Youth diversion has a legitimate place in Victoria’s justice system. A warning or caution can be the proportionate response to a young person’s first, low-level offence. It can avoid unnecessary contact with courts and custody without compromising community safety.

But diversion is a tool, not a complete youth-crime strategy.

The policy error begins when evidence about low-risk young people is extended to repeat, serious or deliberately commissioned offending. Those groups are not interchangeable. Victoria needs a system capable of telling them apart—and responding accordingly.

What changed during my time in policing

When I started policing in 1999, most of the young people I encountered for offending had been caught shoplifting. My response area included a major shopping centre, so my experience undoubtedly over-represented that category. Typically, however, these were children in their early teens apprehended by a store loss-prevention officer.

The next level commonly involved a small amount of cannabis or possession of alcohol while underage. At the more serious end was a much smaller group, usually in their later teens, stealing cars. The worst cases involved burglaries, armed robberies and, occasionally, serious violence involving a knife.

At law, a child could be considered for a caution or diversionary process. In practice, that option diminished as the offending became more serious or the history became longer. That was an important exercise of discretion: distinguish the child who had made an early mistake from the young person presenting a repeated or escalating risk.

I processed hundreds of young people by way of caution for a first offence. I can only assume that, for most of them, their contact with me was their only contact with the criminal justice system. Their “life of crime” began and ended with an attempted theft of a video game, earrings or some other low-value item.

That assumption is not blind faith. It is informed by what I saw when their parents arrived at the police station. Most were furious with their child. They understood the seriousness of the moment and made it clear that the police caution would not be the only consequence.

There were exceptions. Some parents made excuses. Others could not be bothered attending, leaving police to arrange for an independent adult—often a complete stranger—to sit in on the interview. Those cases did not determine a child’s future, but they could reveal a weaker protective environment. It was less surprising when police encountered the same young person, or sometimes the same family, again.

This is anecdotal experience, not a controlled study. It does, however, illustrate why selection effects matter. The family engagement visible at the police station may influence both whether a caution is considered appropriate and what happens after it, yet it is not easily captured in administrative data.

Vehicle theft at the start of my career illustrates the difference. Older vehicles such as the Mazda 626 and Ford Telstar were notoriously easy to steal. They were often taken at night using simple tools, usually for transport rather than as part of a sophisticated criminal market. The vehicle was the means of getting somewhere, not necessarily the prize or a tool within a broader offending plan.

Manufacturers responded with steering locks, alarms, immobilisers, vehicle identification technology and tracking. That made the older method harder—but it did not remove the demand or the incentive. Offending adapted.

By the time I returned to Crime Command as a Detective Senior Sergeant, police were dealing with what we called Youth Networked Offending: the institutional term for group-based offending that many members of the public would simply describe as gangs. The offence labels were familiar, but the organisation, methods, targets and level of violence had changed substantially.

It is important not to rewrite that history. Serious youth offending did not suddenly appear in 2026. By 2015, police were already confronting young people involved in aggravated burglaries, armed robberies, serious violence and coordinated offending across identifiable networks. The seriousness was already there.

What continued to change was the reach and mobility of the offending. Earlier young offenders often remained within their own suburb or surrounding area, constrained by familiarity, transport and local associations. Networked groups became increasingly willing and able to cross the entirety of Melbourne—and travel into regional Victoria—to identify targets, steal vehicles and commit further offences. A crime series that began in one police region could quickly move through several others.

That geographic reach matters. Local policing models can miss the pattern when each offence is initially treated as an isolated event in a different area. The network becomes visible only when intelligence, co-offending, vehicle movements, communications and methods are analysed across boundaries.

The vehicles being targeted were newer, faster and more valuable. They carried status, could be used to commit other offences and, in some cases, created opportunities for rebirthing and sale. They were also more difficult to steal by old methods, so offenders found more ingenious—and sometimes more violent—ways to obtain them.

