CFR & AAR financial returns matched to the KS4 panel dataset
School financial data comes from the DfE Financial Benchmarking and Insights Tool, which publishes two types of returns:
Both are published as Excel workbooks, one per academic year. The data covers income, expenditure, staffing costs, and balance sheet items for every state-funded school in England.
The full financial_returns.rds file contains all schools (primary, secondary, special), not just KS4 schools in the panel.
finance %>%
count(academic_year, source) %>%
pivot_wider(names_from = source, values_from = n, values_fill = 0) %>%
mutate(Total = CFR + AAR) %>%
kable(format.args = list(big.mark = ","))| academic_year | AAR | CFR | Total |
|---|---|---|---|
| 2021-2022 | 10,001 | 12,390 | 22,391 |
| 2022-2023 | 10,292 | 12,023 | 22,315 |
| 2023-2024 | 10,955 | 11,537 | 22,492 |
| 2024-2025 | 0 | 10,701 | 10,701 |
tibble(
Metric = c("Total records", "Unique schools (URNs)", "Years covered",
"Variables", "File size"),
Value = c(
comma(nrow(finance)),
comma(n_distinct(finance$URN)),
paste(sort(unique(finance$academic_year)), collapse = ", "),
as.character(ncol(finance)),
paste0(round(file.size(here::here("data", "financial_returns.rds")) / 1e6, 1), " MB")
)
) %>% kable()| Metric | Value |
|---|---|
| Total records | 77,899 |
| Unique schools (URNs) | 23,485 |
| Years covered | 2021-2022, 2022-2023, 2023-2024, 2024-2025 |
| Variables | 35 |
| File size | 9.3 MB |
AAR data for 2024-25 has not yet been published by the DfE, so only CFR (LA maintained) schools are available for that year.
The harmonised dataset extracts comparable fields from both CFR and AAR returns.
var_info <- tribble(
~Variable, ~Description, ~Type,
"URN", "School unique reference number", "Identifier",
"school_name", "School name", "Identifier",
"la_code / la_name", "Local authority code and name", "Identifier",
"academic_year", "Academic year (e.g. 2023-2024)", "Identifier",
"source", "Data source: CFR or AAR", "Identifier",
"phase", "School phase (Primary, Secondary, etc.)", "Characteristic",
"num_pupils", "Number of pupils (FTE)", "Characteristic",
"num_teachers", "Number of teachers (FTE)", "Characteristic",
"pct_fsm", "% pupils eligible for FSM", "Characteristic",
"total_income", "Total income (GBP)", "Income",
"grant_funding", "Grant funding from government", "Income",
"self_generated", "Self-generated income (lettings, catering, etc.)", "Income",
"pupil_premium", "Pupil premium income (CFR only)", "Income",
"total_expenditure", "Total expenditure (GBP)", "Expenditure",
"staff_costs", "Total staff costs", "Expenditure",
"teaching_staff", "Teaching staff costs", "Expenditure",
"premises_costs", "Premises costs", "Expenditure",
"energy", "Energy costs", "Expenditure",
"learning_resources", "Learning resources (not ICT)", "Expenditure",
"supply_staff", "Supply staff costs", "Expenditure",
"catering_costs", "Catering expenses", "Expenditure",
"educational_supplies", "Educational supplies total", "Expenditure",
"bought_in_professional", "Bought-in professional services", "Expenditure",
"in_year_balance", "In-year balance (income - expenditure)", "Balance",
"revenue_reserve", "Revenue reserve", "Balance",
"total_income_pp", "Total income per pupil", "Per-pupil",
"total_expenditure_pp", "Total expenditure per pupil", "Per-pupil",
"staff_costs_pp", "Staff costs per pupil", "Per-pupil",
"teaching_staff_pp", "Teaching staff costs per pupil", "Per-pupil",
"premises_costs_pp", "Premises costs per pupil", "Per-pupil",
