← More or Less
More or Less
Are refugees more likely to commit crime?
00:28:10
Asylum seekers and crime, nuclear fish, the Quiet Revival and Sir David Attenborough.
Loading summary
Transcript129 lines
- [00:00]
A
Hello and welcome to More Or Less. We're your guide to the numbers all around us in the news and in life. And I'm Tim Harford. This week, we fight misinformation wherever we find it, even if it does require us to take on the national treasure of all national treasures. And we examine further rather extraordinary developments on the story of the quiet revival in Christianity. But first, the columnist and former politician Annunziata Rees Mogg took to the social media platform formerly known as Twitter to post a claim about the proportion of sex offences in Dorset that are committed by asylum seekers.
- [00:41]
B
Asylum seekers make up 0.8% of Dorset's population and 44% of alleged sex offences.
- [00:49]
C
So.
- [00:50]
B
Unbelievable. I had to check.
- [00:52]
A
And so did we. Lizzie McNeill has been on the case. Hello, Lizzie.
- [00:55]
B
Hi, Tim. Well, first things first. This 0.8% of the population figure is specifically counting asylum seekers in asylum hotels now in Dorset. These are mainly in the Bournemouth area.
- [01:07]
A
Got it. And they're said to be committing 44% of the alleged sex offences. So where's that figure from?
- [01:15]
B
Unclear. So, last year, Dorset police replied to a Freedom of Information request, asking them how many people living in migrant hotels in Dorset have been charged with a sexual offence. And the answer was one.
- [01:28]
A
One?
- [01:28]
B
Yeah. That was the most recent data when the post was made. However, they then sent me an update to say that eight people living in asylum hotels have been investigated for sex crimes from 1st January 2025 until 31st March 2026.
- [01:44]
A
So investigated for crimes is not the same as committing crimes, but hopefully there is some kind of connection. So how many sex crimes were reported in Bournemouth in that time?
- [01:55]
B
Eight hundred and eight.
- [01:56]
A
Yikes. So really not 44%? More like 1%.
- [02:00]
D
Yeah.
- [02:01]
A
Now, so far we've been talking about asylum seekers living in migrant hotels. So what about asylum seekers who don't live in hotels?
- [02:10]
B
Well, the data isn't that helpful. Police don't break the crime statistics down by immigration status. We could instead look at crimes committed by British nationals versus foreign nationals.
- [02:21]
A
And what does that tell us?
- [02:22]
B
Well, the most recent data from Dorset Police from 2024 and 2025 shows that 7% of all sexual offence crimes were committed by foreign nationals. If you add on people with dual British nationality, then that rises to 8%.
- [02:37]
A
Right. And these people are not necessarily asylum seekers, they're just people who aren't British. Yeah.
- [02:43]
B
So these figures include people from Yemen, Afghanistan, Iran, as well as those from countries like New Zealand, Italy, Spain and Canada.
- [02:50]
A
Do we have any idea where the 44% figure came from.
- [02:55]
B
Dorset Police Don't. I asked them if they had any record of this figure and they responded,
- [03:00]
E
these are not Dorset Police statistics and
- [03:02]
A
the information is incorrect.
- [03:04]
B
I also contacted Annunziata Rees Mogg, but she didn't respond to me. But she did later tweet an apology stating that she'd used AI to check the sources.
- [03:13]
A
Uh oh yeah.
- [03:15]
B
And even just after she posted, people were in her comments section asking Grok, which is Elon Musk's AI, whether Rees Mogg's statement was true and Grok said it was and cited police data.
- [03:27]
F
Dorset Police's FOI disclosure confirms that 51 of 116 males awaiting prosecution for sexual offenses in the Bournemouth area are asylum seekers from migrant hotels. Which calculates to about 44%.
- [03:41]
A
Right. Don't worry, Lizzy, I'll take it from here. Deep in the basement of New Broadcasting House, the more or less interrogation centre. Bring in the suspect. The date is the 20th of May and the time is just after 9am I'm Tim Harford. With me in the room is the Grock AI. All answers are spoken by GROK in its own voice. Where did you get this stat? Out of 116 males awaiting prosecution for sexual offences in the Bournemouth area, 51, that's 44% were asylum seekers from migrant hotels.
