 values. These observations are marked with triangular markers in the plots below.

2 The National Model

The results presented in this case study are derived from a multilevel linear model (also known as a mixed-effects or hierarchical model) fitted to all mainstream secondary schools in England over four academic years (2021–22 to 2024–25). The model predicts log-transformed Attainment 8 as a function of nine school-level characteristics, with random intercepts that allow the baseline to vary by year, Ofsted rating, and local authority (nested within region).

The model equation can be written as:

\[ \log(\text{ATT8}_{ij}) = \beta_0 + \sum_{k=1}^{9} \beta_k \, x_{kij} + u_{\text{year}} + u_{\text{Ofsted}} + u_{\text{region}} + u_{\text{LA|region}} + \varepsilon_{ij} \]

where \(i\) indexes schools, \(j\) indexes school-year observations, the \(\beta_k\) are fixed-effect coefficients, and the \(u\) terms are random intercepts. Because the outcome is on the log scale, each coefficient represents the proportional change in Attainment 8 associated with a one-unit change in the predictor (or, for log-transformed predictors, an elasticity — the percentage change in ATT8 for a 1% change in the predictor).

Table 1 presents the full set of estimated coefficients from the All Pupils model.

TipAdmissions policy dummy variables

ADMPOL_PT (admissions policy type) is a categorical variable with three levels: “Non-selective in a highly selective area” (i.e. a non-selective school in an area with grammar schools), “Other non-selective”, and “Selective” (grammar schools). Because it is categorical, lmer() automatically creates two dummy (indicator) variables, with “Non-selective in a highly selective area” as the reference category (its effect is absorbed into the intercept). Each dummy coefficient shows the difference in log(ATT8) relative to that reference group, all else being equal.

Table 1: Fixed-effect coefficients from the national multilevel model (Analysis E — All Pupils)

	Variable	Estimate	Std. Error	t value	Pr(>|t|)
(Intercept)	Intercept	4.63282	0.03688	125.61	< 0.001
log(PTFSM6CLA1A)	log(% FSM eligible)	-0.06748	0.00271	-24.90	< 0.001
log(PERCTOT)	log(% overall absence)	-0.21323	0.00463	-46.03	< 0.001
log(PNUMEAL)	log(% EAL)	0.00586	0.00119	4.92	< 0.001
PTPRIORLO	% low prior attainment	-0.00575	0.00014	-41.35	< 0.001
ADMPOL_PTOTHER NON SEL	Admissions: Other non-selective (ref: non-sel. in highly sel. area)	0.00057	0.00730	0.08	0.938
ADMPOL_PTSEL	Admissions: Selective (ref: non-sel. in highly sel. area)	0.10795	0.00666	16.20	< 0.001
gorard_segregation	Gorard segregation index	-0.03289	0.04739	-0.69	0.488
remained_in_the_same_school	Teacher retention %	0.00050	0.00005	10.20	< 0.001
teachers_on_leadership_pay_range_percent	% teachers on leadership pay	-0.00109	0.00020	-5.39	< 0.001
log(average_number_of_days_taken)	log(avg teacher sickness days)	-0.01456	0.00236	-6.17	< 0.001

NoteReading the coefficients

The four log-transformed predictors (% FSM, % absence, % EAL, teacher sickness days) have coefficients that act as elasticities: a 1% increase in the predictor is associated with a \(\beta\) % change in Attainment 8. The remaining continuous predictors enter on their raw scale, so their coefficients represent the proportional change in ATT8 for a one-unit increase. The two admissions policy rows are dummy variables for a three-level categorical predictor — each coefficient shows the difference relative to the reference category (“non-selective in a highly selective area”). In all cases, the effect on ATT8 in points depends on the school’s starting ATT8 score — multiply the proportional change by the school’s current ATT8 to get the point change.

3 Brighton & Hove’s Position Among All Local Authorities

Before examining individual schools, it is important to understand where Brighton & Hove sits as a local authority in the national model. The caterpillar plot below displays the estimated random intercepts for each local authority (nested within region), ordered by magnitude. Error bars show 95% confidence intervals (±1.96 posterior SE). Local authorities whose interval does not cross zero are meaningfully different from the overall mean, after accounting for fixed effects and other random effects. Brighton and Hove is highlighted in pink.

Caterpillar plot of LA random intercepts (imputed full panel model, All Pupils). Error bars show 95% confidence intervals. Brighton and Hove is highlighted in pink

ImportantBrighton & Hove’s random intercept

Brighton & Hove has the 4th highest positive random intercept among all 152 local authorities in England (estimate = 0.0568 on the log scale, 95% CI: 0.0286 to 0.0850).

This means that, after accounting for all nine model predictors — intake deprivation, absence rates, prior attainment, workforce stability, segregation, and admissions policy — Brighton & Hove schools achieve attainment that is systematically higher than what the national model would predict based on their characteristics alone.

In practical terms, this translates to roughly +5.8% on the ATT8 scale — i.e. if the model predicts a score of 50 for a school with Brighton & Hove’s characteristics, the LA effect alone pushes that prediction up to approximately 52.9.

This is not a raw league table position. A high random intercept does not mean Brighton & Hove has the highest raw attainment — it means the authority is doing better than expected given the challenges its schools face. The confidence interval excludes zero, confirming this is a statistically meaningful effect.

Possible explanations for this positive “Brighton effect” could include a range of different factors:

• community factors not captured by the model (parental engagement, cultural attitudes to education, strong school-home ties)
• consistently effective school leadership and governance across most of the schools in the authority.
• local authority governance
• school improvement programmes
• effective school-to-school collaboration networks

While all of these are examples of local authority or school-level factors that could account for the residual variance not captured by the model, new data or additional qualitative work would be required to establish potential causal pathways.

It is also worth noting that we have kept the number of explanatory variables as few as possible and selected just those which have most effect and contribute most to the story. In other associated files on this page, we experimented with other financial variables and those associated with special educational needs. While both added a very small amount of additional explanatory power, most of the their effect was bound up in other variables so we left out in the name of parsimony. Including them will have added an additional 1-2% of explanatory power.

4 Brighton & Hove Schools Map

Brighton & Hove secondary schools coloured by most recent Attainment 8 score (All Pupils)

The map above shows the location of each Brighton & Hove secondary school, coloured by its most recent Attainment 8 score. Click on any marker to see the school’s key characteristics. The spatial pattern reveals how attainment varies across the city, with notable differences between schools in the more affluent west and the more deprived east.

5 School Profiles

Brighton & Hove secondary schools: average characteristics across available years

School	Years	Mean ATT8 (All)	Mean ATT8 (Disadv.)	Mean ATT8 (Non-Dis.)	Mean % FSM	Mean % Absence	Mean % EAL
King’s School	4	56.9	43.1	59.6	16.1	6.6	11.8
Dorothy Stringer School	4	53.5	38.2	57.1	19.1	10.1	7.6
Varndean School	4	53.5	39.9	57.3	21.6	9.2	9.8
Cardinal Newman Catholic School	4	53.1	40.6	55.8	17.2	9.0	22.1
Blatchington Mill School	4	51.0	38.2	53.6	17.3	10.1	15.7
Patcham High School	4	47.8	36.0	51.0	20.4	8.8	6.9
Portslade Aldridge Community Academy	4	45.9	34.4	50.2	27.8	11.7	7.7
Hove Park School and Sixth Form Centre	4	45.2	35.3	49.6	31.3	11.8	28.7
Longhill High School	4	37.2	26.6	42.5	33.6	14.7	7.3
Brighton Aldridge Community Academy	4	36.8	30.8	42.0	47.9	15.8	12.2

6 School Characteristics Map

Use the dropdown below to explore how different outcome and predictor variables are distributed across Brighton & Hove’s secondary schools. Circle colour and size both reflect the selected variable.

Brighton & Hove secondary schools — select a variable to change colour and circle size

Comparing the two maps reveals the geographic correlation between intake deprivation and attainment in Brighton & Hove. Use the dropdown to explore how different variables relate to school location — for instance, switching between % FSM and Attainment 8 shows how schools with higher deprivation in the eastern parts of the city tend to have lower attainment, while schools in the west show the opposite pattern.

7 Non-Linear Effects of Log-Transformed Predictors

A key feature of the multilevel model is that four predictors enter on the log scale: % FSM, % absence, % EAL, and average teacher sickness days. This means the relationship between these predictors and attainment is non-linear on the original (raw) scale — a given absolute change in a predictor has a different predicted effect depending on where the school starts.

