UK Migration — Crime, Trust, and the Debate
A standalone, careful treatment of the topic at the centre of trust collapse in UK migration policy. Confidence labels (high / medium / low) at every claim level. The foundational data gap, what data does exist, the Albanian signal as the strongest finding in the data, the geographic correlation, the grooming gang question, the suppression question, and what the evidence-graded conclusions actually support. Long, deliberate, and uncomfortable in places.
21. Crime, Trust, and the Migration Debate
What this section is
The crime question sits at the centre of why public trust in migration policy has collapsed. This section presents the comprehensive evidence on crime and immigration in the UK, drawing substantially on UK Crime, Immigration & the Casey Audit: An Honest Account of the Evidence (March 2026, 24 primary sources, original geographic regression analysis, ~360 verified data points, compiled through AI-assisted research). That report's findings are integrated here with appropriate attribution; the underlying data sources (MoJ PNC via CMC FOIs, Casey Audit, Bell et al 2013, Home Office returns data, MoJ prison statistics, ONS, NCA) are cited throughout.
- High confidence: official published statistics with clear definitions.
- Medium confidence: official data requiring interpretation, or quality-reviewed secondary analysis.
- Low confidence: FOI-derived, modelled, unpublished, route-inferred, or denominator-sensitive claims.
1. Nationality is not immigration status. A foreign national prisoner may be a long-settled resident, a visa holder, a person with no lawful status, a failed asylum seeker, or another category. 2. Ethnicity is not nationality. Ethnicity data cannot be used as a proxy for immigration status. 3. Prison population is not all crime. Prison data reflect serious offending, remand, sentencing, sentence length, recall policy, and early-removal rules. 4. Valid rate comparison needs a reliable denominator. A per-capita rate needs a population estimate for the group being compared, plus age, sex, region, offence-type, and residence-status controls. 5. FOI-derived conviction data are useful but lower confidence. They can indicate where further official publication is needed, but should not be treated as full official statistics unless quality-assured and methodologically explained.
The section is longer than its data weight alone would suggest because the trust dimension matters as much as the data dimension. The avoidance of crime data in mainstream policy debate has been a real failure with real consequences for community safety, victim protection, and public trust in institutions. Honest engagement is overdue.
The foundational data gap
The most important fact in this debate is not a statistic. It is an absence.
Immigration status has never been recorded by UK police at the point of arrest or charge. This is not a technical limitation — every other significant personal characteristic is recorded. It is a policy choice, made and maintained across four decades, by consecutive governments of different parties.
When asked in 2025 to provide data linking crime to immigration or asylum status, the Office for National Statistics confirmed: "We have not produced any internal assessments relating to the availability or quality of data linking migration or asylum status with recorded crime." The UK Statistics Authority had made no recommendations on this gap.
This absence means the specific question at the centre of public debate — whether illegal migrants or asylum seekers are disproportionately involved in sexual offending — is not directly answerable from any existing UK dataset. The people who claim it can be answered are using proxies and presenting them as direct evidence. The people who claim the question is inherently illegitimate misunderstand that data gaps have political consequences: when you don't collect data, the gap serves whoever can most plausibly fill it with anecdote.
The trust collapse is partly a direct product of this gap. Successive governments — Conservative and Labour — have been reluctant to publish disaggregated data on offending by nationality and immigration status, which has fed the perception that uncomfortable truths are being suppressed. Where data has emerged, it has emerged through Freedom of Information requests by think tanks (Centre for Migration Control's June 2025 release), through media investigations (the Daily Mail's April 2025 sex offence arrest data by nationality), or through Parliamentary Questions — not through proactive government publication.
Each delay reinforces the perception that the data is uncomfortable. The institutional silence has produced its own consequences.
