>>2916748I think your latest response has actually clarified the disagreement rather nicely, because you have now explicitly admitted the central point I have been making throughout:
we do not have the asylum-seeker rape rate.
Everything you are doing after that is an attempt to infer it from other populations.
That is a legitimate thing to attempt. It is not legitimate to present the inference as though it were the measured statistic.
- You are still confusing population size with exposure
You say that reducing 110,000 to 80,000 “makes my estimates worse”.
No. It changes the denominator of your hypothetical calculation. It does not establish that the resulting ratio is an offending rate.
You are taking a number of convictions and dividing it by a number of people who claimed asylum in a particular period.
But convictions are not contemporaneous with asylum applications.
Someone can arrive in 2021, be investigated in 2023 and convicted in 2025. Someone can arrive in 2025 and be convicted in 2026. The denominator therefore needs to correspond to the population at risk over the relevant exposure period, not merely the number of people who happened to claim asylum during one calendar year.
This is not pedantry. It is the difference between calculating a rate and dividing two unrelated quantities.
And your own “80,000” figure actually demonstrates the problem: you have moved from people claiming asylum to your estimate of people who were adult main applicants, while retaining a conviction numerator whose cases need not correspond to that cohort.
- Your small-boat calculation is simply wrong
You now say:
“If we also see that 40% of all asylum seekers are illegally entering the UK, and they are 72% male, then at least 60% of all male asylum seekers come from boats.”
That does not follow.
The current Commons Library figures say that 40% of people who claimed asylum in the year ending June 2026 had arrived by small boat.
Separately, 72% of people detected arriving by small boat were adult men.
Those are two different denominators.
You cannot turn:
40% of asylum claimants came by boat
and
72% of small-boat arrivals were adult men
into:
60%+ of male asylum seekers came by boat.
You would need the sex distribution of the entire asylum-seeker population and the joint distribution of sex × arrival route.
In fact, the latest Commons Library data also show that people claiming asylum after entering the UK on a visa or other leave represented 39% of asylum claims in 2025. So even the broader premise that the asylum population is essentially synonymous with small-boat arrivals is demonstrably false.
Again: the problem isn’t that your conclusion is politically inconvenient. The arithmetic does not establish it.
- Eight convictions from asylum seekers would still not establish the rate you want
You ask whether you should gather eight convictions across 2018–2026.
If you found eight, that would establish that eight asylum seekers were convicted.
It would not by itself establish the rate.
You would then need to know the relevant asylum-seeker population exposed to the risk of conviction over that period.
And if you expand your numerator from 2025 to 2018–2026, you must correspondingly expand and properly define your denominator.
Otherwise you are simply adding observations to the numerator while leaving the denominator conceptually unchanged.
That’s not a stronger estimate.
- “If they don’t have the foreign rate, it is an anomaly” is not evidence
This is the fundamental problem with your argument.
You say:
“Yes, [I assume the subgroup inherits the parent population’s rate] because if it didn’t, it would be anomalous.”
But “anomaly” relative to what?
You’ve defined the foreign-national rate as the expected asylum-seeker rate.
Then you propose to interpret deviation from that assumed rate as anomalous.
That is circular.
Suppose I said:
“I expect asylum seekers to have a lower rape rate than foreign nationals generally. Therefore, if they have the foreign-national rate, that would be anomalous.”
You would quite rightly ask me what empirical evidence justified my expectation.
Exactly the same objection applies to you.
An expectation is not an observation.
A null hypothesis is not a discovered fact.
And “common sense” does not turn an assumption into a measured rate.
- Your own ethnicity argument makes this even clearer
You say your argument is not that all foreigners have identical offending rates.
Fine.
Then we agree that:
P(offending | foreign national)
does not determine:
P(offending | particular subgroup of foreign nationals).
That is precisely my point.
You have shown that national/ethnic groups can have different conviction distributions.
So the relevant question becomes:
What is the rate for asylum seekers?
Not:
What is the rate for foreigners?
And not:
What is the rate for Pakistanis?
And not:
What is the rate for people arriving by small boat?
Those are different populations.
The fact that the parent population is heterogeneous makes subgroup-specific measurement more important, not less.
- Your Pakistan/Bangladesh comparison does not establish what you think it establishes
You give:
Bangladeshis: median age 27, 1% of UK population, 1% of rape convictions
Pakistanis: median age 28, 2.5% of UK population, 4% of rape convictions
Even granting those figures exactly as you’ve presented them, this doesn’t give you the conclusion you want.
First, you are comparing representation in the conviction population with representation in the general population. That can be suggestive of a disparity, but it is not equivalent to measuring the underlying offending rate.