Before exploring these non-linear patterns, it is helpful to see how the model produces a prediction for a single school. The worked example below uses Longhill High School to show how each variable contributes to the predicted Attainment 8 score, and how changes in key variables shift that prediction.

7.1 Worked example: Longhill High School

7.1.1 Step 1: Longhill’s predictor values and their contributions

The table below shows each variable in the model, Longhill High School’s value for the most recent year (2024-25), how that value enters the model (log-transformed where applicable), the national model coefficient, and the resulting contribution to predicted log(ATT8).

Variable	Raw value	Model input (x)	Coefficient (β)	β × x
Intercept	—	1	4.63282	4.6328
% FSM eligible	40.1	3.6914	-0.06748	-0.2491
% overall absence	16.7	2.8141	-0.21323	-0.6000
% EAL	7.4	2.0015	0.00586	0.0117
% low prior attainment	25.3	25.3	-0.00575	-0.1455
Admissions policy (OTHER NON SEL)	OTHER NON SEL	1	0.00057	0.0006
Gorard segregation index	0.129	0.1285	-0.03289	-0.0042
Teacher retention %	40.4	40.4	0.00050	0.0201
% on leadership pay	15.6	15.5556	-0.00109	-0.0170
Avg teacher sickness days	9.3	2.23	-0.01456	-0.0325

Sum of fixed-effect contributions: 3.6169

7.1.2 Step 2: Adding the random effects

The multilevel model also includes random intercepts that adjust the baseline for contextual groupings:

Random effect	Adjustment
Year (2024-25)	-0.03460
Ofsted rating (Requires Improvement)	-0.03205
Region (South East)	-0.01490
LA (Brighton and Hove)	0.05682

Sum of random effects: -0.0247

7.1.3 Step 3: The prediction

Adding the fixed and random components:

\[\log(\widehat{\text{ATT8}}) = 3.6169 + -0.0247 = 3.5922\]

\[\widehat{\text{ATT8}} = e^{3.5922} = 36.3\]

Longhill’s actual Attainment 8 in 2024-25 was 32.9, so the residual (actual − predicted) is -3.4 points.

(Internal check: predict() returns 3.5921 on the log scale = 36.3 on the ATT8 scale.)

7.1.4 Step 4: What happens when a variable changes?

The power of the model is that we can change one variable at a time and see how the prediction shifts. Below we show two realistic scenarios for Longhill, each corresponding to a plausible policy target.

7.1.4.1 Scenario A: % FSM falls to the Brighton & Hove average (28.8%)

Longhill currently has a FSM rate of 40.1%, well above the city average of 28.8%. What if intake redistribution brought Longhill’s FSM rate down to that average?

	% FSM	log(% FSM)	FSM contribution	Predicted ATT8
Current	40.1	3.6914	-0.2491	36.3
City average (28.8%)	28.8	3.3593	-0.2267	37.1
Change	-11.3	-0.3320	+0.0224	+0.8

Reducing Longhill’s FSM rate from 40.1% to the city average of 28.8% changes the log-FSM input from 3.6914 to 3.3593 (a shift of -0.3320). Multiplied by the FSM coefficient (-0.06748), this shifts the predicted log(ATT8) by +0.0224, producing a predicted ATT8 gain of +0.8 points.

7.1.4.2 Scenario B: % absence falls to the national average (8.4%)

Longhill’s absence rate is 16.7% — far above the national mainstream average of 8.4%. What if targeted attendance interventions brought it in line with the national norm?

	% Absence	log(% Absence)	Absence contribution	Predicted ATT8
Current	16.7	2.8141	-0.6000	36.3
National avg (8.4%)	8.4	2.1228	-0.4526	42.1
Change	-8.3	-0.6913	+0.1474	+5.8

Bringing Longhill’s absence from 16.7% down to the national average of 8.4% shifts the log-absence from 2.8141 to 2.1228 (-0.6913). The absence coefficient (-0.21323) translates this into a predicted ATT8 gain of +5.8 points.

TipKey takeaway

For Longhill, reducing absence to the national average is predicted to raise Attainment 8 by +5.8 points — a substantial improvement. Reducing FSM to the city average would add +0.8 points. Both are positive, but the absence reduction produces a larger gain and is directly actionable through attendance policy. The scenario tables in the sections below apply this same arithmetic to every Brighton & Hove school.

NoteA note on the scenario calculations

The worked example above uses the full model prediction for Longhill — that is, it feeds all nine predictor values plus the school’s random effects (year, Ofsted rating, region, and local authority) through the model, then changes one variable at a time and re-computes the prediction. This is the most accurate way to estimate the impact of a change for a specific school.

The policy scenario tables that follow in sections 7.3 and 7.4 use a simplified approximation: they take the model’s coefficient for the variable of interest, compute the proportional change implied by a given shift on the log scale, and then multiply that proportional change by the city-wide mean observed Attainment 8 to convert it to ATT8 points. This is a reasonable shorthand for comparing schools side-by-side, but it produces slightly different point estimates from the full-model approach because it uses an average ATT8 as the base rather than each school’s specific predicted score.

For Longhill, for example, the full-model FSM scenario above predicts a gain of +0.8 points, while the simplified table in section 7.3 reports roughly +1.1 points for the same change. Both use the same coefficient from the same model — the difference is entirely due to the base ATT8 used to convert the proportional effect into points. The full-model approach is more precise for individual schools; the simplified approach is better suited to quick cross-school comparisons.

7.2 What the log transform means in practice

Brighton & Hove schools: raw and log-transformed predictor values (most recent year)

School	Year	% FSM	% Absence	% EAL	log(% FSM)	log(% Absence)	log(% EAL)
Dorothy Stringer School	2024-25	18.2	9.8	7.9	2.901	2.279	2.067
Blatchington Mill School	2024-25	20.1	10.8	16.1	3.001	2.378	2.779
Cardinal Newman Catholic School	2024-25	20.3	8.5	22.7	3.011	2.138	3.122
King’s School	2024-25	20.6	6.6	9.6	3.025	1.881	2.262
Varndean School	2024-25	25.1	9.0	9.7	3.223	2.200	2.272
Patcham High School	2024-25	25.2	7.9	7.5	3.227	2.072	2.015
Portslade Aldridge Community Academy	2024-25	29.7	11.3	8.1	3.391	2.428	2.092
Hove Park School and Sixth Form Centre	2024-25	35.8	12.9	29.2	3.578	2.555	3.374
Longhill High School	2024-25	40.1	16.7	7.4	3.691	2.814	2.001
Brighton Aldridge Community Academy	2024-25	52.6	14.4	12.2	3.963	2.671	2.501

7.3 The diminishing-returns curve: % FSM

Before examining the non-linear effects, it is worth noting where Brighton & Hove sits nationally on this variable. The city’s mean FSM rate across its secondary schools is 28.8%, compared to the national average of 28.3% — placing it at roughly the 52nd percentile among all 152 local authorities. Brighton & Hove is, in terms of overall deprivation levels, an unremarkable authority. This is important context for the policy discussion that follows.

The table above shows how the same raw predictor value maps to very different positions on the log scale. To illustrate the practical consequence, consider two hypothetical Brighton & Hove schools:

• School A with 10% FSM: log(10) = 2.30. If FSM rises by 5 percentage points to 15%, log(15) = 2.71 — a shift of +0.41 on the log scale.
• School B with 40% FSM: log(40) = 3.69. If FSM rises by the same 5 percentage points to 45%, log(45) = 3.81 — a shift of only +0.12 on the log scale.

Because the model operates on the log scale, the same absolute increase in FSM rate produces a predicted effect that is more than three times larger for the lower-FSM school.

Diminishing returns: the same 5pp change in % FSM produces a much larger shift on the log scale at the low end. Brighton & Hove schools are marked in orange.

7.3.1 Policy scenario: redistributing FSM towards the city average

One stated policy ambition for Brighton & Hove has been to reduce the concentration of disadvantaged pupils at the schools with the highest FSM rates — effectively redistributing intake towards a more even spread. The model allows us to estimate the predicted attainment gain if schools above the city average FSM rate were brought down to it.

