Are refugees more likely to commit crime?

Episode Summary

Executive Summary: This episode fact-checks viral claims about asylum seekers and crime in Dorset, corrects an earlier misleading nuclear power fish-cost claim, explains why YouGov retracted its “quiet revival” church-going poll, and debunks a BBC blue-tit feeding statistic. Across all stories, the theme is the same: numbers can be technically sourced yet misleading without careful context, definitions, and triangulation.

Main Topics: Dorset asylum seeker crime claim (Priority: 5/5): The programme investigates a social-media claim that asylum seekers in Dorset made up 44% of alleged sex offences. Reporters find the figure is unsupported and based on a misleading interpretation of police data; the real proportion related to a narrower hotel-based prosecution dataset, not overall sexual offences. Limits of crime data by immigration status (Priority: 5/5): An Oxford Migration Observatory expert explains that police and courts do not routinely record immigration status, so there is no clean public dataset on asylum-seeker offending. The best available proxy is nationality-based convictions, which remains imperfect and only suggestive. Correction of the Hinkley fish-cost claim (Priority: 4/5): The programme revisits an earlier error about Hinkley Point C fish protection costs, clarifying that the quoted quarter-million pounds per fish was based on a selective and oversimplified reading of a report and that the total costs and fish-saved estimates are more complicated. YouGov retracts the ‘quiet revival’ poll (Priority: 5/5): YouGov withdraws a survey claiming a rise in churchgoing, especially among young men, after finding fraudulent respondents and admitting anti-fraud controls were not used optimally. The discussion highlights how non-probability panels can be distorted by bots or low-quality respondents. AI and the future of survey fraud (Priority: 4/5): Experts warn that agentic AI can imitate real participants, including timing, mouse movements, and typing errors, making survey fraud harder to detect. The episode frames this as an arms race between pollsters and bot developers. Blue-tit caterpillar statistic corrected (Priority: 3/5): The BBC corrects a Sir David Attenborough narration claiming each blue-tit chick needs 20,000 caterpillars. Scientific checking shows the likely intended figure was about 20,000 caterpillars per nest, not per chick.

Key Arguments: The Dorset claim is misleading because it compares a hotel-based subset of asylum seekers against all alleged sex offences in the county, producing an exaggerated percentage. Police and court systems generally do not record asylum status, so claims about asylum seekers’ crime rates cannot be verified directly from routine official statistics. Nationality-based conviction data is only a rough proxy, but it suggests some nationalities associated with asylum pathways are overrepresented in conviction data, while still representing a very small share of total UK crime. The Hinkley fish-cost figure was misleading because it blended disputed capital costs, operating costs, and a selective subset of endangered fish, rather than a clear cost-per-fish calculation. YouGov’s “quiet revival” result was undermined by fraudulent responses; this reinforces the need to distrust surprising poll results that are not supported by other evidence. Agentic AI will likely worsen survey fraud by imitating human behavior more convincingly than earlier bots or careless respondents. The blue-tit claim was a unit error: the total number of caterpillars per nest was misreported as being per chick.

Data Points: Asylum seekers in Dorset population share: 0.8% - Figure cited in the social media claim, referring specifically to asylum seekers in hotels in Dorset. Claimed share of alleged sex offences: 44% - The viral claim about alleged sex offences in Dorset attributed to asylum seekers. Dorset police FOI: charged people living in migrant hotels: 1 - Most recent figure available when the post was made, for people living in asylum hotels charged with a sexual offence. Investigations of people in asylum hotels for sex crimes: 8 - Dorset Police update covering 1 January 2025 to 31 March 2026. Reported sex crimes in Bournemouth: 808 - Sex crimes reported in the same period used for comparison against the 8 investigations. Foreign nationals in sexual offence crimes: 7% - Most recent Dorset Police data from 2024-2025 on all sexual offence crimes. Foreign nationals plus dual British nationals: 8% - Expanded Dorset figure including dual nationals. Males awaiting prosecution cited by Grok: 116 - The number Grok initially referenced in defending the misleading 44% framing. Asylum seekers from migrant hotels among males awaiting prosecution: 51 - The figure Grok cited as being 44% of 116 males awaiting prosecution for sexual offences. Fish protection cost cited in earlier item: £700 million - Government-commissioned Fingleton Review case study on Hinkley Point C fish protection measures. Fish disco cost: £50 million - Part of the fish-protection system described in the Hinkley case study. Pipe system cost: £150 million - Another element of the fish-protection system at Hinkley. Low-velocity intake heads cost: £500 million - Largest disputed component of the Hinkley fish-protection costs. Blue-tit chicks per nest: About 10 - Used to show that the caterpillar figure was likely intended for the whole nest, not each chick. Caterpillars per chick per day: About 90 - Derived from RSPB calculations based on scientific literature. Nestling period: 16-20 days - Duration blue-tit chicks remain in the nest. Caterpillars per chick: About 1,800 - Calculated estimate if 90 caterpillars are fed per day over 20 days. Caterpillars per nest: About 18,000 - 10 chicks times about 1,800 each, close to the claimed 20,000 figure once the unit is corrected.

Pivotal Quotes: "The stat comes from a Dorset Police Freedom of Information FOI disclosure... It does not directly state a county-wide or Bournemouth-wide total of 116 males awaiting prosecution for sexual offenses." — Grok AI: The AI initially presents but then partially retracts the evidential basis for the viral Dorset claim. "The 44% figure, as originally presented, is misleading and inaccurate in that framing." — Grok AI: Grok concedes the social-media interpretation of the Dorset data was wrong. "We do not know how many crimes are committed by asylum seekers." — Madeleine Sumption: Explaining the core data limitation in measuring offending by immigration status.

Implications: Listeners should treat striking statistics as provisional until definitions, denominators, and sources are checked. The episode shows how AI, polls, and selective data can mislead public debate, and why triangulation matters more than viral certainty.