Epistemic Enclosure · Part 1 of 11

How the English Planning System Distributes Its Evidence

The system is not neutral. It distributes its protections, its data and its evidence unevenly, and the unevenness tracks wealth for the kinds of protection that carry the most legal weight.

Carlos Arleo · Aitiopoietic Systems · September 2026

When residents contest a planning application, they do so with whatever evidence they can assemble. This audit asks a prior question: how much evidence does the state make available, and to whom?

We took 852 grassroots planning campaigns and passed them through a seven-stage attrition funnel, leaving a locked cohort of 641 contested sites. Each site was then cross-referenced against 65 live spatial layers from 33 APIs held by 8 government bodies. Every downstream number reports against the same 641.

EPISTEMIC ENCLOSUREFIG. 02 · EE‑2026

No metric may cite an unverified denominator

Attrition from gross harvest to the locked canonical cohort

75.2%

of the 852 harvested campaigns survive every stage with an invariant denominator; every downstream chapter reports against the same 641.

Manuscript constants (Slide 2). Sample-bias direction stated as conservative.

EPISTEMIC ENCLOSUREFIG. 01 · EE‑2026

Contested sites across England

Each dot is one site in the canonical cohort, coloured by deprivation band

641

contested sites, plotted at their resolved coordinates (no basemap: the shape of England is the sites themselves).

Equirectangular projection scaled by cos(mean latitude). Computed from canonical CSV.

1Protection is not one thing

Statutory shields are not spread evenly. Ecological protections, the ones that carry the most weight in law, rise with affluence (Mann-Whitney U = 8,611, p = 0.0029). But a second axis runs the other way: sites in the most deprived areas have 7.75 times the odds of sitting beside a Heritage at Risk asset.

EPISTEMIC ENCLOSUREFIG. 03 · EE‑2026

Statutory shields rise with affluence; burdens fall

Mean layers per site, by deprivation band

0.57 → 0.87

Protective shields per site, most deprived → most affluent band. Heritage at Risk (separate axis): 18.9% of deprived-band sites vs 2.9% of affluent-band sites.

Computed from canonical CSV. Sites with any ERROR/blank contributing layer are excluded from that mean. Level 2 (association).

EPISTEMIC ENCLOSUREFIG. 05 · EE‑2026

The shields of decay skew deprived

Sites within 500m of a Heritage at Risk asset (The 5th Shield)

OR 7.75

Deprived sites have 7.75× higher odds of intersecting Heritage at Risk (p < 0.001). The near-mirror image of Ancient Woodland.

Manuscript constants (§7.3). Level 2 (statistical association).

Add heritage to the ecological shields and the composite effect attenuates from p = 0.0029 to p = 0.191. The system does not simply protect the rich. It protects different kinds of value in different places, and the kind that wins appeals is the kind that tracks affluence.

2Speculative capital goes where the soil is best

Development categories differ in their land. Grid-scale energy sites intersect Best and Most Versatile farmland far more often than urban infill does (χ² = 12.40, p = 0.0063).

EPISTEMIC ENCLOSUREFIG. 08 · EE‑2026

Speculative capital targets high-grade soil

Intersection with Provisional Best & Most Versatile (BMV) farmland

71.4%

of grid-scale energy capital sits on BMV farmland, compared to 47.0% for urban infill (χ² = 12.40, p = 0.0063).

Manuscript constants (§8.1). Level 2 (statistical association).

3The data void sits in the commuter belt

Across the cohort, 36.3% of sites fall in ecological data deserts, with fewer than 5,000 historical biological records nearby. There is no detectable difference between deprived and affluent sites (p = 0.8693). The rate is instead non-monotonic, peaking at 48.4% in IMD decile 6: neither urban enough to attract civic data-mapping nor designated enough to attract formal surveys.

EPISTEMIC ENCLOSUREFIG. 07 · EE‑2026

Where ecological data goes missing

Share of sites with < 5,000 NBN records within 1 km, by IMD decile

Decile 6

peaks at 46.7% (n = 92). Overall 218/626 sites (34.8%) sit in a data desert.