Scenario: if schools above the city average (29% FSM) reduced to it — predicted ATT8 effect (All Pupils model)

School	Current % FSM	Target % FSM	Change in log(FSM)	Predicted ATT8 change (%)	≈ ATT8 points
Brighton Aldridge Community Academy	52.6	28.8	-0.603	4.16	1.9
Longhill High School	40.1	28.8	-0.332	2.27	1.0
Hove Park School and Sixth Form Centre	35.8	28.8	-0.219	1.49	0.7
Portslade Aldridge Community Academy	29.7	28.8	-0.032	0.21	0.1

Across the 4 schools above the city average, bringing FSM rates down to 29% would produce a predicted gain of roughly +0.9 ATT8 points per school on average. These are modest gains — and this is the key insight: FSM redistribution alone produces relatively small predicted improvements, because the log transformation compresses changes at the high end of the distribution.

Moreover, this calculation assumes everything else remains constant — in practice, reducing FSM concentration at one school necessarily increases it elsewhere, and the non-linear curve means the receiving schools (at the low end) would experience a larger negative predicted effect per pupil redistributed. The policy therefore has ambiguous net effects on city-wide attainment.

Note: the ATT8 point estimates in this table are approximate — they convert the model’s proportional effect using the city-wide mean ATT8 as a common base. For a more precise school-specific calculation using the full model (all nine predictors plus random effects), see the Longhill worked example above.

WarningThe limits of FSM redistribution as a lever

The model suggests that redistributing disadvantaged pupils more evenly across Brighton & Hove schools would produce only modest predicted attainment gains at the high-FSM schools, while potentially creating larger predicted losses at the low-FSM schools receiving additional disadvantaged pupils (because they sit on the steep part of the curve). This does not mean FSM redistribution is a bad policy — there may be valid equity and social cohesion arguments — but the model does not support it as an effective route to raising overall attainment.

7.4 The diminishing-returns curve: % Absence

The same non-linear logic applies to overall absence, but the policy implications are far more urgent for Brighton & Hove because absence is both a directly actionable variable and because the city is a national outlier on this measure.

WarningBrighton & Hove is a national outlier on absence — but not on disadvantage

In 2024-25, Brighton & Hove’s mean school absence rate is 10.8%, compared to a national average of 8.2% — a gap of +2.6 percentage points. This places the authority 151st out of 152 local authorities in England for absence (where 152 is the worst), in the 99 — near the very worst in the country percentile.

By contrast, Brighton & Hove’s mean FSM rate is 28.8% versus the national mean of 28.3% (difference: +0.5 pp), placing it 79th out of 152 LAs — almost exactly at the national median (52th percentile). The city is unremarkable for its overall level of deprivation.

8 out of 10 Brighton & Hove secondary schools have absence rates above the national school-level average.

Highest-absence local authorities across all panel years (mean school absence %, rank)

Local Authority	2021-22	2022-23	2023-24	2024-25
Knowsley	12.6% (152/152)	11.8% (149/152)	11.9% (151/152)	10.9% (152/152)
Brighton and Hove	10.8% (144/152)	10.5% (137/152)	11.0% (143/152)	10.8% (151/152)
Newcastle upon Tyne	12.2% (151/152)	12.9% (152/152)	11.4% (150/152)	10.6% (150/152)
Southampton	10.1% (119/152)	10.6% (139/152)	11.0% (145/152)	10.5% (149/152)
Bradford	11.5% (149/152)	11.9% (150/152)	11.2% (147/152)	10.3% (148/152)
Plymouth	11.1% (147/152)	10.8% (144/152)	11.3% (149/152)	10.2% (147/152)
Middlesbrough	11.8% (150/152)	12.9% (151/152)	12.2% (152/152)	10.1% (146/152)
Sefton	10.1% (122/152)	10.3% (129/152)	10.5% (133/152)	10.0% (145/152)
Devon	10.8% (145/152)	10.9% (145/152)	10.8% (139/152)	9.9% (144/152)
Dorset	10.3% (134/152)	10.2% (127/152)	10.7% (137/152)	9.9% (143/152)
Halton	10.1% (118/152)	10.3% (131/152)	11.0% (144/152)	9.8% (142/152)
Blackpool	9.8% (103/152)	11.1% (147/152)	11.3% (148/152)	9.8% (141/152)
Hartlepool	10.2% (130/152)	11.1% (148/152)	10.5% (134/152)	9.8% (140/152)
Bristol, City of	10.6% (141/152)	11.0% (146/152)	11.0% (142/152)	9.6% (135/152)
Gateshead	11.4% (148/152)	10.8% (143/152)	11.2% (146/152)	9.5% (130/152)
St. Helens	10.7% (143/152)	9.6% (103/152)	10.0% (122/152)	9.1% (121/152)
Torbay	10.9% (146/152)	10.2% (125/152)	10.3% (126/152)	8.2% (75/152)

This combination — typical deprivation but atypical absence — makes absence reduction an especially important lever for the city. The non-linear effects of the log transformation compound the problem: because most Brighton & Hove schools sit at the high end of the national absence distribution, they are on the flatter part of the curve where each percentage point of improvement buys less predicted attainment gain than it would for schools starting from a lower base. Nonetheless, the sheer scale of the gap means that closing it would deliver substantial improvements.

Diminishing returns: the same 1pp change in absence has a much larger model impact at the low end. Brighton & Hove schools (orange) cluster to the right of the national average, on the flatter part of the curve.

7.4.1 Policy scenario: reducing absence by 1 percentage point

Unlike FSM, absence is something schools and the council can directly influence. The table below shows the predicted ATT8 gain from a 1 percentage point reduction in absence at each Brighton & Hove school:

Scenario: 1pp reduction in absence at each school — predicted ATT8 effect (All Pupils model)

School	Current % Absence	Target % Absence	Change in log(Absence)	Predicted ATT8 change (%)	≈ ATT8 points
King’s School	6.6	5.6	-0.165	3.59	1.9
Patcham High School	7.9	6.9	-0.135	2.91	1.3
Cardinal Newman Catholic School	8.5	7.5	-0.125	2.71	1.4
Varndean School	9.0	8.0	-0.117	2.54	1.2
Dorothy Stringer School	9.8	8.8	-0.108	2.33	1.2
Blatchington Mill School	10.8	9.8	-0.097	2.10	1.1
Portslade Aldridge Community Academy	11.3	10.3	-0.092	1.99	0.8
Hove Park School and Sixth Form Centre	12.9	11.9	-0.081	1.74	0.7
Brighton Aldridge Community Academy	14.4	13.4	-0.072	1.54	0.6
Longhill High School	16.7	15.7	-0.062	1.33	0.4

The non-linearity is immediately visible: a school with low absence (around 4–5%) stands to gain roughly +1.9 ATT8 points from a 1pp reduction, while a school already at higher absence (8–9%) gains only around +0.4 points from the same 1pp reduction. The low-absence school gains more because it sits on the steep part of the log curve.

7.4.2 Policy scenario: closing the gap to the national average

Given that Brighton & Hove’s absence rates are among the worst in England, a more ambitious but entirely reasonable policy target is to ask: what if every school in the city brought its absence rate down to the national school-level average of 8.2%? This is not an aspirational stretch target — it is simply asking Brighton & Hove schools to achieve what the typical English secondary school already does.

Scenario: if all schools above the national average (8.2%) reduced to it — predicted ATT8 effect (All Pupils model)

School	Current % Absence	National Average (Target)	Reduction needed (pp)	Predicted ATT8 change (%)	≈ ATT8 points
Longhill High School	16.7	8.2	8.4	16.25	5.3
Brighton Aldridge Community Academy	14.4	8.2	6.2	12.74	4.6
Hove Park School and Sixth Form Centre	12.9	8.2	4.6	10.01	4.1
Portslade Aldridge Community Academy	11.3	8.2	3.1	7.07	3.0
Blatchington Mill School	10.8	8.2	2.6	5.93	3.0
Dorothy Stringer School	9.8	8.2	1.5	3.71	2.0
Varndean School	9.0	8.2	0.8	1.98	1.0
Cardinal Newman Catholic School	8.5	8.2	0.3	0.64	0.3

8 out of 10 Brighton & Hove schools currently have absence above the national average. The reductions required range from modest (around 1 pp) to substantial (8.4 pp for Longhill High School). Across these 8 schools, achieving the national average would produce predicted gains averaging +2.9 ATT8 points per school, with a city-wide total gain of approximately +23.3 ATT8 points summed across all affected schools.