What data does exist
Despite the foundational gap, substantial data has been extracted that allows a partial picture:
| Source | What it contains | Critical limitation |
|---|---|---|
| MoJ PNC via CMC FOI (2021-2024) | Sexual offence convictions by nationality | ~30% of all convictions; no age-sex standardisation; disputed denominators for small nationalities |
| Met Police FOI (2018-2024) | Sexual offence proceedings by nationality | London is 26% foreign-born, not 9.3% national figure |
| WYP FOIs (2023-2025) | Charges for rape/sexual assault by nationality | Charges, not convictions; single force area |
| Dorset Police FOI 25/3692 | Sexual offences by non-British suspects 2024-2025 | Single force; small denominators |
| MoJ OMSQ Table 1_A_26 (Jul 2025) | Prison population by nationality × offence group | British vs foreign national only; no finer breakdown |
| Casey Audit (Jun 2025) | Grooming gang suspect ethnicity; data gaps | Two-thirds (~66%) of suspect ethnicity not recorded nationally; explicitly states national conclusions not possible |
| Home Office NRM quarterly | Trafficking victims by nationality | Victims, not perpetrators |
| Home Office FNO returns (Ret_D03) | Deportations by nationality | Post-sentence; says nothing about offending rates |
| MoJ OMSQ Oct-Dec 2025 (Apr 2026) | Quarterly prison population by nationality | Most recent official cut (31 March 2026); updated nationality breakdown |
Most recent prison-population figures (MoJ OMSQ Oct-Dec 2025, published 30 April 2026; reference date 31 March 2026): 10,487 foreign nationals in custody in England and Wales — 3,588 on remand, 6,458 sentenced, 441 non-criminal. Foreign nationals account for 12% of the total prison population. The most common non-British nationalities were Albanian (9% of FNO prison population), Irish (7%), Polish (7%), Romanian (6%), and Indian (4%). The Albanian share has fluctuated 9-13% across recent quarterly cuts; Polish, Irish, Romanian, and Indian shares have been broadly stable.
What the conviction data actually shows
The MoJ PNC data via Centre for Migration Control FOIs shows foreign nationals accounting for 15–22% of sexual offence convictions in 2024, against approximately 9–15% of the population.
| Metric | Raw figure | Confidence |
|---|---|---|
| FN sexual offence convictions 2024 | 1,118 of 7,874 (15%) | HIGH |
| Rape of female 16+ — FN share | 155 of 720 (21.5%) | HIGH |
| Sexual assault on female — FN share | 380 of 1,453 (28.5% excl unknowns) | HIGH |
| FN sexual offence conviction rate vs British | +71% higher per 10,000 (2021-23) | LOW — denominator disputed |
| FN conviction rate (age-sex adjusted) | Non-UK nationals slightly UNDER-represented in prison overall | HIGH (Migration Observatory FOI Sep 2025) |
| % change FN sexual offence convictions 2021-2024 | +62% vs +39% British | HIGH |
These figures are widely cited without the caveats that fundamentally qualify them.
Age and sex are not controlled for. Foreign nationals in the UK are disproportionately young adult males — the demographic group that commits more crime universally. The Migration Observatory's age-sex adjusted analysis found non-UK nationals are slightly under-represented in the overall prison population. This does not mean there is no over-representation in specific categories, but it means the aggregate gap is largely — perhaps entirely — demographic rather than behavioural.
Population denominators for small nationalities are disputed by 2-3×. CMC used an Afghan UK population estimate of approximately 12,000. Alternative estimates run to 35,000 or more. A 3× error in the denominator transforms a high per-capita rate into an ordinary one.
The PNC captures ~30% of convictions. It covers indictable offences prosecuted by the CPS only — the most serious end of the spectrum. There is no established reason to assume this selection is nationality-neutral.
The Metropolitan Police dataset has a probable data error. British charges appear recorded as zero for 2018-2020 in the CMC's Met Police spreadsheet — likely because 'United Kingdom' and 'British' were not combined. This artificially inflates the apparent FN share for the period.
What we can say with confidence: foreign nationals are over-represented in serious sexual offence convictions. The direction of this finding is consistent across multiple independent sources and survives basic scrutiny.
What we cannot say with confidence: the magnitude. The per-capita rates for specific nationalities. Whether the over-representation would survive proper age-sex standardisation with correct population denominators.
What the data suggests but cannot prove: the over-representation is concentrated in specific asylum-seeker nationality cohorts — particularly Albanian, and to a lesser but real extent Afghan and Eritrean — rather than being a general immigration phenomenon. EU working migrants do not show this pattern. This distinction is important for policy.