Computed from canonical CSV; every decile plotted from data, none interpolated.

EPISTEMIC ENCLOSUREFIG. 09 · EE‑2026

Deserts track infrastructure, not housing

The flipped archetype: Ecological Data Deserts (<5k records) by category

31.5%

Residential sprawl is under-represented in data deserts (OR = 0.65, p = 0.0109). Deserts concentrate in energy and logistics.

Manuscript constants (§9.3). Level 2 (statistical association).

Deserts track infrastructure, not housing.

4Open in law, closed in operation

The fourth finding concerns the state's own plumbing. Three datasets that planning evidence depends on are inaccessible by three different mechanisms.

Three barriers, three mechanisms
Dataset	What happens	Stated rationale
HSE MAHP zones	No public REST endpoint	National security / critical national infrastructure
HM Land Registry INSPIRE	WFS disabled (`Service WFS is disabled`)	Commercial IP conflict
Coal / mining risk	View-only WMS plus licence gate; 100% vector ERROR	Commercial revenue protection (CON29M model)

EPISTEMIC ENCLOSUREFIG. 11 · EE‑2026

The new enclosure is legally opaque

HM Land Registry title resolution across the canonical cohort

38.2%

of contested sites cannot be matched to a registered freehold title. A direct measurement of Christophers' registration gap.

Manuscript constants (§8.4). Level 1 (direct observation).

The consequence is measurable. 245 of 641 sites cannot be matched to a registered freehold title. Christophers documented the sale of public land; this audit shows the maps of who now holds it have been switched off.

5Where the burdens stack

EPISTEMIC ENCLOSUREFIG. 12 · EE‑2026

From 639 sites to 52 facing both burdens

Bottom 30% nationally for economic deprivation and environmental quality

8.1%

52 of 639 canonical sites are in the bottom 30% on both IMD and IMED (Triple Jeopardy).

Computed from master_cohort_canonical.csv; sites lacking either decile excluded from the denominator.

EPISTEMIC ENCLOSUREFIG. 13 · EE‑2026

The urban compounding core

Proportion of sites located in Labour-held constituencies

OR 2.42

Labour-held constituencies concentrate Triple Jeopardy sites (+21.6 pp disparity, Fisher's exact p = 0.0034).

Manuscript constants (§11.2). Level 2 (statistical association).

52 sites are in the bottom 30% nationally on both economic deprivation and environmental quality. These Triple Jeopardy sites concentrate in Labour-held constituencies (OR 2.42, Fisher's exact p = 0.0034).

EPISTEMIC ENCLOSUREFIG. 04 · EE‑2026

Every site, not just the averages

Log‑Defensibility Ratio L = ln(protective+1) − ln(hazard+1) per site; bar = band mean

0.13 vs 0.31

mean L, most deprived vs most affluent band. The median is 0.00 in every band — over half of sites carry equal shields and hazards — so the mean is the honest summary. Above zero means more shields than hazards.

Computed from canonical CSV; sites with any ERROR/blank contributing layer excluded. Level 2 (association).

6What we did not find

We also tested whether the consultation-stage asymmetry predicts the permit outcome. At full coverage it does not: desert sites are no more likely to be permitted (OR = 1.19, 95% CI 0.66 to 2.13, p = 0.73). That is a result, not a failure. The asymmetry lives in the evidence available to residents, and is not, on this test, visible in the binary decision.

All associations here are Level 1 or Level 2 on the evidence ladder: direct observation or statistical association, not causation.

The four findings
Finding	Headline	Test
Defensibility Gap	Ecological shields skew affluent; heritage skews deprived	U = 8,611, p = 0.0029; OR 7.75
BMV farmland	Energy 71.4%, logistics 64.0%, infill 47.0%	χ² = 12.40, p = 0.0063
Data scarcity	36.3% of sites; peak 48.4% at IMD 6	p = 0.8693 (class)
State API fragility	Three barriers, three mechanisms	Direct observation

Next in the series: The Two-Axis Defensibility Gap. Data: locked canonical cohort v5.0.2, N = 641.