Compare this to the FSM redistribution scenario above, where bringing high-FSM schools down to the city average produced only modest gains. The absence scenario delivers larger predicted improvements — and crucially, it does not have the zero-sum problem of FSM redistribution. Reducing absence at one school does not increase it at another.

For disadvantaged pupils, the gains would be even larger: an average of +3.3 ATT8 points per school (compared to +2.9 for all pupils), because the absence coefficient is 43% larger for disadvantaged pupils. Closing the absence gap to the national average would therefore simultaneously raise overall attainment and narrow the disadvantage gap.

TipThe absence gap is not inevitable

Brighton & Hove’s absence rates are not a necessary consequence of its intake composition. The city has near-average levels of deprivation (FSM rates at the national median) but near-worst absence. Many local authorities with similar or higher FSM rates achieve substantially lower absence. This suggests that the absence problem is driven by local factors — attendance culture, enforcement practices, alternative provision use, or community-level dynamics — rather than being an unavoidable correlate of disadvantage. It is, in principle, solvable.

7.4.3 Comparing the two levers: FSM vs Absence

The model’s coefficients allow a direct comparison of the two policy levers:

• log(FSM) coefficient (All Pupils): -0.0675
• log(Absence) coefficient (All Pupils): -0.2132

When standardised by the Brighton & Hove spread of each variable (SD of log-FSM = 0.35; SD of log-Absence = 0.29), the absence coefficient is 2.6 times as important as FSM for predicting attainment across Brighton & Hove schools.

This has a clear policy implication — and it is reinforced by Brighton & Hove’s position in the national distribution:

• FSM: Brighton & Hove sits at the 52nd percentile nationally (79th of 152 LAs) — the city’s deprivation levels are unremarkable. There is limited scope for FSM-based policy to differentiate Brighton & Hove from its statistical peers.
• Absence: Brighton & Hove sits at the 99th percentile nationally (151st of 152 LAs) — among the very worst in England. There is therefore enormous scope for absence reduction to bring the city into line with its peers, and the model predicts meaningful attainment gains from doing so.

The mismatch between a typical level of disadvantage and an extreme level of absence is what makes absence the standout policy priority for Brighton & Hove specifically.

ImportantKey insight: absence is a far more effective lever than FSM redistribution

Across Brighton & Hove, a one-standard-deviation improvement in absence rates is predicted to improve ATT8 by roughly 2.6 times as much as a one-standard-deviation change in FSM composition. Moreover, absence is directly actionable through attendance strategies, family liaison, and early intervention — whereas FSM redistribution requires changing school admissions patterns, which is politically contentious, operationally complex, and (as shown above) produces ambiguous net effects on city-wide attainment.

The case is further strengthened by Brighton & Hove’s outlier position: the city has the 2nd-highest absence rate among 152 local authorities in England, despite having a near-average FSM rate. Simply converging towards the national mean absence level — something that most LAs already achieve — would deliver predicted attainment gains substantially larger than any feasible FSM redistribution scenario.

The model therefore supports a policy strategy that prioritises:

1. Protecting low-absence schools from any deterioration (the steep part of the curve)
2. Intensive early intervention at moderate-absence schools (where there is most to gain)
3. Sustained absence reduction programmes at high-absence schools (where gains per unit are smaller but the absolute room for improvement is greatest)
4. Benchmarking against national norms — the target should not be the B&H average (which is itself an outlier) but the national average, which most of the city’s schools currently exceed

This is a more promising and directly controllable route to raising attainment than attempting to redistribute intake composition.

7.4.4 Comparing marginal returns: FSM vs Absence on the same scale

Figure 1: Marginal ATT8 gain from a 1 percentage point reduction in each variable, at different starting values. The steeper the curve, the bigger the payoff. Absence (green) delivers larger gains than FSM (purple) at every level, and the gap widens at lower starting values. Brighton & Hove schools are marked on each curve.

Figure 2: Marginal ATT8 gain from a 1pp reduction, with starting values standardised to SDs from the national mean. At the same relative position in the national distribution, absence reduction delivers substantially larger gains than FSM reduction.

8 How Mediation and Confounding Change the Story

The non-linear effects described above tell us how much a change in a predictor matters at different starting points. But a crucial question remains: what are the coefficients actually measuring? The answer depends on which other variables are in the model — and the full imputed panel model (Analysis E, with 9 predictors) tells a fundamentally different story from a simpler model that omits prior attainment and workforce controls.

Two distinct mechanisms are at work:

• Mediation: prior attainment sits on the causal pathway between disadvantage and attainment — disadvantaged pupils arrive at secondary school with lower KS2 scores, which in turn predict lower KS4 outcomes. Controlling for prior attainment removes this indirect effect, revealing the direct (residual) relationship between school-level deprivation and attainment.
• Confounding: absence is correlated with FSM but is not simply caused by it — schools in deprived areas face higher absence for reasons partly independent of deprivation itself. Without controlling for absence, the FSM coefficient absorbs both the deprivation effect and the absence effect.

8.1 The problem: omitted variable bias

Key coefficients from the imputed full panel model (Analysis E) across all three pupil groups

Predictor	All Pupils	Disadvantaged	Non-Disadvantaged
log(% FSM)	-0.06748	0.00765	-0.04180
log(% Absence)	-0.21323	-0.30476	-0.17217
% Low Prior Attainment	-0.00575	-0.00533	-0.00521
log(% EAL)	0.00586	0.02322	0.00798
Teacher Retention	0.00050	0.00003	0.00046
log(Teacher Sickness)	-0.01456	-0.01839	-0.01744

The table above reveals several features that are crucial for interpreting the model in Brighton & Hove.

8.2 Prior attainment absorbs much of the “disadvantage penalty”

The single most important mediating relationship in this model is between % FSM (intake deprivation) and % low prior attainment (PTPRIORLO). Prior attainment is a mediator, not a confounder: the causal chain runs from disadvantage → lower KS2 achievement → lower KS4 attainment. These two variables are strongly correlated because schools with more disadvantaged intakes receive more pupils who arrived with weaker primary-school preparation. In a simpler model without PTPRIORLO, the FSM coefficient absorbs both the direct deprivation effect and the indirect effect operating through prior attainment, making deprivation appear more damaging than it actually is once the prior-attainment pathway is accounted for.

In Brighton & Hove, the correlation between log(% FSM) and % low prior attainment is r = 0.81 — meaning schools with higher deprivation levels consistently receive pupils with weaker primary-school preparation. This is the mediating pathway in action: disadvantage leads to lower KS2 scores, which in turn predict lower KS4 attainment. Without controlling for prior attainment, any model would attribute this entire indirect effect to FSM alone.

In the full model (Analysis E), PTPRIORLO picks up the prior-attainment pathway of the disadvantage penalty, freeing log(% FSM) to capture only the direct residual relationship between intake composition and outcomes. This decomposition reveals something remarkable.

8.3 The FSM sign-flip: a structural finding only visible in the full model

The FSM coefficient flips sign for disadvantaged pupils once prior attainment is controlled for

Pupil Group	Full Model (Analysis E)	Direction
All Pupils	-0.06748	Negative ↓
Disadvantaged	0.00765	Positive ↑
Non-Disadvantaged	-0.04180	Negative ↓

For All Pupils and Non-Disadvantaged, the FSM coefficient is negative as expected: higher deprivation is associated with lower attainment, even after controlling for prior attainment. But for Disadvantaged pupils, the coefficient flips to positive.

This means that, after accounting for prior attainment and all other controls, a school with a higher share of FSM-eligible pupils is predicted to achieve slightly higher attainment for its disadvantaged subgroup. The coefficient is small (+0.00765), but it is consistent across all years in the data, making it a stable structural finding rather than a statistical artefact.

Why does this happen? Two plausible explanations:

1. The “critical mass” effect: Schools serving predominantly disadvantaged cohorts may concentrate more resources — Pupil Premium funding, specialist teaching, pastoral support — on those pupils. Where disadvantaged pupils are a large share of the cohort, interventions are scaled rather than marginal.

2. The composition effect: In schools where almost all pupils are disadvantaged, the disadvantaged subgroup essentially is the school average, removing any within-school composition penalty.