"grant_funding_pp", "Grant funding per pupil", "Per-pupil",
"staff_income_ratio", "Staff costs / total income", "Ratio",
"teaching_expenditure_ratio", "Teaching staff / total expenditure", "Ratio",
"balance_income_ratio", "In-year balance / total income", "Ratio"
)
var_info %>% kable()| Variable | Description | Type |
|---|---|---|
| URN | School unique reference number | Identifier |
| school_name | School name | Identifier |
| la_code / la_name | Local authority code and name | Identifier |
| academic_year | Academic year (e.g. 2023-2024) | Identifier |
| source | Data source: CFR or AAR | Identifier |
| phase | School phase (Primary, Secondary, etc.) | Characteristic |
| num_pupils | Number of pupils (FTE) | Characteristic |
| num_teachers | Number of teachers (FTE) | Characteristic |
| pct_fsm | % pupils eligible for FSM | Characteristic |
| total_income | Total income (GBP) | Income |
| grant_funding | Grant funding from government | Income |
| self_generated | Self-generated income (lettings, catering, etc.) | Income |
| pupil_premium | Pupil premium income (CFR only) | Income |
| total_expenditure | Total expenditure (GBP) | Expenditure |
| staff_costs | Total staff costs | Expenditure |
| teaching_staff | Teaching staff costs | Expenditure |
| premises_costs | Premises costs | Expenditure |
| energy | Energy costs | Expenditure |
| learning_resources | Learning resources (not ICT) | Expenditure |
| supply_staff | Supply staff costs | Expenditure |
| catering_costs | Catering expenses | Expenditure |
| educational_supplies | Educational supplies total | Expenditure |
| bought_in_professional | Bought-in professional services | Expenditure |
| in_year_balance | In-year balance (income - expenditure) | Balance |
| revenue_reserve | Revenue reserve | Balance |
| total_income_pp | Total income per pupil | Per-pupil |
| total_expenditure_pp | Total expenditure per pupil | Per-pupil |
| staff_costs_pp | Staff costs per pupil | Per-pupil |
| teaching_staff_pp | Teaching staff costs per pupil | Per-pupil |
| premises_costs_pp | Premises costs per pupil | Per-pupil |
| grant_funding_pp | Grant funding per pupil | Per-pupil |
| staff_income_ratio | Staff costs / total income | Ratio |
| teaching_expenditure_ratio | Teaching staff / total expenditure | Ratio |
| balance_income_ratio | In-year balance / total income | Ratio |
When joined to the KS4 panel (secondary schools only), the match rate depends on whether both CFR and AAR data are available.
panel %>%
group_by(`Academic Year` = academic_year) %>%
summarise(
`Panel Schools` = n(),
`With Finance` = sum(!is.na(fin_total_income)),
`Missing` = sum(is.na(fin_total_income)),
`Match %` = percent(mean(!is.na(fin_total_income)), accuracy = 0.1),
.groups = "drop"
) %>%
kable(format.args = list(big.mark = ","))| Academic Year | Panel Schools | With Finance | Missing | Match % |
|---|---|---|---|---|
| 2021-2022 | 4,292 | 3,297 | 995 | 76.8% |
| 2022-2023 | 4,281 | 3,305 | 976 | 77.2% |
| 2023-2024 | 4,292 | 3,341 | 951 | 77.8% |
| 2024-2025 | 4,265 | 550 | 3,715 | 12.9% |
panel %>%
filter(!is.na(fin_total_income)) %>%
count(academic_year, fin_source) %>%
ggplot(aes(academic_year, n, fill = fin_source)) +
geom_col(position = "stack") +
scale_fill_manual(values = c(AAR = "#2e6260", CFR = "#e16fca"), name = "Source") +
labs(x = "Academic Year", y = "Schools matched",
title = "Panel schools with financial data") +
geom_text(aes(label = comma(n)), position = position_stack(vjust = 0.5),
size = 3.5, colour = "white", fontface = "bold")
panel_fin %>%
summarise(
