The Second Cohort
Southampton repeats a finding — and takes away the comfort of a false premise.
"If the Jew did not exist, the anti-Semite would invent him."
— Jean-Paul Sartre, Anti-Semite and Jew (1946)
Of the 37 people detained during the disorder in Southampton on 2 June, 22 had previously been reported to police for offences associated with intimate partner violence. That is the substance of the FOI response Hampshire constabulary released on Friday, reported by the Guardian. Fifty-nine per cent. In April I wrote here about the equivalent figure from the 2024 riots — roughly two in five of those arrested had a prior domestic abuse report — and argued that it was not a curiosity of that particular summer but a visible edge of something structural. A single striking number is an anecdote with a denominator. Southampton is the second cohort, and second cohorts are where arguments either firm up or fall over.
What the request returned
The sequence, briefly, for anyone who has managed to avoid it. Henry Nowak, eighteen, was stabbed by Vickrum Digwa, who then told police that he had himself been the victim of a racist attack. Officers arrested Nowak. Bodycam footage released after Digwa was jailed for life shows Nowak handcuffed as he lay dying, telling officers repeatedly that he had been stabbed. The footage went out; on 2 June several hundred people gathered outside a city-centre police station and then set off in the direction of the murder scene and the street where Digwa's family lived. Twelve officers were injured. Vehicles and nearby property were damaged. Thirty-three people have so far been charged. Hampshire has referred itself to the Independent Office for Police Conduct, which is examining whether officers failed to recognise that Nowak needed urgent medical help and why he was arrested and handcuffed before he was given first aid; two officers are under investigation for potential gross misconduct.
Against that background, the arrest cohort. Of the 37 detained, 22 had come to police attention before for intimate partner violence. The individual records the Guardian set out are of the type the April piece described: a thirty-four-year-old given three years for punching riot shields, who had pleaded guilty in 2023 to assaulting a partner he pinned against a bedroom wall; a twenty-nine-year-old with nineteen previous convictions, sentenced to three years for throwing a bin and a metal chair at officers, whose history includes knocking a partner unconscious and breaking her teeth. Farah Nazeer of Women's Aid, responding to the figures, made the point that public violence and private violence are correlated and that nobody working in the field finds this surprising. She is right that nobody in the field finds it surprising. That is rather the problem: it has been operational knowledge for thirty years and analytical knowledge for about eighteen months.
What a small number is worth
Thirty-seven is a small cohort and 59% is a fragile-looking headline, so it is worth being precise about what survives contact with the arithmetic. A 95% interval around 22 of 37 runs from roughly 43% to 74%. That interval sits above the 2024 national figure of about 40% — a cohort this domestic-abuse-heavy would arise from a 40% population about one time in eighty — but it sits comfortably inside the spread of individual 2024 local cohorts, where Bristol reported more than two-thirds. The honest summary is that Southampton is at the high end of a distribution we have already seen, not that it represents an escalation. Anyone reporting 59% as the new number, higher than last year's, is over-reading a sample of thirty-seven men.
The replication is the finding, not the point estimate. A second cohort, a different force, a different city, a different trigger, eighteen months apart, drawn from a crowd assembled by a different set of local networks — and the same disproportion appears. That is what moves a claim from "there was something odd about the summer of 2024" to "this is what these crowds are made of."
Prior domestic-abuse record among disorder arrestees, two waves.
The standing objection to all of this is selection. Police identify rioters from bodycam and CCTV, and the people they can put names to fastest are the people already in the system. A cohort assembled by that method will over-represent men with prior police contact of every kind, domestic abuse included. This objection is real and it is not fully answerable from the Southampton data alone. It is, however, largely answered by the fourth row of the chart: about one in five of those arrested in 2024 were reported for domestic abuse in the period after the riot. Forward-looking incidence cannot be produced by backward-looking identification bias. It can be inflated somewhat — a man who has been through the courts for violent disorder is a man whose partner may be likelier to be believed, and likelier to call — but the mechanism that explains the retrospective figure does not explain the prospective one.
This time the grievance was real
Here is what Southampton adds that Southport did not, and it is the reason I abandoned the piece about court reporting.
The 2024 disorder was triggered by a falsehood. What circulated in the hours after the Southport murders — about who the killer was and how he had arrived in the country — was untrue, and was known to be untrue very quickly. This gave everyone a convenient handle. The crowds had been mobilised by a lie; remove the lie and you remove the crowd; the whole episode could be filed under misinformation, a problem with a technical shape and, in principle, a technical fix.
Southampton removes that handle. Something genuinely went wrong. A stabbed eighteen-year-old was handcuffed while he was dying and told officers what had happened to him while they treated him as the offender, on the strength of his killer's false claim. Two officers are under investigation for gross misconduct and the IOPC is examining precisely the failures that the footage appears to show. Whatever the investigation concludes, the public did not have to be deceived into thinking there was a case to answer. There was one.
And the crowd looked the same. If anything it was more concentrated: a higher domestic-abuse share than the national 2024 figure, from an event with a factual grievance at its centre. The truth-value of the trigger did not filter who turned up.
