verification / datasets / 08-national-fraud-initiative / VARIABLES.md

Variables — National Fraud Initiative (NFI)

Honest limitation up front: the NFI publishes no per-council data. The Cabinet Office / PSFA reports (data/NFI_report_2022-24.pdf, data/NFI_report_2020-22_dec2022.pdf) contain only national totals; per-council match outcomes live in the secure NFI web application and reach the public, if at all, through each council's own audit-committee papers (see data/samples/). So Section 1 is essentially empty at council level; Section 2 lists what a bespoke collection would yield.

1. Extracted — available today

Per-council: nothing from the central publications. What can actually be extracted, all at England (or UK/nation) level per biennial exercise:

Variable Description Source
nfi_total_outcomes_england_gbp Total outcomes (detected + estimated future losses prevented), e.g. £477.5m for 2022–24 Report front section / Appendix 2
nfi_outcomes_by_risk_area_gbp Detected / estimated-prevented / total, by risk area Appendix 2 (p.32, 2022–24 report), with 2020–22 comparators
nfi_total_matches Number of matches per exercise (e.g. 309,593 in 2022–24) Trend table
nfi_recovery_rate_pct Recovered vs not-recovered split of detected overpayments Recovery-rates page

Risk areas in Appendix 2 (2022–24, England, £m total): Pensions 179.5; Council tax single person discount 51.9; Housing benefits 29.7; Blue badges 26.8; Housing waiting lists 17.6; Trade creditors 11.1; Housing tenancy 7.8; Council tax reduction 7.4; Concessionary travel passes 4.0; Right to Buy 1.7; Business validation 1.0; Payroll/Procurement 0.2; Pilots 0.7; Other 0.4; Private sector 137.8; residential care homes (nil 2022–24). Note: totals mix cash detected with estimated future losses prevented (methodology in Appendix 3) — not pure recovered money.

These support national context/benchmarks only, not a per-council metric.

2. Constructed / Prospective — per-council, via bespoke collection

Source routes: (a) scrape ~300+ councils' audit/governance committee papers (ModernGov etc.) for biennial NFI update reports — the data/samples/ Audit Wales papers show the typical shape (per-council matches by data-match area, matches in "key reports"; outcomes usually only at national/Wales level, so outcome values are patchier); (b) FOI/data request to the PSFA (England) or Audit Wales / Audit Scotland / NIAO for per-participant extracts from the NFI web application.

Variable Formula Source route
nfi_matches_total Data matches released to the council per exercise (Bridgend 2024–25: 17,255) (a) committee papers or (b) PSFA/audit-body request
nfi_matches_by_area Matches per data-match area (SPD, creditors, blue badges, payroll, …) same
nfi_key_report_matches Matches in "key reports" with historically high hit rates (a) — present in Welsh sample papers
nfi_matches_investigated / nfi_matches_closed Matches given a closed status per exercise; investigation rate = closed / total (b) — web-app statuses; some committee papers
nfi_outcomes_gbp Fraud/overpayment value recorded per exercise (detected + estimated prevented) (b) preferred; (a) where councils publish their own figure
nfi_detected_gbp Actual detected fraud/error only (excluding estimated prevention) (b)
nfi_outcomes_per_million_spend nfi_outcomes_gbp / (service_expenditure_gbp / 1e6) using net revenue expenditure from ../01-revenue-outturn derived
nfi_investigation_rate nfi_matches_closed / nfi_matches_total derived
nfi_hit_rate nfi_outcomes_gbp / nfi_matches_investigated derived

Comparability caveats for any per-council build: outcome values mix actual and estimated elements; investigation effort is discretionary (risk-based, resource-driven), so low outcomes can mean low fraud or low investigation; periods and formats vary across committee papers; Wales/Scotland/NI run parallel exercises under their own audit bodies.

Variable index

One file per variable in variables/; frontmatter kind marks each as extracted (national publications) or prospective (per-council, bespoke collection only).

Extracted (national-level only):

Prospective (per-council, no central publication):