across(
c(fin_total_income, fin_total_expenditure, fin_in_year_balance,
fin_grant_funding, fin_staff_costs, fin_teaching_staff,
fin_premises_costs, fin_energy),
list(
Median = ~median(.x, na.rm = TRUE),
Mean = ~mean(.x, na.rm = TRUE),
SD = ~sd(.x, na.rm = TRUE),
Min = ~min(.x, na.rm = TRUE),
Max = ~max(.x, na.rm = TRUE)
),
.names = "{.col}|{.fn}"
)
) %>%
pivot_longer(everything(), names_to = c("Variable", "Stat"), names_sep = "\\|") %>%
pivot_wider(names_from = Stat, values_from = value) %>%
mutate(
Variable = str_remove(Variable, "^fin_"),
across(Median:Max, ~comma(.x, accuracy = 1, prefix = "\u00A3"))
) %>%
kable()| Variable | Median | Mean | SD | Min | Max |
|---|---|---|---|---|---|
| total_income | £7,605,000 | £7,786,230 | £3,003,794 | -£22,964 | £26,602,000 |
| total_expenditure | £7,250,120 | £7,467,847 | £2,946,836 | £0 | £27,265,000 |
| in_year_balance | £226,000 | £318,383 | £640,361 | -£4,856,000 | £11,196,000 |
| grant_funding | £7,398,000 | £7,548,952 | £2,890,912 | -£23,286 | £26,269,000 |
| staff_costs | £5,481,000 | £5,617,791 | £2,227,195 | £0 | £21,081,000 |
| teaching_staff | £3,994,000 | £4,100,679 | £1,692,002 | £0 | £14,864,000 |
| premises_costs | £406,000 | £491,929 | £393,389 | -£864,000 | £11,574,000 |
| energy | £154,000 | £167,692 | £108,582 | -£301,000 | £1,030,755 |
panel_fin %>%
summarise(
across(
c(fin_total_income_pp, fin_total_expenditure_pp, fin_staff_costs_pp,
fin_teaching_staff_pp, fin_grant_funding_pp),
list(
Median = ~median(.x, na.rm = TRUE),
Mean = ~mean(.x, na.rm = TRUE),
Q25 = ~quantile(.x, 0.25, na.rm = TRUE),
Q75 = ~quantile(.x, 0.75, na.rm = TRUE)
),
.names = "{.col}|{.fn}"
)
) %>%
pivot_longer(everything(), names_to = c("Variable", "Stat"), names_sep = "\\|") %>%
pivot_wider(names_from = Stat, values_from = value) %>%
mutate(
Variable = str_remove(Variable, "^fin_") %>% str_replace_all("_", " "),
across(Median:Q75, ~comma(.x, accuracy = 1, prefix = "\u00A3"))
) %>%
kable()| Variable | Median | Mean | Q25 | Q75 |
|---|---|---|---|---|
| total income pp | £7,206 | £7,390 | £6,591 | £7,988 |
| total expenditure pp | £6,910 | £7,086 | £6,256 | £7,700 |
| staff costs pp | £5,195 | £5,308 | £4,756 | £5,747 |
| teaching staff pp | £3,793 | £3,853 | £3,471 | £4,176 |
| grant funding pp | £6,987 | £7,164 | £6,387 | £7,752 |
panel_fin %>%
filter(fin_total_income_pp > 0, fin_total_income_pp < 30000) %>%
ggplot(aes(fin_total_income_pp, fill = academic_year)) +
geom_density(alpha = 0.4) +
scale_x_continuous(labels = label_comma(prefix = "\u00A3")) +
scale_fill_manual(values = c("#2e6260", "#abc766", "#f9dd73", "#e16fca", "#4e3c56", "#49a0c4", "#ff7f89"), name = "Year") +
labs(x = "Total income per pupil", y = "Density",
title = "Income per pupil has shifted upward over time") +
theme(legend.position = "bottom")
panel_fin %>%
filter(fin_total_expenditure_pp > 0, fin_total_expenditure_pp < 30000) %>%
ggplot(aes(fin_source, fin_total_expenditure_pp, fill = fin_source)) +
geom_violin(alpha = 0.6, draw_quantiles = c(0.25, 0.5, 0.75)) +
scale_y_continuous(labels = label_comma(prefix = "\u00A3")) +
scale_fill_manual(values = c(AAR = "#2e6260", CFR = "#e16fca"), guide = "none") +
labs(x = "Financial return type", y = "Expenditure per pupil",
title = "Academies (AAR) and LA schools (CFR) have similar per-pupil spending")
panel_fin %>%
filter(!is.na(fin_total_expenditure), fin_total_expenditure > 0) %>%
summarise(
`Teaching staff` = median(fin_teaching_staff / fin_total_expenditure, na.rm = TRUE),
