verification / datasets / 04-efs-s114 / VARIABLES.md

Variables: EFS awards + section 114 notices

Per-council variables this dataset yields. All file paths are relative to data/. Council identity is a free-text name only (col1 in the EFS files, council in the s114 file) — no GSS/E-codes — so joining to sibling datasets requires name matching.

Parsing notes (verified against files)

EFS files (efs-<year>.csv, stacked in efs-all-years.csv) have columns year, table_index, col1, col2 (the stacked file pads to col5, always empty). col1 = local authority name, col2 = amount as free text (e.g. £55.7m (support agreed in-principle), Council provided with in-principle support but withdrew its request). Amounts must be regex-parsed; some rows carry revision notes (£405.7m (from £491.1m ...)).
Header rows are embedded in the data: every table repeats a row with col1 = "Local authority" — drop these.
table_index matters: table 0 is the main per-council list for that year. In efs-2024-25.csv and efs-2025-26.csv, tables 1–2 are additional per-council rows (Norfolk; Lambeth; South Yorkshire MCA police functions). In efs-2026-27.csv, table 2 is not per-council: col1 holds financial years — it is Birmingham City Council's amended prior-year support profile. Exclude it from per-council extraction (or attribute all its rows to Birmingham with the year taken from col1).
s114-notices-compiled.csv is hand-compiled and unofficial (no government register exists): columns council, notice_date, section, notes, source; 10 notices, 8 councils, 2018-02-02 to 2023-11-29. Coverage after Nov 2023 unconfirmed. Treat as indicative, not authoritative.

1. Extracted variables

Variable	File / column (verified)	Unit	Interpretation
`efs_awarded_in_year`	`efs-<year>.csv` (or `efs-all-years.csv` filtered on `year`): council in `col1`, presence of a non-header row	boolean per council-year	Council was granted (or requested) Exceptional Financial Support that year — an official distress flag; a bailout in all but name.
`efs_amount_in_year`	same rows, amount regex-parsed from `col2` (`£X.Xm` pattern; latest figure when a `(from £Y ...)` revision note is present)	£ million per council-year	Size of capitalisation support agreed in principle. Not necessarily drawn down; some entries are HRA support or withdrawn requests (Bexley 2020-21) — the free text says which.
`efs_withdrawn_or_conditional`	same rows where `col2` has no parseable `£` figure or carries qualifying text	boolean / text	Requested but withdrew, or support conditional/in-principle only.
`efs_years_count`	`efs-all-years.csv`: count of distinct `year` values per `col1` council	count (0–7)	Years on EFS, 2020-21 to 2026-27. Repeat appearances signal chronic rather than one-off distress.
`efs_first_year`	`efs-all-years.csv`: min `year` per council	financial year	When distress first became official.
`efs_cumulative_amount`	sum of `efs_amount_in_year` over years per council	£ million	Total EFS agreed in principle across 2020-21..2026-27. Sum latest revised figures; exclude the 2026-27 table 2 restatement or double-counting of Birmingham results.
`s114_issued`	`s114-notices-compiled.csv`: council appears in `council`	boolean	Council's s151 officer declared the budget cannot be balanced — effective bankruptcy. Hand-compiled, unofficial list.
`s114_count`	same file: rows per `council`	count	Number of notices (Croydon 3, Northamptonshire 2, others 1).
`s114_dates` / `s114_first_date`	`notice_date` column (ISO dates)	date(s)	When distress was declared; `section` is s114(3) throughout; `notes` gives the reported cause (equal pay, investment losses, ...).

2. Constructed variables

Only inputs verified to exist are used. Sibling references: dataset 01-revenue-outturn (research/datasets/01-revenue-outturn/data/RSX_LA_Data_2024-25_data_by_LA.ods etc.) provides per-council revenue expenditure; join is by council name.

Variable	Formula	Inputs	Unit	Interpretation
`efs_share_of_revenue_spend`	`efs_amount_in_year / net_revenue_expenditure`	this dataset + revenue expenditure from `01-revenue-outturn` (RSX/RS outturn by LA), matched year	ratio (%)	Scales the bailout to council size: £40m is existential for a district, marginal for a county.
`cumulative_efs_share_of_spend`	`efs_cumulative_amount / net_revenue_expenditure (latest year)`	as above	ratio (%)	Depth of accumulated dependence on exceptional support.
`efs_ever`	`efs_years_count > 0`	this dataset	boolean	Any official EFS distress flag, any year.
`formal_distress_flag`	`efs_ever OR s114_issued`	this dataset	boolean	Union of the two official/reported distress signals. s114 component is from the unofficial hand-compiled list.
`distress_severity_score`	e.g. `2*s114_count + efs_years_count` (ordinal, tunable)	this dataset	ordinal score	Ranks councils by breadth of formal distress; s114 weighted above EFS since it is the harder signal.
`efs_trajectory`	sign of `efs_amount_in_year(t) - efs_amount_in_year(t-1)` for councils with ≥2 EFS years	this dataset	−/0/+	Whether reliance on exceptional support is growing or shrinking.
`s114_to_efs_transition`	`s114_issued AND efs_years_count > 0`, with `efs_first_year` vs `s114_first_date` ordering	this dataset	boolean	Councils whose bankruptcy was followed by the EFS regime (post-2020 s114s have largely been substituted by EFS).

Caveats for scoring

EFS amounts are "agreed in principle", revised retroactively on GOV.UK; re-fetch before final analysis (SOURCE.md re-download script).
Absence from the s114 file is weak evidence of absence: it is a hand-compiled 2018–2023 list, unverified for 2024–2026.
2025-26 includes one non-council row (South Yorkshire MCA police functions); exclude or handle separately.

Variable index

One file per variable in variables/: