Grooming Gangs, Drunk Drivers and Abusive Priests

Simpson’s Paradox and how statistics can be used to avoid difficult truths

There is a rhetorical manoeuvre that has become remarkably common in public life. It lets people appear morally virtuous while avoiding the substance of an argument entirely. I named it in passing in How People Cheat in Arguments, where it sat among several such tactics.

But it has outgrown that list. It does more damage than most others, and it does so more insidiously, which is why it deserves an article of its own here.

It is called Simpson’s Paradox and occurs when the national average tells one story, but the local data tell another, and the average ends up hiding where the real problem lies.

It manifests like this. Someone identifies a pattern of harm. Not a universal claim of harm. Not collective guilt about harm. Not something prejudiced. Merely an observable data-backed concentration of harm.

The knee-jerk response is now immediate:

“But on average in the UK, most of this [identified] harm is committed by white men within the family.”

This sounds reasoned, and it is presented as morally mature and superior.

But often it is none of these.

It is not an answer to the noted pattern of harm, and it is actually a change of subject and a dangerous rhetorical sleight of hand.

Simpson’s paradox warns:

Never assume the whole tells the same story as the parts.

Drunk Drivers

Photo by Anthony Maw on Unsplash

Imagine someone says:

Drunk drivers are disproportionately involved in fatal crashes.

Nobody responds:

Yes, but statistically, most crashes on average involve sober drivers.

Nobody says it because everyone immediately sees the absurdity. The statement may be statistically true and yet entirely irrelevant. The existence of more crashes elsewhere tells us nothing about whether drunk driving is a concentrated risk that warrants attention.

We understand this instinctively. Which raises a more awkward question.

Why, when the subject changes to a protected ideological predilection, do so many otherwise intelligent people suddenly become unable to think?

Part of the answer is that statistics possess an almost magical authority in modern arguments. Introduce a number, and people assume the matter is settled. But very often nothing has been settled at all. The subject has merely been changed.

So often a concern is raised about a known concentration of harm — in a place, an organisation, a context — and immediately somebody produces a national average to beat it over the head with.

Notice the sleight of hand.

The discussion moves: from this place to the whole country; from a cluster to a total population; from known actors to everybody. And because arithmetic has appeared in the sleight of hand, everyone nods gravely and pretends this counts as thought.

Usually it does not. At all.

And usually, it is simply a different question dressed up as an answer to cover the virtue signalling at work. And yes, not everyone making this move is attempting to display virtue. Many are simply making the sort of statistical mistake that modern people are oddly prone to making.

We have acquired a superstitious respect for numbers. Once a statistic appears, especially a national one, people assume the matter has been settled. But statistics are not arguments. They are answers to particular questions. And the truth of an answer depends partly on whether the right question was asked.

Which means that a statistic may be impeccably correct and yet have almost no bearing on the matter under discussion.

That is not deception. It is confusion.

But there is another reason this move is so effective. The moment identity enters the conversation, reason is crowded out by anxiety. People cease asking:

Is this true?

and begin asking:

What sort of person will I look like if I agree/disagree?

That is where broad statistics become irresistible. Not because they explain reality. But because they provide moral cover. They allow people to display decency while avoiding discomfort.

And so the same person who would laugh at anyone excusing drunk driving by pointing to sober drivers will pull exactly the same manoeuvre the moment the pattern touches ethnicity, religion, or some other protected category important to them.

Pakistani Grooming Gangs

Take one of the most painful examples in modern Britain.

Over many years, organised child sexual exploitation cases emerged across several English towns. Subsequent inquiries did not merely uncover criminal acts; they exposed institutional failure. Among the more uncomfortable conclusions was that concerns around accusations of racism, preserving “community relations,” and avoiding reputational damage may, at times, have contributed to hesitation in confronting what was in front of people.

In a large number of these cases, many of the convicted offenders were British Pakistani men.

State this plainly, and one frequently encounters an immediate rejoinder:

“But on average, most sexual offences in Britain are committed by white men in the home.”

Now pause for a moment. Suppose this were entirely correct as a statement about national incidence.

What exactly has it answered? The original claim was never:

British Pakistani men commit most sexual offences.

That would be a different argument altogether. The concern was that identifiable patterns appeared within particular organised networks and that institutions were unwilling to investigate them with sufficient honesty or urgency.

These propositions do not compete. One asks:

What does the overall distribution across the population look like?

The other asks:

Was there a concentrated pattern here that demands investigation?

Those are not rival explanations any more than weather and climate are rival descriptions. A population average may be entirely true and entirely beside the point.

And when a statistic is introduced as though it settles the matter, it ceases to function as evidence. It becomes permission not to look too closely. And it insinuates something about the person raising the concern: that the very act of questioning reveals something immoral about them.

That, I think, is the distinction people sense but struggle to articulate when hit with the national average fallacy. They know something is wrong with the statistic, but they do not know how to say what. And worse, the framing has already shifted the focus onto them. They are no longer the person raising a concern. They are the person who needs to explain why they have one.

Abusive Priests

We have lived through this faulty logic before. Consider the abuse scandals within the Church.

When they began to surface, people stood up and said:

Stop focusing on priests — statistically, most abuse happens elsewhere, in the home by non-priests.

We would recognise that immediately for what it is. Not an analysis but a deflection. That statistic may well have been and still be true. It was also, yet again, completely beside the point.

Because the concern was never that priests commit most abuse. The concern was: is there a concentrated institutional pattern being concealed?

So why do we keep losing that clarity the moment the pattern sits in a less comfortable place?

Moral Cover and Signalling

Because at some point in the exchange, something subtle occurs. The question changes. Almost nobody announces the change. It simply happens, and the discussion ceases to be:

Is this pattern real?

and through rhetorical sophism, things become:

What will it imply if I admit that it is real?

That is the move, where attention shifts from examining reality to managing identity and signalling virtue. A statistic no longer serves as an instrument for understanding the world. It becomes a means of creating distance from an uncomfortable conclusion.

Suppose somebody says:

There appears to have been a repeated pattern in some organised child sexual exploitation cases involving many British Pakistani offenders, and fears of accusations of racism may have contributed to institutional reluctance to investigate.

Notice how often the reply is not:

That pattern does not exist, and here is the data. Instead, the answer becomes:

We must be careful not to stigmatise communities. We must avoid maligning a culture. Victims deserve better. Besides, most sexual offences in Britain are committed by white men.

Notice what has happened. The question has changed.

We are no longer asking:

Was there a concentrated pattern? Were institutions reluctant to investigate? Were the victims failed? We are now asking:

What sort of person might I appear to be if I acknowledge this?

Imagine applying this to abusive priests and drunk drivers from above:

We must be careful not to stigmatise drivers. We must avoid maligning people who enjoy a drink. Victims deserve better. Besides, most road deaths in Britain are caused by sober drivers.

We must be careful not to stigmatise clergy. We must avoid maligning the church. Victims deserve better. Besides, most child abuse in Britain is committed by family members.

The national statistic enters not as evidence but as moral cover. While it may be true nationally, it fails to answer the question.

The Corrective

The corrective is almost embarrassingly simple. Remember Simpson’s Paradox and become suspicious whenever someone says:

“Overall…”

When someone reaches for a broad national average, ask one question:

That may be true nationally, but are you saying that because more cases may exist elsewhere, known patterns with known victims and known offenders become less important to investigate?

If it answers the concern, good, then it belongs in the conversation. If it doesn’t, then it isn’t working as evidence. At that point, the statistic may be doing several things at once: signalling moral superiority, obscuring the original question, reflecting a misunderstanding of what averages can tell us, or revealing how quickly the fear of appearing prejudiced can override our normal standards of reasoning.

A mature society should be able to hold two things in its head at the same time:

That most members of any group are innocent and decent.

And that concentrated patterns of offending should be investigated honestly, wherever they appear and whoever they implicate. These two truths do not compete. They depend on each other. The first is what prevents pattern recognition from curdling into prejudice. The second is what stops compassion from fermenting into cowardice.

Naming a pattern is not collective guilt. Refusing to see one is not justice.

If every uncomfortable observation is treated as evidence of racism, then racism ceases to mean anything, and reality becomes impossible to discuss. Once accusation replaces argument, the facts stop mattering, and only moral posturing remains.

Truth does not become prejudice simply because it makes us uncomfortable.