By about 2015, the so-called “Audi key” had become part of the operational picture. In the cases police were seeing, a spare or emergency key supplied with a new vehicle had sometimes been left inside the car after purchase, rather than moved to a secure place in the home. That simple oversight created an opportunity. Once owners became aware of the vulnerability and began removing the spare key, offenders did what career offenders repeatedly do: they adapted. Some shifted to entering homes for the primary keys; later methods included electronically enabled theft using devices available outside traditional criminal supply chains. The detail is less important than the pattern. Each time a security gap closed, the offending method evolved.

Stolen vehicles could then support further offending, generate attention online or be used in ram raids against increasingly hardened retail targets. Supermarkets, tobacconists and other businesses improved their physical security; offenders responded by using stolen vehicles to defeat it.

Social media changed behaviour as well. Earlier offenders generally feared seeing their face in a newspaper or on television. Some later offenders sought the visibility, recording or publishing their own crimes—including extremely serious conduct—to build status and credibility.

This is not nostalgia for a simpler period of crime, nor a claim that serious youth offending is new. It existed in 1999, was clearly present in networked form by 2015 and caused real harm throughout. The operational change is that technology, mobility, online status and adult criminal markets can now connect a young offender to more capable methods, more valuable targets, a much wider operating area and a much larger audience.

That history is why youth offending cannot sensibly be treated as a single category. The offence label may look familiar while the method, motivation, network and risk are fundamentally different.

What the evidence actually supports

The Crime Statistics Agency’s 2017 study, The Cautious Approach, examined 5,981 alleged offenders aged 10 to 17. At first glance, its findings appear decisive: 26.8 per cent of cautioned young people were recorded as reoffending within 12 months, compared with 57.6 per cent of those charged.

After propensity-score matching, the gap narrowed. The matched comparison found reoffending rates of 35.9 per cent for cautioned young people and 47.8 per cent for those charged.

I am surprised that recorded reoffending among cautioned children was that high. I do not reject the figure, but I would want to see the deeper data behind it before drawing broad policy conclusions. What offences followed? Were some young people diverted again for different offending? What did the trajectory look like after two years or five years rather than only 12 months? Which people in the cautioned group accounted for repeated events?

That remains an important finding, but it is not proof that cautioning will produce the same result across the entire offending population.

The groups were different before the justice response was applied. Most cautioned young people had no prior history. The charged group included many who had already received a caution or had accumulated substantial prior offending. Only 917 matched pairs were included in the adjusted comparison, leaving much of the prolific end outside the like-for-like analysis.

Selection also matters. A young person who admits an offence and has an engaged parent available may differ in important ways from one who does not. Family stability, contrition, demeanour and informal police interventions are difficult to capture in administrative data, but they can influence both the decision to caution and the likelihood of later offending.

The fair conclusion is not that diversion fails. It is that lower reoffending among diverted young people does not, by itself, prove the diversion caused the difference. The policy is applied to the people already most likely to succeed, many of whom may stop because they mature rather than because a program transformed them. The apparent success of that majority cannot be used to obscure the harm caused by the smaller group who reoffend. A percentage is an analytical tool; every new offence has a victim who experiences it as an event, not a data point.

The age–crime curve may be doing the program’s work

Most young people who offend do not become persistent adult offenders. Offending tends to rise during adolescence and fall as people mature. Time, changing peer groups, work, relationships and ordinary neurological development will move many young people away from offending without an intensive program.

That is not a minor qualification. It goes to the heart of whether diversionary and youth programs can claim to have changed behaviour.

When a program accepts young people already likely to desist, a high apparent “success rate” does not establish causation. Without a credible comparison group, it is impossible to separate the program’s contribution from natural maturation and participant selection. The program may have helped. It may also be taking credit for what time would have done anyway.

Even if a sound evaluation found a ten-percentage-point improvement—for example, 10 additional participants in every 100 did not reoffend within 12 months because of the intervention—we would still need to ask whether that is a successful outcome. What did the program cost per additional non-reoffender? What was the seriousness of the offending prevented? Did the effect survive for two or five years? Were victims and community safety included in the return-on-investment calculation?