The Albanian signal: the strongest finding in the data
One finding is robust enough to be stated without the usual caveats. It is confirmed by six independent sources, none designed to find this result:
| Source | Albanian finding | Independence |
|---|---|---|
| MoJ OMSQ Table 1_A_26 (Jul 2025) | 11-13% of foreign national prison population — largest single nationality (MoJ OMSQ Q2 2025: 10% Albanian + 1-3% additional Kosovar varies by source; PA News Jun 2025 gives 11.1%) | Official published statistics |
| Home Office Ret_D03 (Oct 2025) | 26% of FNO deportations — 2,481 individuals | Official quarterly returns |
| Lords 'Lost in Translation' (Mar 2025) | Albanian interpreter demand surged in 2023-24 | Parliamentary inquiry |
| NCA Strategic Assessment | Albanian networks dominant in UK sexual exploitation economy | Intelligence assessment |
| Home Office NRM 2024 | Albanians 15% of trafficking referrals — primarily trafficked by fellow Albanians | Victim referral data |
| Derived ratio | Materially over-represented (precise multiplier sensitive to denominator) | Illustrative calculations range ~145-195× depending on Albanian UK population estimate used; treat as low-confidence given absence of age/sex/offence-type controls |
Albanian nationals are materially represented in the foreign-national prison population and in enforced-return statistics. That supports targeted enforcement, returns cooperation, and organised-crime disruption. Precise over-representation multipliers (e.g. "150×", "160×", "200×") should be treated as low-confidence illustrative calculations rather than settled findings, because they depend on Albanian UK resident-population denominators that vary by source and lack age/sex/offence-type controls. The qualitative point — that the level cannot be explained by demographics alone — is robust; the precise scaling is not. The NCA has said explicitly that Albanian networks dominate the sexual exploitation economy.
The critical distinction: Albanian over-representation is driven by organised criminal networks — it is a professional criminal enterprise, not poverty crime or integration failure. The German research showing that work rights eliminate two-thirds of the crime effect from conflict-exposed asylum seekers does not apply here. The solutions are law enforcement, bilateral deportation agreements, and financial disruption of criminal networks. These are not the solutions being proposed by most commentators on this topic.
This finding is being underdiscussed by both sides. The restrictionist camp finds it convenient but conflates it with the general immigration picture. The opposition camp finds it inconvenient and rarely addresses it head-on. The result is that the strongest and most actionable signal in the data is being used as general fuel for an immigration debate rather than as the specific law enforcement intelligence it actually represents.
The geographic correlation: original analysis
A regression of asylum seeker concentration against police-recorded sexual offence rates across 50 local authorities, controlling for deprivation using the Index of Multiple Deprivation 2019:
| Model | Asylum coefficient | p-value | R² | Plain English |
|---|---|---|---|---|
| Bivariate (asylum only) | 0.031 | 0.0005 *** | 0.225 | Significant positive correlation |
| Partial correlation (controlling IMD) | -0.016 | 0.915 (n.s.) | — | Correlation drops to zero |
| Full model (asylum + IMD) | -0.001 | 0.915 (n.s.) | 0.492 | Asylum vanishes; deprivation dominates; R² doubles |
The finding: the bivariate correlation that drives political narrative — areas with more asylum seekers have higher sexual offence rates — is real and statistically significant. But it is entirely explained by deprivation. Once deprivation is controlled for, the asylum coefficient is statistically indistinguishable from zero.
The mechanism is straightforward: the Home Office disperses asylum seekers to the UK's most deprived local authorities. Deprived areas have higher crime rates for all offences — violence, drug, theft, sexual — driven by unemployment, overcrowding, alcohol, fragmented families, and inadequate services. These conditions pre-exist the asylum seekers. They would produce elevated crime rates with or without them.
The policy implication: if you replaced every asylum seeker in Bradford or Middlesbrough with retired schoolteachers from Hampshire, you would not expect sexual offence rates to fall to Hampshire levels. The crime is driven by the conditions of deprivation, not by who is housed in those conditions.
The grooming gang question
The grooming gang question requires more care than almost any other topic in this section, because the stakes are highest — children's safety — and the temptation to project partial data onto a national picture is strongest precisely where it would be most misleading.