- [04:26]
F
The stat comes from a Dorset Police freedom of information FOI disclosure, specifically responses in their 2025 disclosure logs under categories like migrant sexual offenses and related migrant hotel queries.
- [04:40]
A
Funny, the FOI request does not contain the 116 males awaiting prosecution figure.
- [04:46]
F
You're correct. The specific FOI request 01 FIN 253546, dated around September 2025, focuses on how many sexual offenses have been recorded resulting in charges of males residing in migrant hotels in Dorset. It does not directly state a county wide or Bournemouth wide total of 116 males awaiting prosecution for sexual offenses.
- [05:10]
A
Give me a link to a website that shows the 116 figure.
- [05:14]
F
No single public webpage directly quotes the exact phrasing. 116 males awaiting prosecution for sexual offenses as a standalone official statistic from Dorset police.
- [05:24]
A
So the 44% figure is wrong?
- [05:26]
F
Yes. The 44% figure as originally presented is misleading and inaccurate in that framing.
- [05:32]
A
Did the error spread because GROK said it was true?
- [05:35]
F
No, the error did not spread because Grok said it was true. The misleading 44% framing originated and circulated widely before my response in that X thread.
- [05:45]
A
But people used GROK to prove it was true.
- [05:48]
F
Yes, some People did cite my earlier responses, including the one you linked, to prove the claim was true. That's a fair point. And I take responsibility for the imprecision in those initial replies.
- [06:00]
A
Take em down, boys. This whole story does raise another question. We know that the claim that Dorset asylum seekers are committing 44% of alleged sexual offences is wrong. We've got another figure about the number of foreign nationals committing sexual offences in Dorset, 7 or 8%. But without any more information, we don't really know what to make of that figure. So what can we say about the general sentiment here, the idea that asylum seekers are more likely to commit crimes than other groups? Who better to ask? But friend of the programme and director of the Migration Observatory at Oxford University, Madeleine Sumption, do we have good crime statistics for asylum seekers?
- [06:47]
G
We do not. So we don't know how many crimes are committed by asylum seekers. One reason for that, actually, is that most crimes never lead to a conviction, and so the perpetrator isn't known and therefore the status of the perpetrator. But where people are convicted, and we could in theory have good statistics in practice, the immigration status and whether someone's an asylum seeker or they're on a work visa or a student visa or anything else is not routinely recorded.
- [07:10]
A
Right. So even if they've been fined, they've been imprisoned, we don't necessarily know what their immigration status is.
- [07:18]
G
That's correct. What we do have is some imperfect but vaguely serviceable data on cautions and convictions by nationality. And there are some nationalities where people. People are particularly likely to be asylum seekers. So people coming from, for example, Afghanistan, Syria, Iran, Eritrea, in most cases they will have come either as a refugee, could have brought to the UK by the government, or through the asylum system.
- [07:44]
A
Right. So we've got some sort of sense of how many Syrians are committing crimes, how many Eritreans are committing crimes, how many Germans are committing crimes. And we just have to kind of make an educated guess as to whether these people might be asylum seekers or not.
- [07:58]
G
That's right.
- [07:58]
D
The.
- [07:59]
A
The Germans, probably not.
- [08:01]
G
Exactly.
- [08:02]
A
So you said it was serviceable data. What does it tell us?
- [08:06]
G
Well, I should say the data aren't routinely published. We have obtained data up to 2024 using Freedom of information requests.
- [08:13]
A
You have to drag it out of them, is what you're saying.
- [08:16]
G
Yes, indeed. And apparently there are plans to publish more of this officially in future. But based on what we have so far up to 2024, the overall picture is that we see quite big variations in the rates at which People appear to be convicted of crimes, nationality. And it is true that some of those nationalities where most people tend to come through the asylum system, like Afghanistan, Iran, Iraq, for example, have higher rates of convictions overall and also for sexual offences specifically. The challenge that we have is being precise about how high those rates are is really difficult because we don't have good data on how many people of each nationality are actually living in the population.