That is the sentence to sit with. In April I argued that the overlap between private and public violence reflects a shared disposition — the enforcement of an order the perpetrator believes he is owed — and that grievance supplies the occasion rather than the motive. Southport was consistent with that but did not test it, because a fabricated grievance is exactly what the theory would predict a disposition to seize on. Southampton tests it. Given a real injustice, an actual failure of policing, a genuine subject for public anger, the same population turned out and did the same things. Sartre's line has to be inverted: this time nobody needed to invent anything, and it made no observable difference to the composition of the crowd.
The behaviour on the night is the corroborating detail. Several hundred people did not stay outside the police station, which is where the institutional failure was located and where a demand for accountability would have addressed itself. They moved toward the street where the killer's family lived. Family members who had done nothing were the more satisfying target, because the object was never redress. It was punishment, applied to whoever was proximate and belonged to the right category. That is the same grammar as the closed fist at home: the person available to be punished stands in for the order that has been violated.
Two claims that are not the same claim
Both available misreadings of the FOI data were in circulation within hours of publication, and they are mirror images.
The first treats the 59% as a verdict on the grievance. It is not. Whether Hampshire police failed Henry Nowak is a question about what officers did on the night they found him, and it will be answered by the IOPC on the evidence of the bodycam and the sequence of decisions taken. The criminal records of the people who later threw bins at riot shields have no bearing on it whatsoever. A police force does not acquire a clean bill of health because the men who came to shout at it turned out to be domestic abusers. The temptation to use the statistic that way is strong, and it should be resisted for the ordinary reason that it is a non-sequitur, and for the tactical reason that people notice when an institution answers an allegation about itself by producing the criminal records of its critics.
The second reads the causation backwards: because the grievance was legitimate, the crowd inherits its legitimacy. It does not. A real failure by the police does not transfer virtue onto men who travelled to a city centre to punch shields, and it certainly does not extend to marching on a family's home. The claims of "two-tier policing" made after the footage was released by Nigel Farage and others have more purchase in this case than they did in 2024 — that is precisely why the distinction matters. You can hold that officers made serious errors and that the men arrested that night were, in substantial part, drawn from a population with a documented history of hurting women. Neither fact does any work on the other. The desire to collapse them in one direction or the other is the whole of the difficulty here.
The instrument is wrong
We now have two cohorts. We have them because a newspaper filed freedom of information requests, which is not a research method. It is a method for getting one number out of one force about one event, defined however the person answering the request chose to define it — and "reported for offences associated with intimate partner violence" is a category doing a great deal of unexamined work, spanning everything from a withdrawn allegation to a conviction for strangulation.
What is missing is the denominator. Fifty-nine per cent of thirty-seven men is a striking figure only against some background rate, and nobody has published one. The right comparison is not the general population, in which the prevalence of recorded domestic abuse perpetration is low; it is men of similar age, from similar neighbourhoods, with similar records of general offending. It is entirely possible that the comparison group runs at forty per cent and that what the FOI figures capture is mostly the ordinary versatility of offenders rather than anything specific to the disposition I have been describing. I do not think that is the case — the mass-shooter literature, where the same comparison has been made properly, points the other way — but I cannot demonstrate it from these data, and neither can anyone else.
The fix is not complicated and does not require new legislation. A standing protocol: for every public-disorder cohort, a pre-specified record linkage between arrest data and domestic abuse reports, protective orders and risk assessments, with a matched comparison group and an interval published alongside the estimate. Run by the Home Office or a university with the access, not by whichever journalist thought to ask. Otherwise each event yields one headline number, everyone argues about it for a week, and the next event starts the cycle again with no accumulated knowledge — which is roughly where we now are, one year and one cohort on.
What follows, and what doesn't
The Home Office response to the Southampton figures pointed to its violence against women and girls strategy and to the rollout of domestic abuse protection orders. These are reasonable instruments and they are aimed at individuals: this man, this partner, this risk. The finding is about a population. If the men who make up disorder cohorts are substantially the same men who appear in domestic abuse records, then public-order planning and domestic abuse work are not two policy areas that occasionally intersect. They are one problem being addressed by two departments that do not share a spreadsheet.
What does not follow — and I want to be exact, because this is where the argument gets dangerous — is that a domestic abuse record should be treated as an indicator of public-order risk in any operational sense. The inference does not run that way. Most men on those databases will never go near a riot, the base rates do not support individual prediction, and building a watchlist out of this finding would be both ineffective and an obvious injustice to the large majority. Being reported for domestic abuse is not a licence to be policed as a rioter-in-waiting. The finding licenses seriousness and it licenses resources. It does not license suspicion of particular people.
In April I ended with the observation that the thrown brick and the closed fist tend to belong to the same hand. Southampton adds one line to that, and it is not a comfortable one: the hand does not need to be lied to. Give it a real grievance, an actual failure, a genuine subject for public anger, and it does the same thing it was always going to do. The lie in 2024 was never the mechanism. It was just the nearest available occasion.