Scope note: the Casey Audit is relevant to group-based child sexual exploitation and abuse, data-recording failures, institutional response, and the specific question of ethnicity and culture in that form of offending. It is not a general migration-crime evidence base and should not be used as a proxy for one. The audit concerns a specific abuse model with specific data failures and specific local evidence; generalisation to migration policy or ethnic-group claims beyond CSEA is not supported by what Casey actually established.
| Finding | Implication |
|---|---|
| Two-thirds (~66%) of grooming gang suspects nationally have no ethnicity recorded | Cannot draw national conclusions from this data |
| The 2020 Home Office claim 'most offenders are white' is not evidenced by research or data | A ministerial statistical claim was made from data where two-thirds of suspect ethnicity was not recorded, and was wrong |
| Pakistani-heritage men over-represented as suspects in specific northern towns | Local pattern confirmed; NOT a national pattern |
| 'Pakistani' removed from one suspect file during audit | Local data suppression documented in at least one case |
| National conclusions cannot be drawn from existing data | The specific question 'are Pakistani men nationally over-represented?' cannot be answered from existing data |
Both of the following are supported by the evidence and are not contradictory:
(1) Pakistani-heritage men are over-represented as suspects in group-based child sexual exploitation in specific northern towns. This is documented in Rotherham (Op Stovewood: ~2/3 of 323 suspects), and confirmed by Casey in data from Greater Manchester, South Yorkshire, and West Yorkshire. This pattern was systematically under-investigated for years. The people responsible for that under-investigation should be held accountable.
(2) Pakistani arrest rates nationally are essentially identical to the national average (11.3 vs 11.2 per 1,000). The grooming gang pattern is localised to specific communities in specific towns, not a national Pakistani or Muslim pattern. White Irish networks operated in Telford. White British networks have been prosecuted across multiple areas. The local over-representation reflects specific cultural networks in specific places.
The political debate has mostly treated these as mutually exclusive. They are not. Holding both simultaneously — taking the local pattern seriously while refusing to project it nationally — is precisely what good analysis requires and what has been most notably absent from public debate.
Because ethnicity was not recorded for two-thirds of suspects (Casey: COCAD 2023 records self-defined ethnicity for only 34% of suspects; 66% are recorded as not declared), the national picture is genuinely unknowable from existing data. Casey recommended — and the government accepted — mandatory ethnicity and nationality recording for all CSE suspects from 2025/26. If that data is collected properly and published as official statistics by Q3 2027, we will know for the first time whether the local pattern generalises nationally.
If publication is delayed, published in a form that obscures nationality breakdowns, or quietly omitted from routine statistics, the 40-year institutional pattern will have reasserted itself. That is the correct test to apply.
The suppression question
The most accurate description of the institutional failure is this: a durable coalition of institutional interests — none of which individually required a conspiracy to maintain — produced the same outcome as deliberate suppression over 40 years.
| Level | What happened | Assessment |
|---|---|---|
| Local (specific cases) | At least one case where 'Pakistani' was literally removed from a suspect file | Individual act of suppression. Documented. Real. |
| Local institutions (systemic) | Pattern of poor ethnicity recording was worst in areas where findings would be most politically uncomfortable | Institutional culture, not individual conspiracy |
| National statistics | 2020 Home Office claim 'most offenders white' made from data where two-thirds of suspect ethnicity was not recorded, without stating the limitation | Not honest statistical communication |
| Policy (data gap) | Immigration status has never been recorded at arrest or charge across 40 years | Policy choice. Different governments made the same choice. Structural |
| Central conspiracy | Coordinated national programme to hide specific findings | NO EVIDENCE. Structural explanation is sufficient and more consistent with the pattern |
It was not a cover-up in the sense of a coordinated programme with architects and instructions. It was something more insidious — an institutional culture in which the costs of finding and publishing uncomfortable data fell on the people who found it, while the costs of not finding it fell on victims who had no institutional voice. That culture produced a cover-up's results without requiring a cover-up's organisation.
This is not exculpatory. It is in some ways more serious than a conspiracy, because it has no single author to hold accountable and no single intervention that fixes it.
What causes the over-representation where it exists
This is the most important question for policy and the one most consistently avoided in the political debate. If you misidentify the cause, you prescribe the wrong treatment.
Work restrictions (German research, multiple studies 2017-2021): offering asylum seekers the right to work eliminates approximately two-thirds of the crime effect attributable to conflict exposure. Bell et al (2013): work-restricted asylum seekers in the UK increased property crime; EU migrants with full work rights reduced it. The mechanism is consistent across multiple countries and replications.