- [09:00]
A
So we don't know exact numbers, but we can estimate ballpark crime rates for different nationalities using data from the 2021 census and compare against the crime rate for the overall UK population. And that gives us a rough idea.
- [09:14]
G
If you look at the top nationalities that are the sources of refugees to the uk, it does seem that when they're in the uk, the rates of convictions for sexual offences are higher, so they're overrepresented in those figures. They are, of course, a very small percentage of the overall UK population. So looking at probably a few percent of overall sexual offences convictions being committed by those top, say, 18 or so nationalities, it's a small share overall. But they're overrepresented because there aren't that many of them in the uk.
- [09:44]
H
Right.
- [09:44]
A
Okay. So a fairly small number of people responsible for a fairly small number of offences, but nevertheless, it's not outrageous to say that some of them are overrepresented.
- [09:54]
G
That's correct.
- [09:56]
A
Another question that arises is whether the demographic profile of people seeking asylum is different from the demographic profile of British residents. I'm thinking in terms of the gender and the age.
- [10:10]
G
Yes. Yeah. There are lots of differences. So people seeking asylum are more likely to be men and men of any nationality are much, much more likely to be convicted of offences than women. Those differences are really stark. We also see that people in general migrating to the uk, including in the asylum category, are in the younger age group, people in their 20s and 30s. And that's also where we tend to see higher rates of criminal offending in general. There are other risk factors for criminal offending in general, some of which also you would expect asylum seekers to be more exposed to. So, for example, being exposed to violence as a teenager and young adult having mental health problems, being socially isolated, all of these are things that in the research, in general, we know are associated with higher rates of offending, and they are also things that many asylum seekers will have experienced.
- [11:04]
A
Our thanks to Madeleine Sumption and thanks also to Lizzie McNeil. Oh, no, the more or less error. SIREN I can't believe we're only 10 minutes into the new series and Already we've made a mistake.
- [11:18]
E
Actually, Tim, the error was in the last series.
- [11:20]
A
Ah, Richard Varden, editor of More or Less. Does your appearance in the studio mean it was your mistake?
- [11:28]
E
Oh, no, more or less. We are like the Cabinet. We take collective responsibility.
- [11:32]
A
Oh, really? Well, let's hear the grisly details in
- [11:35]
E
our Numbers of the year programme. At the end of last year, one of the numbers involved the spiralling costs of infrastructure projects in the uk. In that section we said that the Hinkley Sea nuclear power station involved hundreds of millions of pounds being spent on measures to protect fish.
- [11:49]
A
And is that wrong?
- [11:50]
E
No, that bit is correct. The problem is the next bit, when we say that the cost is something like a quarter of a million pounds per fish saved.
- [11:57]
A
So what should we have said?
- [11:59]
E
We should have said that if you look at the costs over 25 years, according to the highest estimates, it would cost something like a quarter of a million pound per endangered fish saved. But actually it's a bit more complicated than that.
- [12:11]
A
Well, that's already quite complicated. So go on then, remind us all where these numbers come from.
- [12:15]
E
The figures come from a government commissioned report on the regulation of the nuclear industry known as the Fingleton Review. In a case study in that report they say that edf, the company building the nuclear power plant, are going to spend £700 million on fish protection measures. The report also lists the number of different kinds of fish that would be saved by the protection measures every year. These numbers are very small, things like 0.028 Seatrout 6, River Lamprey 18, Alice Shad. If you add them all up, assume a 25 year lifespan for the project and divide by the total cost, you get about a quarter of a million pounds per endangered fish.
- [12:56]
A
Right, So a few things to figure out there. First of all is £700 million definitely the cost of the fish protection measures.