Crucially, this sign flip is invisible in a simpler model. Without PTPRIORLO in the model, log(% FSM) is forced to absorb the indirect effect operating through prior attainment as well, and the net effect is negative for all three groups. The structural relationship is suppressed because the mediating pathway has not been separated out.

8.4 What this means for Brighton & Hove schools

Brighton & Hove schools: decomposing the disadvantage-related penalty into FSM, absence (confounder), and prior attainment (mediator) components (log-ATT8 scale)

School	% FSM	% Absence	% Low Prior	FSM effect (All)	Absence effect (All)	Prior att. effect	Combined	FSM effect (Disadv.)
Dorothy Stringer School	18.2	9.8	19.1	-0.196	-0.486	-0.110	-0.792	0.022
Blatchington Mill School	20.1	10.8	16.8	-0.202	-0.507	-0.097	-0.806	0.023
Cardinal Newman Catholic School	20.3	8.5	17.6	-0.203	-0.456	-0.101	-0.760	0.023
King’s School	20.6	6.6	12.3	-0.204	-0.401	-0.071	-0.676	0.023
Varndean School	25.1	9.0	18.5	-0.217	-0.469	-0.106	-0.792	0.025
Patcham High School	25.2	7.9	16.4	-0.218	-0.442	-0.094	-0.754	0.025
Portslade Aldridge Community Academy	29.7	11.3	24.9	-0.229	-0.518	-0.143	-0.890	0.026
Hove Park School and Sixth Form Centre	35.8	12.9	24.2	-0.241	-0.545	-0.139	-0.925	0.027
Longhill High School	40.1	16.7	25.3	-0.249	-0.600	-0.146	-0.995	0.028
Brighton Aldridge Community Academy	52.6	14.4	29.4	-0.267	-0.569	-0.169	-1.005	0.030

The table above decomposes the model’s predicted penalty from intake composition into three parts for each Brighton & Hove school:

• FSM effect: the direct (residual) contribution of log(% FSM) × its coefficient
• Absence effect: the contribution of log(% absence) × its coefficient — a confounder correlated with FSM but not simply caused by it
• Prior attainment effect: the contribution of % low prior attainment × its coefficient — a mediator on the causal pathway from disadvantage to attainment

For Brighton Aldridge Community Academy (53% FSM, 14.4% absence, 29% low prior), the FSM effect on all-pupil attainment is -0.267 log points, the absence effect is -0.569, and the prior attainment effect is -0.169 — meaning prior attainment and absence together account for the bulk of the total intake-related penalty.

But look at the last column: the FSM effect for disadvantaged pupils at the same school is +0.030 — positive. This school’s disadvantaged pupils are predicted to do slightly better precisely because it is a high-FSM school. Conversely, Dorothy Stringer School (18% FSM) has a FSM effect for disadvantaged pupils of +0.022 — smaller, because there are fewer disadvantaged peers.

8.5 Absence hits disadvantaged pupils hardest

The absence coefficient is largest (most negative) for disadvantaged pupils

Pupil Group	Coefficient	Ratio to All
All Pupils	-0.21323	1.00
Disadvantaged	-0.30476	1.43
Non-Disadvantaged	-0.17217	0.81

The absence coefficient for disadvantaged pupils (-0.30476) is 43% larger in absolute terms than for all pupils (-0.21323). This means every unit of absence hits harder for the pupils who can least afford it.

Combined with the non-linear log transformation discussed above, this creates a double disadvantage:

1. Disadvantaged pupils are more sensitive to absence (larger coefficient)
2. Schools with lower absence rates see the biggest marginal effect (steeper curve)

For Brighton & Hove, this double disadvantage is compounded by a third factor: the city is a national outlier on absence (ranked 151st out of 152 LAs, with a mean school absence rate of 10.8% versus the national average of 8.2%). This means Brighton & Hove’s disadvantaged pupils face a triple penalty: they are more sensitive to absence per unit, they attend schools with absence rates far above the national norm, and the high baseline absence places their schools on the flatter part of the log curve where improvement is harder to achieve.

Crucially, this is not a simple correlate of deprivation. Brighton & Hove’s FSM rate sits at the 52nd percentile nationally — almost exactly at the median. Many local authorities with similar or higher levels of disadvantage achieve substantially lower absence rates. The absence problem is therefore amenable to local policy action rather than being an inevitable consequence of the city’s socioeconomic profile.

Absence reduction strategies targeted at schools with high absence rates — particularly those serving disadvantaged cohorts — therefore represent the single most effective lever the model identifies for closing the attainment gap, and the lever where Brighton & Hove has the most room for improvement.

Predicted ATT8 gain from a 1pp absence reduction, by pupil group — the gap between disadvantaged and non-disadvantaged illustrates how absence reduction disproportionately benefits vulnerable pupils.

ImportantMediation, confounding, and the story they tell

The full model reveals that what appears to be a simple “disadvantage penalty” is actually a composite of distinct mechanisms:

1. Prior attainment (mediation) — the primary mediating pathway between deprivation and attainment. Schools with more disadvantaged intakes receive pupils with weaker KS2 preparation, and this indirect effect accounts for a large portion of the apparent FSM–attainment relationship. Once this mediator is included, the FSM coefficient captures only the direct residual effect of school-level deprivation.

2. Residual deprivation — once prior attainment is accounted for, higher FSM rates actually help disadvantaged pupils slightly (the sign flip), while still penalising overall and non-disadvantaged attainment.

3. Absence (confounding) — the strongest directly actionable predictor, hitting disadvantaged pupils hardest. Absence is correlated with FSM but is not simply caused by it, making it a confounder that inflates the apparent deprivation penalty in simpler models. Unlike FSM composition, absence can be influenced through policy and intervention, and the non-linear log scale means the greatest gains come from protecting schools with moderate absence rates. Brighton & Hove’s position as a national outlier on absence — but not on deprivation — makes this the defining policy challenge for the authority.

For Brighton & Hove, the practical conclusion is clear: investing in absence reduction and KS2–KS3 transition support will deliver far more than attempts to redistribute intake composition across schools. The fact that the city has near-average deprivation levels but near-worst absence rates demonstrates that the absence problem is not an inevitable consequence of its intake profile — and is therefore, in principle, solvable through targeted local action.

9 Observed vs Predicted

How well does the national model predict attainment for Brighton & Hove schools specifically? The scatter plot below shows all schools nationally in 2024-25 as grey points, with Brighton & Hove schools highlighted in pink and labelled by name. The dashed line marks perfect prediction — schools above the line are outperforming the model’s expectations; schools below are underperforming. The three panels show All Pupils, Disadvantaged, and Non-Disadvantaged outcomes.

Observed vs predicted Attainment 8 for all schools (2024-25). Brighton & Hove schools highlighted in pink and labelled.

The plot reveals how each Brighton & Hove school sits relative to the full national distribution of predicted attainment. Schools well above the dashed line have positive residuals — they are achieving more than the model predicts given their intake and context. Schools below are underperforming relative to expectations.

9.1 Brighton & Hove schools across all years

The following plot focuses on Brighton & Hove schools only, showing how each school’s observed and predicted attainment tracks across all four years of the panel. Triangular markers indicate years where carry-forward imputed predictors were used (2024-25).

Observed vs model-predicted Attainment 8 for Brighton & Hove schools, faceted by pupil group. Points above the dashed line are outperforming expectations.

9.2 Interactive observed vs predicted

10 Residuals by School

Model residuals — the difference between observed and predicted Attainment 8 — reveal which schools are outperforming (positive residual) or underperforming (negative residual) the model’s expectations, given their intake characteristics and context.

Model residuals (observed − predicted) for Brighton & Hove schools by outcome group and year. Schools above zero are outperforming the model’s expectations.

10.1 Interactive residuals

10.2 Residual map

Brighton & Hove schools coloured by mean model residual across all panel years. Use the dropdown to switch between outcome groups. Green = outperforming, red = underperforming.

Use the dropdown above the map to switch between outcome groups: All Pupils, Disadvantaged, and Non-Disadvantaged. The residuals shown are mean values across all panel years — the same figures used to rank schools in the value-added league tables in Section 12 below. Schools shown in green are outperforming the model’s expectations given their intake, context, and workforce characteristics, while schools in red are underperforming. Circle size reflects the magnitude of the residual. Click any marker to see the full breakdown of residuals across all three groups, including observed and predicted scores and the number of years contributing to the average. The spatial pattern can reveal whether there are area-level factors — community resources, transport links, local employment conditions — that the model does not capture.

