Project Anchor — objective + scope

The canonical, single anchor for the UK Regulatory Burden Measurement project. It fixes both what the project counts (the objective: aim, unit, methodological commitments, output) and the scope boundary (which layers of binding rules the measure reaches, how content is typed, and how it is counted at the edge). Subsequent methodological, scoping, or technical decisions must be anchored here; refer back explicitly when proposing changes.

Reconciled 2026-07-16. This document absorbs the former project_scope_anchor.md (created 2026-07-14 in the v2.5 scope batch; extended 2026-07-15 with the v2.6 capacity-axis batch). The separate scope-anchor file is retired — this is now the one anchor. Items from the original objective anchor that were overtaken by later decisions are superseded inline and flagged (superseded …).

Core aim

To produce the first comprehensive, validated measure of prescriptive regulatory burden falling on private actors across the entire UK statute book.

Unit of measurement

The count of legally binding obligations and prohibitions imposed on private actors — businesses, individuals, and third-sector organisations — by UK legislation. Not a word count, page count, or document count: a count of provisions that impose compliance cost on non-governmental actors.

What counts — scope ruling (statutory private-law duties, both sides)

The measure counts statutory obligations and prohibitions as defined in the rubric, not a pre-sorted “regulation” subset. A statutory obligation or prohibition with legal force on a private actor counts whether its parent statute is “regulatory” or private-law, whether it binds the trader or the consumer side, and whether it is a mandatory term or a displaceable default (e.g. CRA 2015 s.25 — the consumer’s duty to pay at the contract rate counts).

Deeming distinction. Interpretive deeming (labels something for a scheme’s purposes — “treated as hazardous for the purposes of any legislation”) is excluded. Substance-creating deeming (a statutory implied/deemed contract term bringing a live duty between private parties into existence — “every contract is to be treated as including a term that…”) is counted and re-attributed to the obligated party (CRA 2015 s.13 → 1 trader burden vs Commission Decision 2000/532 → 0).

Four key methodological commitments

1. Private actor focus. Systematically excluding obligations on public bodies (ministers, regulators, courts, named officeholders). The public/private actor distinction is the bedrock filter; everything else builds on it.

2. Six-category taxonomy (numeric IDs 1–6 canonical, rename-proof; names presentational — see category_mapping.md). Every classified provision falls into one of:
   • 1 direct — a standing obligation/prohibition on a private actor
   • 2 conditional_direct — Conditional burden (operational): triggered by an event within the actor’s own activities
   • 3 implied_burden — obligation revealed through a defence provision (“it is a defence to prove…”)
   • 4 implied_burden_active — implied burden requiring an active, pre-existing compliance programme (“adequate procedures”)
   • 5 conditional_burden — Conditional burden (regulator-triggered): triggered by an external regulator’s act (notice, inspection, order)
   • 6 ambiguous — genuine category uncertainty, flagged for review

   Plus a polarity attribute (obligation / prohibition / review) on every burden (rubric §2A). (Superseded the original Cat-2/Cat-5 names “conditional obligation” / “conditional burden”, which conflated the trigger-control distinction.)

3. Manual validation. Every methodological decision validated against line-by-line ground truth across multiple Acts. The validation workbook is the authoritative ground truth; classifier outputs are assessed against it; methodological adjustments require workbook evidence.

4. Section-anchored extraction; the unit of count is the distinct burden. (Supersedes the original “sentence-level classification” commitment.) After the 2026-06-14 pipeline reframe and the 2026-07-03 extraction rebuild, candidates are surfaced at the section/provision grain (DOM-anchored; eight material_types — see the implementation plan), and the unit of measurement is the distinct burden (rubric §1 unit rule): one section can carry several burdens and one burden can span several sentences. Not word-counting, not document-level, and no longer sentence-level — the count tracks burdens, not grammar.

Capacity axis — economic vs personal (v2.6)

Each burden carries a capacity tag, applied after the private/public gate. (Supersedes the earlier activity-profile formulation.) Capacity turns on whether the duty attaches to voluntary market activity on the burdened actor’s side:

• economic — the actor is exchanging or producing-for-exchange (selling, buying, letting, hiring, employing, working for pay, producing/supplying for the market), formal or informal, either side of the transaction. Private eBay seller → economic; business buyer → economic; consumer’s duty to pay → economic.
• compulsion carve-out (side-specific) — a scheme-compelled transaction does not confer economic capacity on the compelled party; classify by the underlying activity. Dog-chip fee, MOT, compulsory motor insurance → personal for the owner/driver; the provider (vet, garage, insurer) → economic.
• personal — no market activity on the actor’s side (ownership, status, conduct, consumption).
• both/either and ambiguous — unchanged.
• Splits: a role-split within one exchange (trader/consumer) → both economic, the split lives in obligated_party; a market-vs-non-market split of an activity (driving for hire vs privately) → the split decides capacity.
• Floor: economic floor = burdens on voluntary market activity (includes consumer-side transactors); the business-burden cut = economic ∧ business-side obligated_party, reported as its own line.