A ten-point causal improvement may justify investment. It may not. The answer depends on cost, duration, harm prevented and how the intervention compares with other uses of the same money. Announcing a participation number or an uncontrolled “success rate” is not an answer.

When the measurement creates the result

There is a second problem. Diversionary and non-punitive systems can partly manufacture their own apparent success when the outcome measure depends on criminal justice processing.

If authorities intentionally charge fewer young people, the recorded number of charged young offenders will fall by design. That may be good policy, particularly for minor first offending, but the statistical reduction cannot then be presented as independent proof that underlying behaviour changed.

The same measurement effect appears when conduct is moved from a criminal framework into a health or administrative framework. Recorded drug charges may fall immediately because police and courts are processing fewer people—not necessarily because drug use or drug-related harm fell at the same rate. A change in classification can be justified while still changing the statistic used to judge success.

This is why evaluation must measure behaviour and harm, not merely system contact. It should include victimisation, offence severity, self-reported conduct where reliable, repeat events, emergency and health consequences, and outcomes over several years. Otherwise, policy risks proving itself by changing the definition of the thing being counted.

The Bear Patrol test

The Simpsons captured this causal error neatly in the 1996 episode Much Apu About Nothing. After one bear wanders into Springfield, the town creates an elaborate and expensive Bear Patrol. When no further bears appear, Homer treats their absence as proof that the patrol works. Lisa demonstrates the flaw with a supposedly tiger-repelling rock: there are no tigers nearby either, but their absence does not prove the rock caused it. Homer misses the lesson and offers to buy the rock.

It is funny because the reasoning is familiar. An intervention is introduced, the feared outcome is not observed, and the intervention claims the credit—without asking what would have happened anyway or whether the expense produced any additional benefit.

Youth programs should be required to pass the Bear Patrol test. A lower offending rate after participation is not enough. Government must show the program caused an improvement beyond natural desistance, selection effects, changed charging practices and broader crime trends. It must then show that the additional benefit was worth the cost.

The Victorian Auditor-General’s September 2026 review shows why this standard matters. Victoria had spent more than $40 million on the Youth Crime Prevention Program since 2016, supporting more than 2,700 participants. The Department of Justice and Community Safety described it as proven and evidence-based. VAGO found no evidence that the program was reducing offending.

The audit found that the department’s evaluations had not measured performance against meaningful targets or benchmarks. A claimed 29 per cent reduction relied on data covering COVID-19 lockdowns, when crime generally fell. In VAGO’s own matching exercise, more than two-thirds of identified participants offended within six months of starting the program. Among participants who reoffended, neither the frequency nor severity of alleged offending reduced over that period.

That does not prove that mentoring, case management, education, employment support or other individual services have no value. It proves that government cannot honestly claim this program reduced offending on the evidence it collected.

After more than $40 million, “it may help” is not an adequate accountability standard. A program should have a defined counterfactual, measurable community-safety outcomes, cost per additional positive result and a sunset clause if those results are not demonstrated.

Doing less to a low-risk child can still be sound policy. It is cheaper, proportionate and less likely to cause harm. But restraint should be defended as restraint—not embellished into an unproven claim of transformation, and not used to justify the same response to repeated aggravated burglary, serious violence or paid criminal tasking.

Victoria’s new framework needs operational discipline

The Youth Justice Act 2024 was fully implemented on 30 September 2026. Current Victoria Police guidance describes warnings, cautions and diversion as responses for low-level, less serious offending and first-time offenders. It also states that police will continue to charge and seek remand for serious and violent offending.

That distinction is important and should remain operationally clear.

A minimum-intervention framework can improve consistency and prevent unnecessary escalation. But frontline decision-making must retain sufficient weight for offence seriousness, prior conduct, victim harm, escalating patterns and credible intelligence about co-offending or adult direction.

The test should not be whether diversion is generally desirable. It should be whether it is appropriate for this young person, this offence and this risk.