11 Brighton & Hove in National Context

Brighton & Hove schools: position within national quartiles (most recent year)

School	% FSM	FSM Quartile	% Absence	Absence Quartile	% EAL	EAL Quartile
Blatchington Mill School	20.1	Q2	10.8	Top 25%	16.1	Q3
Brighton Aldridge Community Academy	52.6	Top 25%	14.4	Top 25%	12.2	Q3
Cardinal Newman Catholic School	20.3	Q2	8.5	Q2	22.7	Q3
Dorothy Stringer School	18.2	Q2	9.8	Q3	7.9	Q2
Hove Park School and Sixth Form Centre	35.8	Q3	12.9	Top 25%	29.2	Top 25%
King’s School	20.6	Q2	6.6	Bottom 25%	9.6	Q2
Longhill High School	40.1	Top 25%	16.7	Top 25%	7.4	Q2
Patcham High School	25.2	Q3	7.9	Q2	7.5	Q2
Portslade Aldridge Community Academy	29.7	Q3	11.3	Top 25%	8.1	Q2
Varndean School	25.1	Q3	9.0	Q3	9.7	Q2

The table above places each Brighton & Hove school within the national distribution of key predictors. This context is essential for understanding the non-linear effects discussed in Section 7: schools in the lower quartiles sit on the steep part of the log curve where small absolute changes have the largest predicted impact.

12 An Alternative League Table: Value Added by School

Traditional school league tables rank schools by raw Attainment 8 scores. But this conflates school quality with school intake: schools serving affluent, well-prepared cohorts will always appear near the top, while schools facing greater challenges are penalised regardless of the value they add.

The model residual — observed ATT8 minus predicted ATT8 — offers an alternative. A positive residual means a school is achieving more than the model predicts given its intake composition, absence rates, prior attainment, workforce stability, and area characteristics. A negative residual means it is achieving less. This is conceptually similar to a value-added measure: how much is the school contributing beyond what would be expected from its circumstances?

NoteHow to read these tables

• Positive residuals (green) → the school is outperforming expectations. The larger the number, the more the school is adding above what the model predicts.
• Negative residuals (red) → the school is underperforming expectations. The school may still have good raw results, but its circumstances predict even better outcomes.
• Near zero → the school is performing roughly as the model expects.
• Schools are ranked by their mean residual across all available years for the panel model.

12.1 All Pupils

All Pupils: value-added rankings (panel model residuals by year)

Rank	School	2021-22	2022-23	2023-24	2024-25	Mean Residual
1	Dorothy Stringer School	+3.5	+3.8	+2.8	+5.2	+3.8
2	Varndean School	+4.0	+3.3	+6.0	+1.2	+3.6
3	King's School	+4.8	+0.1	+5.9	+0.8	+2.9
4	Portslade Aldridge Community Academy	+2.9	-0.4	+3.1	-0.5	+1.3
5	Brighton Aldridge Community Academy	+2.0	+1.2	+0.4	-0.0	+0.9
6	Blatchington Mill School	-2.2	-0.8	+2.5	+3.1	+0.6
7	Hove Park School and Sixth Form Centre	+0.6	+2.0	+0.4	-0.6	+0.6
8	Cardinal Newman Catholic School	+0.4	-1.7	+1.2	-1.3	-0.3
9	Longhill High School	+0.4	-2.2	-2.2	-3.4	-1.8
10	Patcham High School	-3.2	-2.7	-4.9	-3.4	-3.6

All Pupils: value-added rankings (per-year model residuals)

Rank	School	2021-22	2022-23	2023-24	2024-25	Mean Residual
1	Dorothy Stringer School	+3.9	+5.2	+4.2	+7.2	+5.1
2	Varndean School	+4.6	+4.5	+7.2	+3.1	+4.9
3	King's School	+6.3	+1.0	+6.9	+2.7	+4.2
4	Portslade Aldridge Community Academy	+3.4	+0.5	+4.4	+1.1	+2.3
5	Blatchington Mill School	-1.9	+0.2	+3.6	+5.1	+1.8
6	Brighton Aldridge Community Academy	+2.3	+2.2	+1.4	+1.2	+1.8
7	Hove Park School and Sixth Form Centre	+1.1	+2.9	+1.4	+0.9	+1.6
8	Cardinal Newman Catholic School	+0.6	-0.4	+2.8	+0.6	+0.9
9	Longhill High School	+0.5	-1.0	-1.1	-2.0	-0.9
10	Patcham High School	-2.1	-1.8	-3.8	-1.4	-2.3

12.2 Disadvantaged Pupils

Disadvantaged Pupils: value-added rankings (panel model residuals by year)

Rank	School	2021-22	2022-23	2023-24	2024-25	Mean Residual
1	Brighton Aldridge Community Academy	+3.2	-1.3	+4.1	+5.2	+2.8
2	Varndean School	+3.2	+3.6	+4.5	-1.9	+2.3
3	Dorothy Stringer School	+2.9	+3.2	+0.3	+1.7	+2.0
4	Cardinal Newman Catholic School	+2.4	-5.0	+6.5	+2.5	+1.6
5	King's School	+5.8	+0.9	+4.4	-5.4	+1.5
6	Blatchington Mill School	-5.4	+1.3	+5.2	+3.4	+1.1
7	Hove Park School and Sixth Form Centre	+2.4	-1.7	+1.3	+0.5	+0.6
8	Portslade Aldridge Community Academy	+2.6	-4.0	+4.7	-1.3	+0.5
9	Patcham High School	+3.8	+1.6	-11.9	-3.7	-2.6
10	Longhill High School	-4.6	-6.2	+0.8	-1.4	-2.9

Disadvantaged Pupils: value-added rankings (per-year model residuals)

Rank	School	2021-22	2022-23	2023-24	2024-25	Mean Residual
1	Brighton Aldridge Community Academy	+3.1	-0.2	+4.2	+5.8	+3.2
2	Varndean School	+3.6	+5.4	+4.5	-1.0	+3.1
3	Dorothy Stringer School	+3.0	+5.0	+0.5	+2.7	+2.8
4	King's School	+7.3	+2.9	+3.8	-4.4	+2.4
5	Cardinal Newman Catholic School	+2.1	-3.1	+7.0	+3.1	+2.3
6	Blatchington Mill School	-5.3	+3.0	+5.1	+4.7	+1.9
7	Hove Park School and Sixth Form Centre	+2.7	-0.4	+1.3	+1.3	+1.2
8	Portslade Aldridge Community Academy	+3.1	-2.8	+4.8	-0.4	+1.2
9	Patcham High School	+4.4	+3.1	-12.3	-2.8	-1.9
10	Longhill High School	-4.6	-4.8	+1.0	-0.5	-2.2

12.3 Non-Disadvantaged Pupils

Non-Disadvantaged Pupils: value-added rankings (panel model residuals by year)

Rank	School	2021-22	2022-23	2023-24	2024-25	Mean Residual
1	Dorothy Stringer School	+3.3	+3.5	+3.1	+5.1	+3.7
2	Varndean School	+3.7	+2.9	+6.0	+2.0	+3.7
3	King's School	+3.9	-0.6	+5.3	+2.0	+2.7
4	Portslade Aldridge Community Academy	+2.4	+0.4	+2.2	-0.4	+1.2
5	Brighton Aldridge Community Academy	+2.6	+4.5	-0.9	-2.2	+1.0
6	Hove Park School and Sixth Form Centre	-0.4	+3.1	-0.1	-0.9	+0.4
7	Blatchington Mill School	-2.3	-2.2	+0.7	+1.8	-0.5
8	Longhill High School	+3.5	-1.4	-3.8	-3.7	-1.4
9	Cardinal Newman Catholic School	-0.5	-1.6	-0.6	-2.8	-1.4
10	Patcham High School	-4.7	-4.4	-3.3	-3.3	-3.9

Non-Disadvantaged Pupils: value-added rankings (per-year model residuals)

Rank	School	2021-22	2022-23	2023-24	2024-25	Mean Residual
1	Dorothy Stringer School	+3.8	+4.8	+4.0	+6.5	+4.8
2	Varndean School	+4.4	+4.0	+6.8	+3.5	+4.7
3	King's School	+5.8	+0.4	+5.8	+3.4	+3.8
4	Portslade Aldridge Community Academy	+3.3	+1.3	+2.8	+0.8	+2.1
5	Brighton Aldridge Community Academy	+3.1	+5.7	-0.3	-1.3	+1.8
6	Hove Park School and Sixth Form Centre	+0.4	+4.0	+0.5	+0.2	+1.3
7	Blatchington Mill School	-1.8	-1.0	+1.1	+3.5	+0.5
8	Cardinal Newman Catholic School	-0.1	-0.4	+0.3	-1.4	-0.4
9	Longhill High School	+3.9	-0.0	-3.2	-2.7	-0.5
10	Patcham High School	-3.8	-3.4	-2.7	-1.8	-2.9

12.4 Summary: Combined Value-Added Heatmap

Value-added heatmap: panel model residuals for all Brighton & Hove schools, by year and pupil group. Green = outperforming, red = underperforming, white = as expected.