- [13:03]
E
That's disputed. The nuclear power station is water cooled, which involves sucking in a lot of water from the Bristol Channel and then spitting it out again, killing lots of fish in the process. To lower the number killed, they are building three things, one of which has had a lot of media attention. An acoustic fish deterrent known of course as the fish disco. That's only going to cost 50 million pounds according to the Fingleton Review. Then there's a system of pipes so some of the fish that get sucked in can escape another 150 million and then 500 million for low velocity intake heads, which are basically like the nozzle on a vacuum cleaner, sucking in the water These are specially designed so they have lower suction and more fish can escape. But there's an argument about this last component because you'd need to build intake heads anyway. And it's unclear what proportion of the costs should be classified as fish protection. There's also an argument about whether all of this is enough and it's entirely possible the costs and the timescale will change.
- [14:03]
A
What about the 25 year time scale?
- [14:05]
E
Well, that sounds like a good long term perspective, but in nuclear power station terms it isn't. The Government is guaranteeing the price it pays EDF for energy from Hinkley C for 35 years. But EDF told us the 700 million includes the operating cost over the 60 year lifespan of the power station.
- [14:23]
A
And how about the number of fish being saved? Are the figures in the Fingleton Review an exhaustive list of all the fish that will not be killed because of the protection measures?
- [14:33]
E
No, Fingleton focuses on endangered fish such as the sea trout Lamprey and Alice Shadow. If you look at all fish, you would get a much bigger number.
- [14:41]
A
Such as?
- [14:42]
E
There are lots of different estimates flying around, but it seems like Hinckley C might kill more than 100 tonnes of fish per year without any safety measures. And two of the protection measures without the fish disco might save more than half of them.
- [14:55]
A
That's quite a lot of fish.
- [14:57]
E
Yes, this equates to millions of mostly tiny fish. But to put that into context, a single medium sized commercial fishing trawler might catch hundreds of tonnes of fish each year. So one fishing trawler kills more fish than a nuclear power station. Which is why the Fingleton Report focuses on the impact on endangered fish. However, what we said in the original item was misleading and in the spirit of more or less, we hold up our hands and we admit it when we're wrong.
- [15:25]
A
Quite right too. Thank you, Richard. Oh, no. What now?
- [15:32]
C
Don't worry, tim. That's the YouGov error.
- [15:34]
A
Siren that's a relief. YouGov's error, not ours. It's Tom Coles, Cat correspondent extraordinaire, occasional producer of the series. What have YouGov done?
- [15:44]
C
Well, remember we talked back in July last year about a report from the Bible Society called the Quiet Revival?
- [15:51]
A
This was the one that said there'd been a big increase in churchgoing in the uk, particularly among young men. And that wasn't reflected in the actual churchgoing data from the Anglican or the Catholic churches. And it wasn't reflected in other polling data either.
- [16:06]
C
Yeah, that's the one. We raised an eyebrow, it's fair to say. And then polling Godfather Sir John Curtis came on the program and he did the same.
- [16:15]
A
And I seem to remember that YouGov told us they had checked everything and that their result was entirely legit.
- [16:22]
C
Well, they've now gone back and found that it wasn't legit at all. They've retracted the survey and apologised, although they didn't want to come onto the program and tell us what happened.
- [16:33]
D
Right.
- [16:33]
A
So what did happen?
- [16:34]
C
In a statement, they said that they looked again at the original survey data and found that specific demographic groups contained a number of respondents who they can now identify as fraudulent.
- [16:47]
B
They said some of the anti fraud measures available in 2024 were not administered in the optimal way.
- [16:54]
D
Hmm.
- [16:55]
A
The Quiet Revival really became quite a thing in some Christian circles. So non optimal data processing is a bit of an understatement.
- [17:05]
C
Right. But it was welcome news to Professor David Vose, an emeritus professor at the UCL Social Research Institute, who told us the quiet Revival wasn't plausible back in my first report.
- [17:17]
I
I'm glad to be right. And it's unusual that something is retracted to demonstrate the correctness of one's initial skepticism. On the other hand, I do find it exasperating that it's taken so long, given that at the outset I was saying not only that the findings were very likely to be incorrect, but why they were likely to be wrong and what was behind the surprising results. And it turns out that it was indeed a matter of bogus responses on the survey.
