Supported asylum stays in the language of stock
Local supported-asylum figures are end-of-period support counts. They do not tell you how many distinct people passed through an area or how many claims were processed there.
Tell users what is known, what is estimated, and what is only publicly evidenced.
This is the rules page behind the place-led product. It has to be stricter than a normal dashboard because it mixes official releases with public hotel evidence and archive recovery.
What matters first
4 evidence layers, 6 asylum interpretation rules, and 4 inclusion tests stop this site from turning official data, public hotel evidence, and archive recovery into one fake certainty.
Route scope decides what is publishable
General council finance, procurement, or supplier data kept as research context and never merged into public asylum or refugee charts unless a row is explicitly tied to a route or scheme.
Archive and public-evidence layers do not bypass corroboration
Archive-derived leads and public hotel evidence are useful, but they stay visibly distinct from official datasets and documented entity links so the product never fakes certainty.
Evidence layers
Different source types can coexist on the same site only if the reader can tell exactly what kind of claim each one supports.
Official
Official
Published Home Office, ONS, NAO, or parliamentary sources with stable release or document provenance.
Derived estimate
Derived estimate
Rates, tariff estimates, or calculated shares built from published source inputs and documented methods.
Archive-derived
Archive-derived
Recovered historical regional material that must always keep archive URL, capture date, and validation notes.
Public evidence
Public evidence
Council statements, FOIs, planning records, or parliamentary material that help identify hotel sites or hotel presence.
Rules this product assumes
These are the editorial and structural rules that stop the site drifting into vague outrage or false precision.
Never imply the hotel list is complete
Named hotels are a public-evidence ledger, not an official estate register.
Separate spend families
National spend, tariff-based estimates, and local manual evidence should never be merged into one clean-looking line.
Differentiate route, scheme, and status
Small boat arrivals, asylum support, refugee resettlement, refugee family reunion, and Ukraine routes should not be collapsed into one bucket.
Treat supported asylum as stock, not throughput
Local supported-asylum figures are end-of-period support counts. They do not tell you how many different people moved through the system, how many claims were processed locally, or whether the support total matches the backlog exactly.
Public spending must pass a route test
Whole-council budgets, procurement feeds, and transaction corpora are context only unless a row explicitly ties back to a route or scheme.
Keep release streams separate
National asylum chapters and local authority tables update on different schedules and should not be flattened into one update date.
Integrity claims need document trails
Owner, operator, and supplier risk signals belong in the product, but only after the site-to-entity link and the documentary chain are strong enough to publish.
Separate owner, operator, and prime provider
A hotel owner, a hotel operator, and the regional asylum accommodation prime can all be different entities. The site should keep those roles distinct instead of flattening them into one supplier label.
How supported asylum numbers work
Quarter-end stock, not throughput
Supported asylum counts show how many people were receiving support at the end of the period. They are not new claims, arrivals, or the number of distinct people seen across the period.
Support is not the same thing as the backlog
People awaiting an initial decision and people on asylum support overlap, but they are not identical groups. Some supported people are further into appeals or on other support routes, while some people awaiting an initial decision are not in the published support count.
Processing and exits move the stock
The stock changes when people enter support and when they leave it through case progression, including grants, refusals, withdrawals, departures, or moves out of the published support categories.
Flat local lines can still hide churn
A place hovering around the same level can mean low movement, or heavy turnover with inflows offset by exits. The published local tables do not show how many different people passed through support in that area.
Initial outcomes are not latest outcomes
Initial decisions show operational output at the time of decision. Latest outcomes by claim year can change later through appeals, reconsiderations, or subsequent case progression, so the two measures should not be treated as interchangeable.
People, claims, and claim cohorts are different units
Home Office asylum tables mix quarterly claim flows, quarter-end stock counts, and claim-year cohorts. The site keeps those units separate and does not add them together as if they were one common total.
Why backlog language is not enough
A local support count can stay flat because the backlog is static, but it can also stay flat because new entrants are being offset by grants, refusals, withdrawals, departures, transfers, or other exits.
The published local authority asylum tables do not identify how many different people passed through support in a place over a year. That is why the site treats flat local lines as ambiguous unless a better flow measure is published.
Route scope classes
Route or scheme specific
Official datasets that directly identify an asylum route, refugee resettlement scheme, refugee family route, or humanitarian route.
Local route-relevant
Council statements, FOIs, planning records, or local documents that explicitly refer to asylum hotels, refugee resettlement, Afghan arrivals, Ukraine arrivals, or related local response.
Context only
General council finance, procurement, or supplier data kept as research context and never merged into public asylum or refugee charts unless a row is explicitly tied to a route or scheme.
Public inclusion test
Include publicly if the source explicitly identifies an asylum route, refugee scheme, family route, or humanitarian route.
Keep small boat arrivals separate from refugee resettlement schemes and from Ukraine humanitarian routes.