12.5 Orientation Map: Schools and Proposed Catchments

Brighton & Hove secondary schools with 2025/26 catchment boundaries. Click a logo or marker to see the school name.

The map above shows each Brighton & Hove secondary school at its geographic location, marked with the school’s logo, and overlaid with the 2025/26 catchment boundaries. Shared catchment areas (e.g. Varndean/Dorothy Stringer, Hove Park/Blatchington Mill) are shown as single zones. Click on any logo or dot for key school statistics.

12.6 School-by-School Commentary

Dorothy Stringer School

Good-rated school | Mean ATT8: 53.5 | Mean % FSM: 19.1% | Mean % Absence: 10.1%

Dorothy Stringer School is substantially outperforming the model’s expectations, with residuals consistently positive in every year (mean residual: +3.8 ATT8 points for all pupils). The residuals show an improving trend over time. The school’s value-added is weaker for disadvantaged pupils (mean residual +2.0) than for non-disadvantaged (+3.7), suggesting room for more targeted support.

Varndean School

Good-rated school | Mean ATT8: 53.5 | Mean % FSM: 21.6% | Mean % Absence: 9.2%

Varndean School is substantially outperforming the model’s expectations, with residuals consistently positive in every year (mean residual: +3.6 ATT8 points for all pupils). The residuals show no clear trend over time. The school’s value-added is weaker for disadvantaged pupils (mean residual +2.3) than for non-disadvantaged (+3.7), suggesting room for more targeted support.

King’s School

Good-rated school | Mean ATT8: 56.9 | Mean % FSM: 16.1% | Mean % Absence: 6.6%

King’s School is outperforming the model’s expectations, with residuals consistently positive in every year (mean residual: +2.9 ATT8 points for all pupils). The residuals show no clear trend over time. The school’s value-added is weaker for disadvantaged pupils (mean residual +1.5) than for non-disadvantaged (+2.7), suggesting room for more targeted support.

Portslade Aldridge Community Academy

Good-rated school | Mean ATT8: 45.9 | Mean % FSM: 27.8% | Mean % Absence: 11.7%

Portslade Aldridge Community Academy is outperforming the model’s expectations, with residuals positive in 2 of 4 years (mean residual: +1.3 ATT8 points for all pupils). The residuals show no clear trend over time.

Brighton Aldridge Community Academy

Requires Improvement-rated school | Mean ATT8: 36.8 | Mean % FSM: 47.9% | Mean % Absence: 15.8%

Brighton Aldridge Community Academy is performing broadly in line with the model’s expectations, with residuals positive in 3 of 4 years (mean residual: +0.9 ATT8 points for all pupils). The residuals show a declining trend over time. Notably, the school’s value-added for disadvantaged pupils (mean residual +2.8) is stronger than for non-disadvantaged pupils (+1.0), suggesting particularly effective support for its most vulnerable students.

Blatchington Mill School

Good-rated school | Mean ATT8: 51.0 | Mean % FSM: 17.3% | Mean % Absence: 10.1%

Blatchington Mill School is performing broadly in line with the model’s expectations, with residuals positive in 2 of 4 years (mean residual: +0.6 ATT8 points for all pupils). The residuals show an improving trend over time. Notably, the school’s value-added for disadvantaged pupils (mean residual +1.1) is stronger than for non-disadvantaged pupils (-0.5), suggesting particularly effective support for its most vulnerable students.

Hove Park School and Sixth Form Centre

Good-rated school | Mean ATT8: 45.2 | Mean % FSM: 31.3% | Mean % Absence: 11.8%

Hove Park School and Sixth Form Centre is performing broadly in line with the model’s expectations, with residuals positive in 3 of 4 years (mean residual: +0.6 ATT8 points for all pupils). The residuals show a declining trend over time.

Cardinal Newman Catholic School

Good-rated school | Mean ATT8: 53.1 | Mean % FSM: 17.2% | Mean % Absence: 9.0%

Cardinal Newman Catholic School is performing broadly in line with the model’s expectations, with residuals positive in 2 of 4 years (mean residual: -0.3 ATT8 points for all pupils). The residuals show no clear trend over time. Notably, the school’s value-added for disadvantaged pupils (mean residual +1.6) is stronger than for non-disadvantaged pupils (-1.4), suggesting particularly effective support for its most vulnerable students.

Longhill High School

Good-rated school | Mean ATT8: 37.2 | Mean % FSM: 33.6% | Mean % Absence: 14.7%

Longhill High School is underperforming the model’s expectations, with residuals positive in 1 of 4 years (mean residual: -1.8 ATT8 points for all pupils). The residuals show a declining trend over time. The school’s value-added is weaker for disadvantaged pupils (mean residual -2.9) than for non-disadvantaged (-1.4), suggesting room for more targeted support.

Patcham High School

Good-rated school | Mean ATT8: 47.8 | Mean % FSM: 20.4% | Mean % Absence: 8.8%

Patcham High School is substantially underperforming the model’s expectations, with residuals negative in every year observed (mean residual: -3.6 ATT8 points for all pupils). The residuals show no clear trend over time. Notably, the school’s value-added for disadvantaged pupils (mean residual -2.6) is stronger than for non-disadvantaged pupils (-3.9), suggesting particularly effective support for its most vulnerable students.

TipUsing this alternative league table

These residual-based rankings provide a fairer comparison between schools than raw Attainment 8 scores. A school ranked first in this table is not necessarily the highest-scoring school in Brighton & Hove, but it is the school that is adding the most value relative to what would be expected given its circumstances.

For the council and school improvement teams, the most actionable insights come from:

• Consistently positive schools — what are they doing differently that could be shared across the authority?
• Consistently negative schools — where might additional support or intervention be most impactful?
• Schools with divergent group residuals — where the value-added for disadvantaged pupils differs markedly from non-disadvantaged, either positively or negatively.
• Year-on-year trends — schools whose residuals are improving or declining over time may signal changing effectiveness or emerging challenges.

13 Interpretation: What the Model Tells Us About Brighton & Hove

13.1 The FSM sign-flip

The model reveals a striking pattern in how deprivation relates to attainment across different pupil groups. For the All Pupils and Non-Disadvantaged models, the coefficient on log(% FSM) is negative — schools with higher deprivation levels achieve lower predicted attainment, as expected. But for the Disadvantaged model, the coefficient flips to positive: schools with higher FSM rates actually receive a modest upward adjustment for their disadvantaged pupils’ predicted attainment.

In Brighton & Hove, this plays out clearly. Schools with lower FSM rates benefit from the negative coefficient for overall attainment (lower deprivation → higher predicted score). But for disadvantaged pupils specifically, schools with higher FSM rates — where disadvantaged pupils form a larger share of the cohort — get a slight prediction boost. The model is capturing the idea that disadvantaged pupils may do relatively better in schools where they are not a small, potentially marginalised minority.

TipPolicy implication

Schools with lower FSM proportions should consider whether their disadvantaged pupils receive sufficiently targeted support, or whether Pupil Premium resources are diluted across the whole cohort. The model suggests that concentrated, well-resourced intervention may be more effective than spread-thin approaches.

13.2 Absence: the biggest lever

Across all three pupil groups, overall absence rate is consistently one of the strongest predictors of Attainment 8. The log-transformed relationship means that gains from reducing absence are largest for schools with already-moderate absence rates — the steep part of the curve.