Second, the CPS data are defendant self-declared ethnicity data for rape-flagged prosecutions/convictions, not nationality data and not a census of actual rapes committed. The CPS itself attaches caveats concerning errors and omissions in ethnicity data.
Third, you haven’t controlled for the other differences between these populations.
So at most this demonstrates that the crude representation of these groups in this criminal-justice dataset differs from their population representation.
It does not tell you why.
And it certainly doesn’t tell you the asylum-seeker rate.
- Your “same ethnicity, therefore foreignness predicts it” claim doesn’t follow either
You say:
“A foreigner of the same ethnicity as a UK citizen will be more criminal.”
That’s the claim you need to demonstrate.
Your ethnicity figures do not demonstrate it.
If anything, they illustrate the problem.
Ethnicity and nationality are different variables.
A British-born British citizen of Pakistani ethnicity and a Pakistani national who arrived six months ago can belong to the same ethnic category while having radically different migration histories.
The CPS table doesn’t establish that the difference between those people is caused by nationality.
And the UK government’s criminal statistics do not even give you the joint dataset you would need to establish that cleanly.
The Migration Observatory notes that government crime statistics do not contain country-of-birth information, and naturalised citizens are recorded as British in nationality statistics. It also notes that there is no published data identifying convicted offenders’ immigration status as asylum seeker, refugee, work migrant, student, etc.
That is precisely why your proposed inference remains an inference.
- “But you haven’t provided a counter-example” is not how the burden of proof works
This is another important point.
You say:
“I have given my numbers, and you just can’t provide counter-examples.”
But the absence of a counter-example does not establish your proposition.
If you assert:
Asylum seekers have a higher rape rate than citizens
the evidential burden is to establish that comparison.
You cannot say:
“You haven’t proved that they have a lower rate, therefore my higher-rate hypothesis stands.”
That is an argument from ignorance.
The appropriate conclusion from missing data is:
the rate is uncertain.
Not:
therefore use the rate from another population.
- And no, demographic adjustment does not mean “crime becomes equal”
You ask:
“You seem to be saying the same thing; if you account for demographic differences, crime becomes equal.”
No.
I am saying something much more modest.
Suppose population A has twice the crude conviction rate of population B.
If A is also substantially younger and more male than B, then part of the crude difference may be attributable to age and sex composition.
After adjustment, three things are possible:
- the difference disappears;
- the difference becomes smaller;
- the difference remains or becomes larger.
We don’t know which without doing the analysis.
That is exactly why the Migration Observatory’s statement that the non-citizen share of the prison population is lower relative to citizens when age is taken into account matters. It doesn’t prove that nationality has no effect. It demonstrates that crude nationality comparisons can change materially after accounting for age.
You are treating “age-adjustment matters” as though I had claimed “age explains everything”.
I haven’t.
- Your “all aggregates are sums of individual data points” point is irrelevant
Of course an aggregate consists of individual observations.
The question is whether you can infer the conditional distribution of one subgroup from the aggregate distribution.
You cannot.
Imagine a university where 20% of students fail an exam.
That tells you nothing by itself about whether:
* first-years fail at 30%;
* second-years fail at 10%;
* international students fail at 25%;
* domestic students fail at 18%;
* students from one faculty fail at 40%.
The aggregate is mathematically composed of the subgroups, but the subgroup rates are not recoverable merely from the aggregate.
You need the subgroup data.
That is the exact statistical mistake you’re making.
- “Syllogistic logic” doesn’t rescue an invalid premise
A syllogism is only as good as its premises.
Your implicit syllogism is roughly:
- Foreign nationals have a higher observed conviction representation.
- Asylum seekers are foreign nationals.
- Therefore asylum seekers have the foreign-national conviction rate.
But premise 3 does not follow.
Being a member of a population does not entail having the population’s mean rate.
If it did, there could be no subgroup differences at all.
But you yourself are relying on subgroup differences when you discuss Pakistanis, Bangladeshis, Afghans and other nationalities.
So you are simultaneously saying:
“subgroups can have substantially different rates”
and:
“the asylum-seeker subgroup should be assumed to have the parent group’s rate.”
Those propositions are incompatible unless you supply evidence for why asylum seekers should be an exception to the subgroup principle.
- What I actually concede about the 1,250 men
Here I think we should be completely explicit, because this is where we genuinely agree.
If Piddington receives 1,250 additional people who are predominantly young men, then, holding age/sex-specific offending rates constant, the expected number of offences can increase simply because the population has increased in a demographic category with a higher baseline rate of offending.
Yes.
That is a perfectly reasonable demographic observation.