`Other staff` = median((fin_staff_costs - fin_teaching_staff) / fin_total_expenditure, na.rm = TRUE),
`Premises` = median(fin_premises_costs / fin_total_expenditure, na.rm = TRUE),
`Energy` = median(fin_energy / fin_total_expenditure, na.rm = TRUE),
`Other` = 1 - median(fin_staff_costs / fin_total_expenditure, na.rm = TRUE) -
median(fin_premises_costs / fin_total_expenditure, na.rm = TRUE)
) %>%
pivot_longer(everything(), names_to = "Category", values_to = "Share") %>%
mutate(
Category = fct_reorder(Category, Share),
label = percent(Share, accuracy = 0.1)
) %>%
ggplot(aes(Share, Category, fill = Category)) +
geom_col(show.legend = FALSE) +
geom_text(aes(label = label), hjust = -0.1, size = 4) +
scale_x_continuous(labels = percent, limits = c(0, 0.65)) +
scale_fill_manual(values = c("#2e6260", "#abc766", "#f9dd73", "#e16fca", "#4e3c56", "#49a0c4", "#ff7f89")) +
labs(x = "Share of total expenditure", y = NULL,
title = "Teaching staff dominate school spending")
trends <- panel_fin %>%
group_by(academic_year) %>%
summarise(
`Income pp` = median(fin_total_income_pp, na.rm = TRUE),
`Expenditure pp` = median(fin_total_expenditure_pp, na.rm = TRUE),
`Staff costs pp` = median(fin_staff_costs_pp, na.rm = TRUE),
`Grant funding pp` = median(fin_grant_funding_pp, na.rm = TRUE),
.groups = "drop"
) %>%
pivot_longer(-academic_year, names_to = "Metric", values_to = "Value")
ggplot(trends, aes(academic_year, Value, colour = Metric, group = Metric)) +
geom_line(linewidth = 1.2) +
geom_point(size = 3) +
scale_y_continuous(labels = label_comma(prefix = "\u00A3")) +
scale_colour_manual(values = c("#2e6260", "#abc766", "#f9dd73", "#e16fca", "#4e3c56", "#49a0c4", "#ff7f89")) +
labs(x = "Academic Year", y = "Per pupil (median, GBP)",
title = "Per-pupil income and expenditure trends",
colour = NULL) +
theme(legend.position = "bottom")
panel_fin %>%
filter(!is.na(fin_balance_income_ratio),
abs(fin_balance_income_ratio) < 0.5) %>%
ggplot(aes(fin_balance_income_ratio, fill = academic_year)) +
geom_histogram(bins = 50, alpha = 0.6, position = "identity") +
geom_vline(xintercept = 0, linetype = "dashed", colour = "red") +
scale_x_continuous(labels = percent) +
scale_fill_manual(values = c("#2e6260", "#abc766", "#f9dd73", "#e16fca", "#4e3c56", "#49a0c4", "#ff7f89"), name = "Year") +
labs(x = "In-year balance / total income",
y = "Number of schools",
title = "Most schools run a small surplus; some are in deficit") +
theme(legend.position = "bottom")
panel_fin %>%
filter(!is.na(fin_in_year_balance)) %>%
group_by(`Academic Year` = academic_year) %>%
summarise(
Schools = n(),
`In Deficit` = sum(fin_in_year_balance < 0),
`% in Deficit` = percent(mean(fin_in_year_balance < 0), accuracy = 0.1),
`Median Deficit (GBP)` = comma(
median(fin_in_year_balance[fin_in_year_balance < 0]),
prefix = "\u00A3"
),
.groups = "drop"
) %>%
kable()| Academic Year | Schools | In Deficit | % in Deficit | Median Deficit (GBP) |
|---|---|---|---|---|
| 2021-2022 | 3297 | 629 | 19.1% | -£155,873 |
| 2022-2023 | 3305 | 853 | 25.8% | -£152,000 |
| 2023-2024 | 3341 | 967 | 28.9% | -£194,000 |
| 2024-2025 | 550 | 298 | 54.2% | -£208,461 |
panel_fin %>%
filter(academic_year == "2023-2024",
!is.na(ATT8SCR), ATT8SCR > 0,
fin_total_income_pp > 2000, fin_total_income_pp < 25000) %>%
ggplot(aes(fin_total_income_pp, ATT8SCR)) +
geom_point(aes(colour = fin_source), alpha = 0.3, size = 1.5) +
geom_smooth(method = "loess", se = TRUE, colour = "black", linewidth = 1) +
scale_x_continuous(labels = label_comma(prefix = "\u00A3")) +
scale_colour_manual(values = c(AAR = "#2e6260", CFR = "#e16fca"), name = "Source") +