For Brighton & Hove schools with relatively low absence rates, even a small absolute increase (e.g. 4.5% to 5.5%) translates into a meaningful drop in predicted attainment. In contrast, a school already at 8% or 9% absence would need a larger absolute rise to see the same predicted impact.

This has two implications:

1. Protecting low-absence schools from deterioration is just as important as tackling chronic absence elsewhere
2. Early intervention before absence becomes entrenched is more efficient than remediation later

13.3 Workforce effects

The imputed full model includes teacher retention (remained_in_the_same_school) and teacher sickness days (average_number_of_days_taken) as predictors. In Brighton & Hove:

• Schools with higher staff stability receive a positive adjustment to predicted attainment
• Schools with higher teacher absence see a penalty

These effects are visible in the school-level residuals: schools that previously appeared to over- or under-perform the core model (which lacked workforce variables) may now be better explained by the additional workforce predictors. Where persistent residuals remain after controlling for workforce factors, they point to genuinely unmeasured factors — leadership quality, curriculum design, community resources — rather than being artefacts of missing controls.

13.4 What the residuals mean

Schools that consistently sit above the diagonal on the observed-vs-predicted plot are outperforming the model’s expectations given their intake. Those consistently below are underperforming relative to their statistical peers. With the imputed full model’s nine predictors, persistent residuals point to factors genuinely beyond the model’s scope.

NoteA note on 2024-25

Points shown as triangles (▲) in the plots use carry-forward imputed values for four predictors. These predictions should be treated with somewhat more caution than those based on directly observed data.

14 Variable Importance: National vs Brighton & Hove

14.1 National rankings

National variable importance: standardised coefficients (Analysis E)

Rank	display_name	All Pupils	Disadvantaged	Non-Disadvantaged	Avg |Std. Coef|
1	Overall Absence Rate	-0.0610	-0.0871	-0.0492	0.0658
2	% Low Prior Attainment	-0.0587	-0.0543	-0.0531	0.0554
3	% Disadvantaged (FSM)	-0.0415	0.0047	-0.0257	0.0240
4	% EAL	0.0066	0.0261	0.0090	0.0139
5	Teacher Retention	0.0106	0.0006	0.0098	0.0070
6	Teacher Sickness Days	-0.0057	-0.0072	-0.0068	0.0066
7	Leadership Pay %	-0.0053	-0.0038	-0.0055	0.0049
8	Gorard Segregation Index	-0.0015	-0.0002	-0.0025	0.0014

14.2 Brighton & Hove rankings

The same standardised-coefficient approach is repeated using only Brighton & Hove schools. Because the local predictor distributions may differ from the national picture (e.g. narrower absence range, different FSM spread), the variable importance rankings may shift.

Brighton & Hove variable importance: standardised coefficients (local predictor SDs, national model)

Rank	display_name	All Pupils	Disadvantaged	Non-Disadvantaged	Avg |Std. Coef|
1	Overall Absence Rate	-0.0544	-0.0778	-0.0439	0.0587
2	% Low Prior Attainment	-0.0400	-0.0370	-0.0362	0.0378
3	% Disadvantaged (FSM)	-0.0248	0.0028	-0.0154	0.0143
4	Teacher Retention	0.0146	0.0008	0.0135	0.0096
5	% EAL	0.0028	0.0112	0.0038	0.0059
6	Teacher Sickness Days	-0.0041	-0.0051	-0.0049	0.0047
7	Leadership Pay %	-0.0037	-0.0027	-0.0038	0.0034
8	Gorard Segregation Index	-0.0002	0.0000	-0.0003	0.0002

14.3 Comparison

All Pupils: national vs Brighton & Hove variable importance rankings

display_name	National Rank	National Std.Coef	B&H Rank	B&H Std.Coef	Rank Change
Overall Absence Rate	1	-0.0610	1	-0.0544	0
% Low Prior Attainment	2	-0.0587	2	-0.0400	0
% Disadvantaged (FSM)	3	-0.0415	3	-0.0248	0
% EAL	4	0.0066	5	0.0028	-1
Teacher Retention	5	0.0106	4	0.0146	1
Teacher Sickness Days	6	-0.0057	6	-0.0041	0
Leadership Pay %	7	-0.0053	7	-0.0037	0
Gorard Segregation Index	8	-0.0015	8	-0.0002	0

Disadvantaged Pupils: national vs Brighton & Hove variable importance rankings

display_name	National Rank	National Std.Coef	B&H Rank	B&H Std.Coef	Rank Change
Overall Absence Rate	1	-0.0871	1	-0.0778	0
% Low Prior Attainment	2	-0.0543	2	-0.0370	0
% EAL	3	0.0261	3	0.0112	0
Teacher Sickness Days	4	-0.0072	4	-0.0051	0
% Disadvantaged (FSM)	5	0.0047	5	0.0028	0
Leadership Pay %	6	-0.0038	6	-0.0027	0
Teacher Retention	7	0.0006	7	0.0008	0
Gorard Segregation Index	8	-0.0002	8	0.0000	0

14.3.1 Key differences between national and local rankings

The comparison tables highlight where Brighton & Hove’s local context shifts the relative importance of predictors:

• % Disadvantaged (FSM): Brighton & Hove schools span a wide FSM range (from under 10% to over 50%), which may amplify or dampen this variable’s local importance depending on the SD of log(FSM) in the authority.

• Overall Absence: If Brighton & Hove has a narrower spread of absence rates than the national distribution, absence will appear less important locally — not because it matters less per unit of change, but because schools are more similar to each other on this dimension.

• Teacher Retention and Sickness: Brighton & Hove’s workforce stability profile — shaped by coastal-city labour market conditions, London proximity for recruitment, and academy vs maintained school mix — may produce different local variance in these predictors.

• Gorard Segregation: As a single local authority, all B&H schools share the same segregation index value in any given year, so the within-B&H variance is zero or near-zero. This variable’s local importance will therefore be very low. Note that even at national level, the Gorard segregation coefficient is statistically insignificant in the imputed full model.

15 Policy Recommendations

The variable importance rankings, combined with the non-linear effects and Brighton & Hove’s local context, suggest several priority areas.

15.1 For school leaders

1. Absence reduction is the single biggest lever.

The log-transformed relationship means that gains from reducing absence are largest for schools with already-moderate absence rates. For a school currently at 6% absence, reducing to 5% has a larger predicted effect than a school at 10% reducing to 9%. This argues for intensive early-intervention strategies before absence becomes entrenched, and for protecting low-absence schools from deterioration.

2. Support for pupils with low prior attainment.

The % low prior attainment coefficient is strongly negative. While schools cannot change their intake, they can invest in targeted catch-up programmes at the KS2-KS3 transition, diagnostic assessment, and small-group intervention during KS3.

3. Workforce stability matters.

Teacher retention shows a positive association with attainment; teacher sickness shows a negative one. Retention strategies (competitive CPD, workload management, mentoring for early-career teachers) and staff wellbeing programmes can both contribute.

4. The FSM sign-flip for disadvantaged pupils.

Schools with lower FSM proportions should consider whether their disadvantaged pupils receive sufficiently targeted support, or whether Pupil Premium resources are diluted across the whole cohort.

15.2 For Brighton & Hove Council

1. Tackle absence strategically across the authority through coordinated data sharing, targeted outreach, and shared best practice between schools.

2. Address the disadvantaged attainment gap directly by ring-fencing and monitoring Pupil Premium impact, commissioning cross-school programmes, and tracking the FSM gap as a headline KPI using the model’s predicted-vs-actual framework.

3. A note on segregation: The Gorard segregation index is statistically insignificant in this model. While reducing segregation may be a worthwhile policy goal for broader equity reasons, this model does not provide reliable evidence that it would improve Attainment 8 scores at school level.

4. Support workforce stability through LA-wide recruitment initiatives, monitoring teacher turnover across schools, and ensuring schools in challenging contexts receive additional staffing support.

5. Use the model to target support via the Shiny app’s policy simulator, which allows exploration of school-level what-if scenarios. The residual analysis identifies schools outperforming expectations — learn from their practices — and those underperforming — target support where it is most needed.

WarningImportant caveat

These recommendations are based on associations identified by the multilevel model. The model controls for observable confounders but cannot establish causation. A school reducing absence may not see exactly the predicted ATT8 gain if other unmeasured factors change simultaneously. The rankings and effect sizes should be treated as evidence-informed starting points for policy discussion, not as causal guarantees.