But it is not the same proposition as:
“The asylum seekers have a higher rape propensity.”
Nor is it:
“The asylum seekers have a 100%, 200% or 300% higher rape rate.”
Nor is it:
“Piddington residents have empirical evidence that these asylum seekers constitute a particular sexual threat.”
The first proposition is an expectation based on population composition.
The latter propositions require measurements that you do not have.
- And your “minimal case” is weaker than you think
You now say that your argument is deliberately minimal:
“The act in itself warrants at least the most basic caution according to the most basic threat.”
Fine.
Then call it that.
There is a very important distinction between:
“Adding 1,250 predominantly young men changes the demographic risk profile of a small village.”
and:
“The asylum seekers themselves are demonstrably more likely to commit rape.”
I accept the first.
You have not established the second.
But notice what happens if you retreat to the first proposition.
You no longer need the foreign-national conviction figures at all.
You don’t need your Pakistan/Bangladesh comparison.
You don’t need the eight convictions.
You don’t need the boat-crossing statistics.
You don’t need the CPS ethnicity data.
The entire evidential superstructure you’ve spent this debate constructing becomes unnecessary.
You’re simply making a demographic population-change argument.
That is a much narrower proposition.
- And it does not follow that every increase in expected crime validates every fear
This is another distinction you keep collapsing.
Suppose adding a large number of young men increases the expected number of sexual offences from, say, 1.0 to 1.2.
That is a 20% increase in expected offences.
It does not follow that residents should therefore believe:
“These men are unusually likely to rape women.”
Nor does it follow that a particular individual is substantially more dangerous.
Nor does it establish that the increase will actually materialise.
Expected value is not realised outcome.
And it certainly doesn’t establish the magnitude of the increase.
You keep saying I’m hiding behind the word “unusually”.
I’m not.
The magnitude is precisely the thing you haven’t measured.
If your claim is simply “the expected number could increase”, I agree.
If your claim is “the increase is sufficiently large to empirically validate residents’ sexual-threat fears”, then you need to tell me how large it is.
That is not moving the goalposts.
That is asking what your conclusion actually means.
- Your “at least 60% of male asylum seekers are boat crossers” claim is especially important to withdraw
Because this is no longer a philosophical disagreement.
It is an arithmetic error.
The latest Commons Library figures say:
* 40% of people claiming asylum in the year ending June 2026 had arrived by small boat.
* 72% of small-boat arrivals were adult men.
You are reversing the conditioning.
You have:
P(boat | asylum) ≈ 40%
and:
P(male | boat) ≈ 72%.
You are claiming:
P(boat | male asylum seeker) ≥ 60%.
That does not follow.
Those are different conditional probabilities.
You would need the joint distribution to calculate the latter.
This is actually a textbook example of why I have been objecting to your repeated use of “deduction”.
- Finally, your position has now become unfalsifiable
This is the most serious methodological problem.
You’ve said:
* if asylum seekers have the foreign-national rate, that supports your hypothesis;
* if they have a higher rate, that supports your hypothesis;
* if they have a lower rate, that’s an “anomaly”;
* if there aren’t enough data, you use pattern recognition;
* if I ask for a subgroup denominator, you say the missing data shouldn’t prevent deduction.
What empirical result would actually make you revise the hypothesis?
If asylum seekers were shown to have a lower sexual-offence conviction rate than citizens, would you accept that as evidence against your claim?
If asylum seekers were shown to have the same rate, would you accept that?
If young male asylum seekers had the same age-specific rate as young male citizens, would that change your conclusion?
If small-boat arrivals had a lower rate, would that change your conclusion?
Those questions matter because a statistical hypothesis has to be capable of being contradicted by evidence.
Otherwise “anomaly” simply becomes a word for any observation that conflicts with the prior assumption.
So I think the actual disagreement is now very simple
There are three separate propositions:
A. Adding a large number of predominantly young men to a small population can increase the expected number of sexual offences.
Yes.
B. Some foreign-national or ethnic groups have different observed conviction distributions from others.
Yes, although the CPS data have important methodological caveats and those distributions are not themselves proof of underlying offending rates.
C. Therefore the particular asylum seekers being placed at Piddington have a higher rape rate, or that residents’ specific fear of them as a sexual threat has been empirically validated.
You have not demonstrated that.
And your latest argument doesn’t fill that gap.
It simply says:
“I don’t have the subgroup rate, so I will assume the parent-population rate, and call any deviation an anomaly.”
That isn’t a deduction from the available evidence.
It is an assumption about the missing evidence.
And once we distinguish those two things, I don’t think there is actually much left to disagree about.