- [17:55]
A
Yes, it is nice for once to actually get a clear answer on something. But Tom, I'm interested. Does the failure of this poll tell us anything important about polling in general?
- [18:04]
C
Maybe, yeah. The particular problem in this survey is one we've talked about before, the way non probability polls sometimes come up with very strange answers, particularly coming from young men. Quick refresher on the terminology here. The best quality polling uses a probability sample, where the polling company works out the people they need for a representative sample and then recruits them to take part. But the Quiet Revival survey used a non probability sample. YouGov recruits a panel of volunteers who sign up to do the surveys for a small financial reward and then you invite the people you need to take part from that panel.
- [18:42]
I
We have good reason to believe there's a lot of academic research that's gone on over the last few years on this, that these panels have increasingly been infiltrated by bogus respondents. Now, these could be real people who are just not who they claim to be. They might be real people who don't really care about the subject matter of the surveys and just racing through them as quickly as possible. And at least since 2022, they may well be AI chatbots that can be trained to simulate real people.
- [19:20]
C
Now, when you recruit participants for a survey like the YouGov one, you have a particular problem with the demographic groups that don't volunteer in large numbers, one of which is young men.
- [19:32]
I
If you're organising an army of AI chatbots or you happen to be some person sitting in a basement in India somewhere trying to earn your 50p by doing a survey and doing as many as possible, what you will claim to be is, is a young man in Britain, probably in a big city. And I suspect that that's why we're getting such anomalous results, particularly in that age range and in that gender.
- [20:02]
A
And Tom, it was young men that saw a big increase in their church going in. The very quiet revival survey that YouGov have now just admitted didn't actually catch a lot of bogus respondents.
- [20:13]
C
Indeed, YouGov haven't told us exactly which responses are likely fake, but David says it's quite possible that answ from bots pretending to be people could skew the result in this group upwards. Other surveys show the chances of young men going to church are very low, so they would answer a question about church going at the extreme end of the scale. But a bot might try to avoid the extreme ends of the scale to look more middle of the road and human, meaning they might push the average upwards.
- [20:45]
A
Okay, so always be sceptical about interesting poll results on the views of young men. Got it. Now, I couldn't help but notice a fleeting reference to AI there as one of the things that polling companies have to deal with. Tom, that feels to me like something that might be a very big problem coming down the tracks.
- [21:05]
C
Yeah, it does have that feel. I've been talking to Annette Jecla about that. She's a professor of survey methodology at the University of Essex. She told me that the AI problem is becoming really quite acute with the development of something called agentic AI. This is an AI tool that is quite different to the things like ChatGPT and Grok, where you type in a question and the LLM, that's the large language model, spits out the answers.
- [21:34]
H
With agentic AI, what you can do is you can ask the AI tool to do things for you independently. So the AI tool would be able to operate a browser, would be able to click to navigate between pages to submit the tool at the end. So whilst doing that, the LLM would itself be making decisions about what to do next and how to complete tasks.
- [21:57]
C
This makes it very difficult for polling companies to tell the difference between real people and agentic AI.
- [22:04]
H
Not only can the AI take on a particular Persona and answer all the questions in a consistent way, including text questions, the agentic AI can also mimic human behavior. So it can mimic how long it would take a respondent to read and answer a question on a page. That would be an indicator for fraudulent survey if it's being answered too quickly, more quickly than would be humanly possible. But the AI can mimic that, or it can even mimic mouse movements on a web page. It can mimic responses, including typos, the mistakes that humans would make and correct. The AI can mimic all of those things. There is evidence of this happening. It seems like a real arms race at the moment between the bot developers trying to outsmart the survey researchers in developing methods of detecting this kind of fraudulent survey completion.
- [22:57]
A
Long story short, in the age of AI, if you see a survey response that you can't triangulate with other data and trends, then you need to be very wary.