labs(x = "Total income per pupil", y = "Attainment 8 score",
title = "Weak relationship between per-pupil income and attainment") +
theme(legend.position = "bottom")
panel_fin %>%
filter(academic_year == "2023-2024",
!is.na(ATT8SCR), ATT8SCR > 0,
fin_staff_income_ratio > 0.4, fin_staff_income_ratio < 1.1) %>%
ggplot(aes(fin_staff_income_ratio, ATT8SCR)) +
geom_point(alpha = 0.3, size = 1.5, colour = "#49a0c4") +
geom_smooth(method = "loess", se = TRUE, colour = "black", linewidth = 1) +
scale_x_continuous(labels = percent) +
labs(x = "Staff costs / total income", y = "Attainment 8 score",
title = "Schools spending a higher share on staff tend to have lower attainment") +
theme(legend.position = "bottom")
panel_fin %>%
filter(academic_year == "2023-2024",
!is.na(PTFSM6CLA1A), !is.na(fin_total_income_pp),
fin_total_income_pp > 2000, fin_total_income_pp < 25000) %>%
mutate(PTFSM6CLA1A = as.numeric(PTFSM6CLA1A)) %>%
filter(!is.na(PTFSM6CLA1A)) %>%
ggplot(aes(PTFSM6CLA1A, fin_total_income_pp)) +
geom_point(alpha = 0.3, size = 1.5, colour = "#e16fca") +
geom_smooth(method = "loess", se = TRUE, colour = "black", linewidth = 1) +
scale_y_continuous(labels = label_comma(prefix = "\u00A3")) +
labs(x = "% disadvantaged pupils (FSM6CLA1A)",
y = "Total income per pupil",
title = "Schools with more disadvantaged pupils receive higher per-pupil funding") +
theme(legend.position = "bottom")
panel_fin %>%
filter(academic_year == "2023-2024", !is.na(gor_name),
!is.na(fin_total_income_pp)) %>%
group_by(Region = gor_name) %>%
summarise(
n = n(),
median_pp = median(fin_total_income_pp, na.rm = TRUE),
.groups = "drop"
) %>%
filter(n >= 10) %>%
mutate(Region = fct_reorder(Region, median_pp)) %>%
ggplot(aes(median_pp, Region)) +
geom_col(fill = "#49a0c4", alpha = 0.8) +
geom_text(aes(label = comma(median_pp, prefix = "\u00A3", accuracy = 1)),
hjust = -0.1, size = 3.5) +
scale_x_continuous(labels = label_comma(prefix = "\u00A3"),
limits = c(0, NA), expand = expansion(mult = c(0, 0.15))) +
labs(x = "Median income per pupil", y = NULL,
title = "London schools receive significantly more per-pupil funding")
cor_vars <- c("ATT8SCR", "PTFSM6CLA1A", "TOTPUPS",
"fin_total_income_pp", "fin_total_expenditure_pp",
"fin_staff_costs_pp", "fin_grant_funding_pp",
"fin_staff_income_ratio", "fin_balance_income_ratio")
cor_data <- panel_fin %>%
filter(academic_year == "2023-2024") %>%
select(all_of(cor_vars)) %>%
mutate(across(everything(), as.numeric)) %>%
drop_na()
cor_labels <- c(
ATT8SCR = "Attainment 8",
PTFSM6CLA1A = "% Disadvantaged",
TOTPUPS = "Total Pupils",
fin_total_income_pp = "Income pp",
fin_total_expenditure_pp = "Expenditure pp",
fin_staff_costs_pp = "Staff costs pp",
fin_grant_funding_pp = "Grant funding pp",
fin_staff_income_ratio = "Staff/Income ratio",
fin_balance_income_ratio = "Balance/Income ratio"
)
cor_mat <- cor(cor_data, use = "pairwise.complete.obs")
rownames(cor_mat) <- cor_labels[rownames(cor_mat)]
colnames(cor_mat) <- cor_labels[colnames(cor_mat)]
# Simple heatmap
cor_df <- as.data.frame(as.table(cor_mat)) %>%
rename(Var1 = 1, Var2 = 2, Correlation = 3)
ggplot(cor_df, aes(Var1, Var2, fill = Correlation)) +
geom_tile(colour = "white") +
geom_text(aes(label = sprintf("%.2f", Correlation)), size = 3) +
scale_fill_gradient2(low = "#7b132b", mid = "#f3f3f3", high = "#2e6260",
midpoint = 0, limits = c(-1, 1)) +
labs(x = NULL, y = NULL, title = "Correlation matrix: finance and attainment") +
theme(axis.text.x = element_text(angle = 45, hjust = 1))