UK asylum seekers are legally barred from working for at least 12 months after their initial application, and then only in shortage occupations. The average time to initial asylum decision has exceeded 12 months for years. During this work-restriction period — which is the period captured in the conviction data — the international evidence consistently predicts elevated crime rates.
The implication: a substantial proportion of the over-representation observed in the PNC data may be reducible through a domestic policy change that has nothing to do with restricting immigration — namely, extending work rights to asylum seekers within six months of application.
The intervention is within existing Home Office powers and would cost less than the hotel accommodation programme. It would also test the international finding empirically.
This finding is absent from the restrictionist policy programme. It is also barely discussed by those on the other side of the debate. Its absence from both conversations is diagnostic.
Conflict exposure: the nationalities with the highest per-capita sexual offence conviction rates — Afghanistan, Eritrea — are countries with prolonged, intense armed conflicts. Swedish and German research consistently finds that conflict exposure is associated with elevated violence and sexual offending, and that this effect is substantially reduced (though not eliminated) by work rights and integration support.
This does not excuse offending. It identifies a mechanism that can be addressed. Young men who have witnessed or experienced severe violence, who are housed in hotels in deprived areas with no legal work, no routine, no stake in their community, and no prospect of decision, are being placed in conditions that the international criminology literature reliably predicts will produce elevated offending. This was a predictable outcome of a predictable policy. It was predicted.
The Albanian exception, again: the work restrictions and conflict exposure mechanisms do not adequately explain the Albanian signal. Albanians are not typically conflict refugees. They are not characterised by work restrictions in the way Afghan asylum seekers are. The NCA assessment is unambiguous: this is professional organised criminality operating as a business.
These are different problems. Treating them as one "immigration problem" produces policy that addresses neither.
The 18-month pipeline
One of the most striking findings is how consistently a single mechanism appears across multiple nationalities. There is an approximately 18-month lag between peaks in asylum arrivals by nationality and corresponding surges in court interpreter demand and conviction data for the same nationality.
| Arrival cohort | Arrival peak | Court/prison signal | Lag |
|---|---|---|---|
| Albanian (small boats) | 2021 | Albanian interpreter demand surge; FNO prison share rises to 13% | ~18 months |
| Afghan (post-Taliban) | 2022 | Pashto/Dari interpreter demand rising; Afghan convictions in CMC data | ~18-24 months |
| Eritrean (ongoing) | 2023-24 | Tigrinya interpreter demand rising | ~18 months (projected) |
| Iraqi Kurdish | 2022-23 | Sorani Kurdish interpreter demand surging 2023-24 | ~18 months |
The 18-month lag is mechanistically explicable: 6-18 months for asylum decision, then time for offences to be investigated, charged and reach Crown Court. The government had data on arrivals. It could have predicted this pipeline. The policy response was inadequate.
Policy implication: criminal justice pressures from a given arrival cohort are not visible in courts for over a year. Current court statistics are always 18 months behind current arrivals. Restricting arrivals today would not reduce court pressure for at least 18 months — the pipeline is already full from 2024-25 arrivals. Conversely, extending work rights now could begin reducing the crime signal in courts from mid-2026 onward.
Evidence-graded conclusions
- Albanian nationals are over-represented in the UK criminal justice system at a scale that cannot be explained by demographics. This is organised criminality, not integration failure.
- The 18-month pipeline from asylum arrival to court appearance is real and consistent across multiple nationalities. It was predictable and was not adequately acted on.
- The 2020 Home Office claim that most grooming gang offenders are white was made from data where two-thirds of suspect ethnicity was not recorded, and was not evidenced by research or data.
- The geographic correlation between asylum concentration and sexual offence rates is entirely explained by deprivation clustering.
- Work rights for asylum seekers substantially reduce — the German evidence suggests by approximately two-thirds — the crime effect attributable to conflict exposure.
- Foreign nationals are over-represented in serious sexual offence convictions relative to their population share. The direction is consistent. The magnitude — the 48-86% raw over-representation figures — is substantially inflated by lack of age-sex adjustment and disputed denominators.
- Afghan and Eritrean asylum seekers appear to have elevated sexual offence conviction rates. Direction consistent; specific magnitude unreliable; causal mechanism most consistent with conflict exposure compounded by work restrictions.