Treat local hotel statements, resettlement housing statements, FOIs, and planning records as local route-relevant evidence, not as national totals.
Keep general council finance and procurement as background context unless a row can be attributed to a specific route or scheme.
Route families the site should split
The route model is explicit because the point of the product is to stop incompatible flows being presented as one moralised bucket.
Irregular asylum route Not a refugee scheme
Small boat arrivals
Use small boat or illegal entry route language for the route. Public local asylum-support data does not usually say which supported people arrived this way.
Asylum support system Not a refugee scheme
Asylum seekers receiving support
Best local asylum pressure series, but not a route split. It includes people on support regardless of how they arrived.
Refugee resettlement and relocation Scheme family
Afghan Resettlement Programme
Public local data usually combines Afghan pathways at programme level rather than separating ARAP and ACRS in every table.
Refugee resettlement Scheme family
UK Resettlement Scheme, Mandate, and Community Sponsorship
Public resettlement outputs often group UKRS and Mandate together, with community sponsorship included within that total.
Protection-linked family route Not a refugee resettlement scheme
Refugee Family Reunion
This is a family route connected to people who already have refugee status or humanitarian protection in the UK.
Humanitarian route Not a refugee resettlement scheme
Homes for Ukraine
Useful local route comparison layer but should never be merged into asylum accommodation totals.
Hotel entity-resolution rule
Use this only when the ledger can tie a site to both owner-side and operator-side entities with a strong public source trail.
Use this when one side of the chain is known, such as an owner group or operator brand, but the full site-level entity picture is still incomplete.
Use this when a hotel is publicly named but the starter ledger still lacks a publishable owner or operator match. That is a transparency finding, not a reason to hide the row.
Money-ledger display rule
These rows show who controls a region or visible set of current sites. They are not automatically a disclosed contract value.
These rows can be tariff rates or guidance components. They should not be shown as actual spend totals unless multiplied through with transparent placement counts and method notes.
These rows are valuable accountability facts, but they belong in a distinct display state from contracts and tariffs.
Demographic projections (v7.0)
Hamilton-Perry v7.0 single-year-of-age model with Monte Carlo uncertainty, spatial smoothing, and 13 socioeconomic dimensions. 278 local authorities, 20 ethnic groups, Census-direct base, validated against Census 2021 (MAE 1.72pp).
Hamilton-Perry with single-year-of-age Cohort Change Ratios (CCRs) from Census 2011 → 2021. 20 ethnic groups. Census 2021 base from ONS custom dataset (direct observations, no IPF). Census 2011 base from DC2101EW (18 ethnic groups × 21 age bands × sex × 348 LAs, interpolated to single-year; Roma split from Gypsy/Traveller using 2021 proportions). Child-Woman Ratios for birth cohorts. Every ratio traces to Census observations. No hardcoded demographic rates.
1,000 Monte Carlo simulations (Yu, Sevcikova, Raftery & Curran 2023). Each simulation perturbs CCRs from Normal(observed, sigma) where sigma = 0.02 (calibrated from v7.0 Hamilton-Perry backcast MAE 1.71pp across 269 areas). Sigma scales with projection horizon: sigma_t = sigma_base x sqrt(years/10), so 2061 intervals are approximately 2x wider than 2031 intervals. James-Stein shrinkage (k=50) pulls small-population CCRs toward the national average. Reports median + 80%/95% prediction intervals per area. Spatial smoothing: projections blended 70/30 with regional averages.
Ethnicity (20 groups projected, expanded from 12 in v5.0). Religion (8 categories projected). Nativity (UK-born/foreign-born projected). Plus observed Census 2021 cross-tabs: economic activity, housing tenure, qualifications, health (all by ethnic group per LA). English proficiency. Diversity indices. Shift-share decomposition. Migration maturity profiling.
Hamilton-Perry v7.0 backcast validated against Census 2021 across 269 areas using the same 20-group CCR/CWR methodology as the forward projections. Our MAE: 1.71pp (RMSE 1.87pp), a 30% improvement over v6.0 (2.45pp) due to Census 2011 DC2101EW replacing proportional splitting from NEWETHPOP 12 groups. Compared against: NEWETHPOP cohort-component model (MAE 2.58pp) and national-average CCRs only (MAE 2.32pp). Our local-CCR model outperforms NEWETHPOP by 33%. Local CCRs better in 50% of individual areas. Most demographic change is driven by national-level trends. Sigma calibrated at 0.02. SNPP 2022-based envelope constraint. DfE School Census 2024/25 calibration applied (20% damped, ages 0-5). NOTE: the backcast is partially circular because CCRs were derived from the same 2011→2021 endpoints. The national-CCR baseline is a better measure of genuine predictive ability.