What is in and out of measure

In measure — the centrally published statute book: in-force UK legislation that legislation.gov.uk exposes as machine-readable text (the Tier-1 corpus; see the coverage note for digitisation coverage within this layer).

Out of measure — distinguished, not lumped:

• Binding out-of-corpus instruments. Byelaws, traffic regulation orders (TROs), permit / licence / authorisation conditions, and regulator rulebooks (the FCA and PRA Handbooks the heavyweight cases). Legally binding, but not part of the centrally published statute book — their internal contents are not counted.
• Non-binding regulator guidance. Codes of practice, guidance notes and the like. Out of measure on a different ground (non-binding) — never conflated with the binding layer above.

The frontier principle — deepest layer the measure can see

Count each burden once, at the deepest layer the measure can see (count-at-source, generalised across the scope boundary):

• An in-measure compliance/contravention hook (duty to comply with in-corpus legislation) stays excluded — counted at its own target.
• A statutory duty to comply with an out-of-measure binding instrument IS counted, once, as a frontier proxy — it stands at the visible frontier for the invisible layer behind it (which would otherwise be captured nowhere). One proxy per out-of-measure target; never an enumeration of that target’s internal contents.
• Frontier proxies are tagged (frontier_hook, frontier_target_type) so the population is enumerable and reportable as “N frontier duties standing proxy for out-of-measure layers.”

The scope-expansion flip

Counts are not additive across scope expansions. If a future phase brings an out-of-measure layer into measure (e.g. a regulator’s rulebook — see Future phases), that layer’s contents are counted there and the corresponding frontier proxies re-classify to counted_at_source — so the same rule is never counted in two scopes at once.

Growth-analysis consequence

Regulation migrates between layers with no net change in burden. The FSMA 2023 REUL→FCA/PRA-rulebook transfer will show as a statute-book financial-services burden decline that is layer-migration, not deregulation; the growth/flow analysis names it as an adjustment when the data exists.

Output

A number — total private-actor obligations and prohibitions in UK in-force legislation — broken down by:

• Legislation type (ukpga, uksi, ssi, wsi, nisr, asp, nia, apni, eur, eudn, etc.)
• Territorial extent (England, Wales, Scotland, Northern Ireland, UK-wide; supports the devolution fragmentation analysis)
• Subject area (employment, environment, financial services, consumer protection, criminal law, etc.)
• Year of enactment / provenance (introduced_by / introduced_year; supports historical flow analysis)
• Obligation type (the six-category breakdown above) and polarity
• Capacity (economic / personal / both / ambiguous — the economic floor and business-burden cut)

Plus the frontier-proxy line (N duties standing proxy for out-of-measure layers) reported alongside, never folded into the statute-book count.

Future phases

The headline measure is Phase 1. Subsequent phases extend the framework to:

• Regulatory rulebooks — extending beyond legislation to the prescriptive content of the FCA Handbook, Ofcom rulebooks, PRA Rulebook, etc. This is the archetypal scope expansion: those layers are currently out of measure (frontier proxies stand in for them), and bringing them in triggers the scope-expansion flip above.
• Historical flow analysis — measuring the rate of obligation and prohibition addition and removal over time, by Government, by department (keyed on provenance; mind the layer-migration adjustment above).
• International comparison — applying the same methodology to other jurisdictions (US, EU, Commonwealth) to support comparative regulatory-burden analysis.

Decision-making rule

When making suggestions about methodology, scope, or technical decisions, ask:

• Does this advance the count of private-actor obligations and prohibitions?
• Does it respect the public/private filter, the six-category taxonomy, the manual-validation requirement, the unit-of-count (distinct burden) rule, and the in/out-of-measure scope boundary?
• Does it move the work closer to the headline number, or does it serve a future phase?

If a proposed change moves the project away from these anchors without clear justification, flag it as a divergence and surface the trade-off explicitly.

What the methodology paper documents

The classification system, the validation results, and the headline number with confidence intervals. The paper is not yet published; the corpus, classifier, methodology, and validation workbook are the inputs.

Cross-references

• docs/methodology.md — full methodology (v14), the six-category taxonomy and key rules
• the validation rubric (current draft; owns the operative labelling rules — §1 unit rule + capacity axis, §2/§2A categories + polarity, §3 count-at-source / frontier proxies, §4 non-operative + deeming)
• category_mapping.md — canonical numeric category IDs ↔ presentational names
• docs/coverage_methodology_note.md — corpus coverage, scope-stratified figures, and the “Regulatory layers beyond the statute book” scope disclosure
• docs/implementation_plan.md — pipeline, schema notes (frontier fields; eight material_types), and the documented eight-type baseline
• Reg Burden Project Validation.xlsx — manual validation workbook (ground truth)
• tna_dataset_comparison.md / tna_crosscheck_methodology.md — comparison with, and validation layer derived from, TNA’s Statutory Powers and Duties dataset