- Pakistani-heritage men are over-represented as suspects in group-based CSE in specific northern towns. This is local data, not a national pattern. Both statements are simultaneously true.
- The institutional data gap was maintained by a durable coalition of institutional interests whose collective effect was identical to suppression, even if no individual component required conspiratorial intent.
- A coordinated national conspiracy to suppress immigration-crime data.
- That the grooming gang pattern in northern towns is a national Pakistani or Muslim pattern.
- That immigration restriction is the primary policy response indicated by the evidence.
- That the specific magnitude of per-capita conviction rates for Afghan, Eritrean, or other small nationality groups is reliable as stated.
What the evidence says should happen
- Extend work rights to asylum seekers within six months of initial application. German evidence shows this eliminates two-thirds of the conflict-exposure crime effect. It is within existing Home Office powers. It costs less than continued hotel accommodation. It is the most evidence-based crime-reduction intervention available.
- Accelerate asylum decisions to under six months. The work-restriction mechanism cannot be eliminated while decision timelines run to 18+ months.
- Pursue Albanian criminal networks as a law enforcement priority. The NCA has identified them. The data confirms the scale. Specific, bounded problem requiring specific, resourced enforcement.
- Publish Table 1_A_26 (nationality × offence group in prison) as a routine quarterly statistic with commentary. It was released quietly in July 2025 as a buried spreadsheet. It should be a headline statistic.
- Implement Casey's mandatory nationality recording for all CSE suspects, with published force-level compliance data.
- Aggregate and publish hotel safeguarding referral data nationally.
- Commission independent replication of the geographic correlation analysis at full 317-LA scale.
- Require the UK Statistics Authority to assess the immigration-criminal justice data gap and publish recommendations.
- Blanket visa restrictions by nationality based on current per-capita conviction rates. The denominator problems are severe enough to make these figures unreliable as policy inputs.
- Treating the grooming gang question as resolved by the Casey Audit. Casey confirmed local over-representation in specific areas and a 40-year data gap. She did not establish a national picture.
- Using deportation as the primary policy response. Deportation is appropriate after conviction. It does not prevent re-entry through irregular routes, does not address the structural causes of over-representation where it exists, and has historically not reduced FNO prison populations significantly.
Final assessment
The honest account of what the data shows:
The problem is real but concentrated. Certain asylum-seeker nationalities — primarily Albanian, and to a lesser degree Afghan and Eritrean — are genuinely over-represented in serious sexual offending. The aggregate foreign national over-representation figure is substantially inflated by methodology, but a real signal survives scrutiny. The Albanian signal is severe and documented beyond reasonable doubt.
The geographic pattern is a dispersal policy artefact. Areas with more asylum seekers have higher sexual offence rates because they are deprived areas. Once deprivation is controlled for, asylum concentration has no independent predictive power. This does not mean asylum seekers do not commit offences — they do. It means the policy prescription of geographical dispersal into deprived areas is itself a driver of whatever general crime elevation exists.
The causal mechanism points toward integration policy, not restriction. Work restrictions, conflict exposure, and conditions of hotel accommodation in deprived areas are the most consistent causal explanations for the over-representation where it exists — except for Albanian organised crime, which requires law enforcement. The strongest evidence-based intervention is extending work rights. It is the one most notably absent from the political debate.
The institutional failure was structural, not conspiratorial. The 40-year absence of this data reflects a durable coalition of institutional interests producing the same result as deliberate suppression without requiring individual conspiracy. Casey's recommendations address the symptom — mandatory recording — but not the underlying incentive structure that made the gap possible and sustainable for so long. Whether the symptom treatment holds is the empirical test of the next two years.
The political debate has been consuming evidence in the wrong direction on both sides. The left has used the institutional failure of data collection to argue there is nothing significant to see. The right has used the real signal in the data to argue for policy responses — restriction, deportation — that the evidence does not support as primary interventions. Both positions have served institutional or political interests rather than victims.
The people who have been most harmed by this debate are the women and children who were abused while institutions argued about whether to count the people abusing them; and the asylum seekers — the vast majority — who did not offend and who have been collectively stigmatised by a debate that cannot distinguish a criminal network from a refugee cohort.