- [23:06]
C
Exactly. One more thing, Tim. When they retracted their survey, YouGov said they were going to run another quiet revival poll with all the anti fraud measures in place. This time it'll be interesting to see what that one finds.
- [23:20]
A
Yep, it's the gift that keeps on giving. Thank you, Tom. And thanks to David Vose and Annette Jecler. Yes, I'm afraid another klaxon, this time with a nature theme. I need to remind you at this point that we hold power to account on more or less, without fear or favour. No one is beyond scrutiny. If our loyal listeners hear a stat that they doubt, we will follow wherever the story leads, no matter who it implicates. So it was when we got this email from Matthew Clark. I heard on the BBC Secret Garden TV programme that it takes 20,000 caterpillars to feed one blue tit chick until it fledges. Blue tits can have several chicks, so the numbers would multiply accordingly. I struggle to believe it's possible. Can you investigate, please? Investigation is what we do, Matthew. So our team went straight to the source. Who was it who made this eyebrow raising claim?
- [24:29]
C
To successfully fledge, each chicken will require 20,000 caterpillars.
- [24:37]
A
It's a tall order that must be completed within three weeks. Wait, we're fact checking? Sir David Attenborough, the national treasure who just celebrated his hundredth birthday. Are you Trying to get me fired.
- [24:54]
C
Without fear or favour, Tim. Without fear or favourite.
- [24:58]
A
Well, who am I to question Our production team, all of whose names are available in this program's credits. If I go down, I'm taking you all down with me. So is it really right that each blue tit chick requires 20,000 caterpillars over the nearly three weeks that it's in the nest? Blue tit nests tend to have about 10 chicks in. That's 200,000 caterpillars for the two parents to find in three weeks. And we spoke to Dr. Malcolm Burgess, principal conservation scientist at the RSPB and more or less his resident blue tit consultant. He looked at the scientific literature on the topic and did some quick calculations for us.
- [25:42]
D
So each chick will be receiving about 90 caterpillars per day. The blue tit nestlings are in the nest for about 16 to 20 days. So say 20 days. 20 times 90 caterpillars per day gives about 1,800 caterpillars per chick.
- [26:01]
A
So nearly 2,000 caterpillars per chick. Much more feasible. But David Attenborough said 20,000. So what could have gone wrong? Well, remember that there's more than one chick in each blue tit nest.
- [26:16]
D
Typically, a nest would have up to 10 chicks in it. Could be more, could be less, but say on average 10. So 10 times 1,800 gives you 18,000 caterpillars.
- [26:26]
A
18,000 caterpillars for the entire nest of 10 chicks. Just shy of Attenborough's 20,000 figure that he said was per chick. So that's the mistake here. They should have said 20,000 caterpillars per nest rather than per chick. After we got in touch with the makers of the Secret Garden documentary, the BBC published an official correction. We have added a clarification strap at the start of the episode and apologize for the error in the narration. And who did that narration again?
- [26:58]
C
When caterpillars are in short supply, spiders
- [27:04]
A
will have to do. So we fact checked David Attenborough. Next on our national treasure list is Paddington Bear. That's all we have time for this week, but please keep your questions and comments coming to more or lessbc.co.uk because there's a whole series to get through and we couldn't do it without your ideas. Until next time, goodbye. More or Less was presented by me, Tim Harford. The producer was Tom Coles with Nathan Gower, Josh McMinn and Lizzie McNeil. The production coordinator was Brenda Brown. The programme was mixed by James Beard and our editor is Richard Varden.
- [27:50]
B
If you've got a scrolling problem, then this is the podcast for you.
- [27:53]
A
It's called Top Comment.
- [27:54]
B
With me, Matt Shea, and me, Mariana Spring. We both investigate social media for a
- [27:58]
A
living, whether it's disinformation, conspiracy theories, Internet culture memes.
- [28:03]
B
We're gonna be getting behind the stuff that is popping up on your feed
- [28:06]
A
on this podcast that's Top Comment on BBC Sounds.