It's who they know

When police record a crime, they capture the who, the what and the where. What they have been slower to capture — and to use — is the with whom. A new study from the University of Cambridge suggests that may be the single most consequential gap in how British policing approaches youth violence.

The research, led by Professor Paolo Campana at Cambridge’s Institute of Criminology, analysed more than 200,000 police records from Cambridgeshire Constabulary covering March 2018 to October 2021.

The dataset spans Peterborough and Cambridge alongside several towns where county lines operations have been active — a geographical spread that gives the findings real-world weight beyond a purely academic setting. Within it, more than 10,000 young people were recorded as crime victims and around 6,000 as suspects. Nearly one in four young victims had also been identified as a suspect in a separate offence, a dual status that sits at the heart of what makes serious youth violence so resistant to conventional intervention.

The core argument is straightforward: the young people most likely to commit serious violence are not simply those with the longest individual criminal records, but those most deeply embedded in criminal networks.

By mapping co-offending relationships — connections formed whenever two individuals appear together as suspects in the same recorded crime — the researchers identified a group they describe as “super-connected”: a top five per cent of youth suspects aged ten to eighteen with more than seven connections to other criminal suspects of any age. The maximum recorded was 21.

Super-connected young people are almost five times more likely to carry knives than the average youth suspect. They are nearly three times more likely to commit violent crime, seven times more likely to commit robberies, and nine times more likely to be involved in organised crime. Crucially, they are also far more likely to become victims themselves: nearly 60 per cent of young people who suffered knife-related injuries in the dataset were also recorded as suspects in separate offences.

For each additional co-suspect in a young person’s criminal record, the likelihood of violent offending rises by around 30 per cent; for knife violence specifically, by 19 per cent. Each connection also increases the probability of becoming a victim of violence by 7 per cent and of knife injury by 11 per cent.

“Traditional approaches treat young offenders in isolation, focusing on individual risk factors such as age, background and previous behaviour,” Campana said. “In reality, we are missing a crucial layer, as youth violence is deeply social, driven by relationships and peer groups.”

Troubled history

The intervention logic the research points toward is not about targeting the most prolific young offenders but about identifying and engaging with young people at the point they become embedded in criminal networks, before the seriousness of their own offending escalates. That means acting on network position, not just offence history.

Network analysis is not new to policing. It has been applied in counter-terrorism and serious organised crime for years, and the Metropolitan Police Service have operated something in this space for over a decade. The Gangs Matrix, introduced in the aftermath of the 2011 riots, attempted to map gang-affiliated individuals by risk score, drawing on police intelligence, offending history and associations. At its peak it held data on around 3,600 individuals, the overwhelming majority of them young black men.

The matrix itself soon became its own cause of concern. Amnesty International’s 2018 investigation found that nearly two-thirds of those on the database had no recorded violence on their record. The criteria for inclusion were opaque, the data-sharing arrangements poorly governed, and the racial disproportionality severe enough to prompt the Information Commissioner’s Office to open a formal investigation.

In 2022, the Metropolitan Police agreed a series of remedial actions with the ICO, including tighter criteria for entry and retention, clearer processes for removal and greater transparency about how the data was used.

Those failures matter for how any new network mapping methodology is received. The Cambridge approach is methodologically distinct as it derives connections from structured police crime records rather than intelligence assessments, and it frames the output explicitly as a tool for directing intervention, not enforcement.

That said, the civil liberties questions are the same, and they are sharpest when the subjects are children. Any deployment at scale would require independent oversight, clear legal gateways for data sharing between agencies, time-limited retention policies and meaningful routes for individuals to challenge their inclusion. None of that infrastructure currently exists in a form fit for purpose.

Operational reality

England and Wales now has a network of Violence Reduction Units covering most major urban areas, funded through the Home Office and tasked with coordinating the kind of cross-agency response that serious youth violence demands. They are the natural institutional home for network intelligence of this kind — and the place where the gap between research promise and operational reality is most visible.

In principle, a VRU analyst with access to co-offending data could run the kind of network mapping Campana’s team demonstrates and produce a ranked output: young people by connection count, flagged where those connections include adult offenders with records for serious violence or organised crime. That list would then inform outreach, not enforcement, prioritising which young people receive early contact from youth workers, mentors or education specialists before their own involvement deepens.

In practice, several things get in the way. Police data systems are not designed for network analysis. The records exist, but extracting and mapping co-offending relationships across a force area requires either bespoke analytical tools or significant manual processing.

Data sharing between police and local authority services — the youth workers and schools who would deliver any intervention — operates under governance frameworks that were not built with this use case in mind. And VRUs, despite their cross-agency mandate, often lack the analytical capacity to move beyond descriptive statistics into the kind of relational mapping the Cambridge study demonstrates.

The researchers suggest the methodology could also be used to track whether interventions are working — by monitoring whether a young person’s network position changes after contact, whether they become less central, less connected to high-risk adults. That is a more sophisticated evaluation approach than most VRUs currently apply, but it is not technically out of reach.

Limits of the evidence

The Cambridgeshire dataset, while substantial, does not reflect the policing environments where youth violence is most acute. London, Birmingham, Manchester and other major cities have different network densities, different gang structures and different institutional histories. Campana and colleagues say the study should be replicated using data from those areas, though they believe the underlying dynamics will hold, a claim supported by comparable findings from co-offending network studies in American cities.

It is also worth being clear about what network mapping cannot do. It can identify young people at elevated statistical risk; it cannot determine intent, individualise that risk or substitute for professional judgement. A teenager appearing in police records for the first time may carry low individual risk on every conventional measure; connected to six adults with organised crime histories, that assessment shifts.

But the shift is probabilistic. Acting on it responsibly, meaning engaging rather than criminalising, diverting rather than surveilling, requires practitioners who understand the difference and institutions that are designed to support it.

“Understanding where a young person sits within a crime network, and who they are connected to, should inform how and when we intervene,” Campana said. “That means building the capacity to map and share network intelligence across police, youth services, schools and community organisations in a responsible way. Violence does not happen in isolation, and tackling youth violence means tackling the networks that sustain it.”

The report ‘Breaking networks of youth serious violence’ is online on the University of Cambridge website.