Independent validation against DfE School Census 2024/25 (8.1 million pupils across 126 LAs). Compares Census 2021 ethnic composition of ages 4-15 (from direct Census observations) against school enrollment 3 years later. Age-specific MAE: 2.36pp (RMSE 3.49pp). By group: White British 6.20pp, Asian 1.93pp, Black 1.78pp, Mixed 0.88pp. The 3-year gap means some divergence is expected from births, migration, and demographic change since the Census. 49 LAs have minority-WBI school populations. Pipeline analysis (primary vs secondary) identifies areas where diversification is accelerating. DfE calibration feeds back into the model: 20% damped CCR adjustment for ages 0-5 in 121 areas with significant gaps.
Hamilton-Perry assumes 2011-2021 cohort change patterns persist. Cannot model future policy changes, migration shocks, or fertility convergence. Census 2011 DC2101EW provides 21 age bands (not single-year), interpolated via uniform distribution within each band. Roma was not a separate Census 2011 category; split from Gypsy/Traveller using 2021 proportions. Census 2021 migration indicator captures COVID-period mobility (March 2020-2021). SNPP constraint applies to England only (Wales LAs unconstrained). SNPP data ends 2047; projections for 2048-2061 use linear extrapolation of the 2043-2047 SNPP trend. 2061 projections are illustrative only and carry very wide confidence intervals. Religion projections use CCRs with James-Stein shrinkage but no age-sex decomposition. ONS empirical TFRs (WBI 1.31, PAK 2.52) validate but do not replace Census CWR-based fertility in the HP model. These are demographic projections, not forecasts.
ONS Linked Births 2024 (567,460 births) provides independent validation of ethnic fertility differentials. Empirical TFRs: White British 1.31, Pakistani 2.52, Black African 2.42, Bangladeshi 2.27, Indian 2.17, Black Caribbean 1.13, Other White 1.58. National ASFR peaks at ages 30-34 (98.0 per 1000 women). The HP model uses Census CWRs (not these TFRs). The ONS data serves as calibration and validation.
ONS Births Table 8 by IMD decile shows a 1.66x deprivation-fertility gradient: the most deprived quintile accounts for 25.5% of births vs 15.4% for the least deprived. Pakistani and Bangladeshi populations are concentrated in deprived areas, so apparent ethnic fertility differentials partly reflect deprivation, not culture. Full ethnicity × IMD cross-tabulation is not available from ONS published data (marginals only).
Census 2021 migration indicator × ethnicity data (308 LAs, 20 groups) reveals ethnic-specific mobility patterns. Most mobile: Roma 26.4%, Chinese 21.4% (high international component from students), Other White 18.8%. Least mobile: White British 9.0%, Pakistani 8.6%, Black Caribbean 8.9%. Per-LA mobility rates are used to scale the CC v2 model's internal migration component. COVID caveat: Census captured March 2020-2021 migration.
Home Office Asy_D04 outcome analysis (Dec 2025) reveals initial decision grant rates significantly understate true protection rates. 70 nationalities see rates rise after appeals, with an average uplift of +14.1pp. Largest uplifts: Sri Lanka 27%→59% (+32pp), DR Congo 33%→60% (+27pp), Iraq 31%→56% (+26pp), Iran 67%→85% (+19pp). The site displays both initial and appeal-adjusted rates.
Hamilton & Perry (1962) original method. Yu, Sevcikova, Raftery & Curran (2023, Demography) probabilistic extension. Hauer (2019, Scientific Data) US county race projections. Wilson et al. (2022) ensemble methods. NEWETHPOP: Wohland, Rees, Norman, Lomax & Clark (2024), CC BY 4.0. ONS 2022-based SNPP. Census 2011 DC2101EW (NOMIS NM_651_1). ONS Linked Births 2024. Home Office Immigration Statistics Asy_D02/D04 Dec 2025. DfE School Census 2024/25. Goodwin & Sherwood (CHSS, 2025).
Cross-domain analysis
Place pages combine asylum, demographic, crime, SEND, and social care data. This section explains how domains are combined and what the pressure index represents.
Crime: ONS police recorded crime statistics via Police Force Area tables. SEND: DfE SEN2 return (Education, Health and Care Plans by local authority). Adult social care: NHS Digital SALT collection and ASCOF framework. Each domain uses the most recent annual publication at time of last update.
The composite pressure index converts each domain metric to a percentile rank (0–100) among tracked areas, then takes the unweighted mean of available domains. It measures co-occurrence of pressures across asylum dispersal, demographic change, crime rates, SEND demand, and social care spend. A high score indicates multiple domains showing elevated readings simultaneously. It does not imply that any domain causes another.
This site presents correlations between local authority characteristics. Correlation does not establish causation. Areas with high asylum dispersal rates also tend to have higher deprivation, lower property values, and different age profiles, all of which independently affect crime rates, SEND demand, and social care costs. The data is presented to inform public accountability, not to attribute causes.