The June 2026 official data also requires care. The Victorian Government reported that, in the year to March 2026, the youth offending rate fell by six per cent and the criminal incident rate by nearly three per cent. Those are welcome movements. They do not, by themselves, identify which policy caused the change, particularly where reforms commenced near the end of the reporting period.

Policy should be judged over time using measures that matter to the community: repeat victimisation, serious harm, network disruption, compliance, perceived safety and the persistence of high-risk cohorts.

Community safety is more than a recidivism rate

The debate about youth justice too often becomes esoteric: which intervention produces the lowest measured recidivism, and by how much? That is an important question, but it has quietly become the objective rather than one means of achieving it.

The primary objective should be community safety.

Reducing reoffending contributes to that objective, but it does not fully describe it. The type and seriousness of the offence matter. So do repeat victimisation, the visibility of disorder, the apparent reach of criminal networks and the confidence people have that serious offending will bring a credible response.

Perception is not a substitute for evidence. Crime statistics still matter, and public anxiety can be influenced by selective reporting or dramatic imagery. But perceptions of safety have real consequences. People change how they travel, where they shop, whether they use public spaces and how secure they feel in their own homes. Fear of crime can damage wellbeing even where a broad crime category is declining.

The Monash Freeway arrests in October 2026 demonstrate the tension. Victoria Police reported that an alleged aggravated burglary in Brighton East involved the theft of two vehicles, followed by another alleged home invasion and a pursuit across Melbourne’s east and south-east. Specialist police intercepted one vehicle on the freeway. Four children and two adults were arrested; the children were later charged with offences including aggravated burglary, armed robbery and vehicle theft.

The police operation deserves recognition. A high-risk arrest was completed without serious injury. Some members of the public will feel reassured by the scale and capability of that response.

Others will draw a more troubling conclusion from the same images: that specialist police, the Air Wing and a major freeway interception are now what can be required to stop alleged youth offending linked to stolen vehicles and home invasions.

Both reactions matter. Public confidence depends not only on whether police can bring a dangerous incident to an end, but on whether government and agencies can show that the underlying pattern is being reduced.

This is also why a 12-month recidivism percentage cannot settle the argument. Even if roughly seven in ten diverted young people are not detected for another offence during that period, legitimate questions remain:

What offences were committed by those who did reoffend?
Were further warnings or diversionary outcomes counted consistently?
What happened over two years and five years?
Was harm concentrated in a much smaller repeat-offending cohort?
Were victims and community confidence measured at all?

My experience suggests that most first-time shoplifters I cautioned probably went on to ordinary lives: jobs, families and responsibilities. Some may now be the people whose homes are burgled or whose cars are stolen. That is precisely why diversion should remain available for the right child—and why policy cannot wait indefinitely for a hardened repeat offender to make the right choice next time.

Prolific offenders require a different response

The central weakness in broad youth-crime debate is its tendency to treat “young offenders” as one population.

They are not.

There is a major operational difference between:

a first-time, low-level offender who is likely to desist;
a young person with complex welfare needs and an escalating pattern;
a prolific offender responsible for repeated serious harm; and
a young person being recruited or tasked by an organised criminal network.

A risk-stratified system should identify these differences early. Victoria should maintain a named, harm-ranked and regularly refreshed cohort of prolific young offenders, with multi-agency ownership and clear thresholds for intensive supervision. Where the legal test is met and the risk warrants it, remand must remain available.

The objective is not punishment for its own sake. It is preventing further harm while concentrating expensive interventions on the people who present the greatest risk.

Community safety should be the primary statutory and operational objective. Rehabilitation matters because it can reduce future victimisation. It is a means to a public outcome, not a substitute for one.

When youth offending becomes a labour market

Some young people involved in serious crime are not simply offending with peers. They are being used.

Organised criminal groups can recruit young people to commit arson, steal vehicles to order, attack premises, move illicit goods or intimidate others. The young person carries the immediate risk while the adult organiser remains removed from the offence.

This changes the analytical question.

It is no longer enough to ask, “How do we divert this young person?” Agencies must also ask, “Who is buying this young person’s labour?”

Indicators may include offences with no clear personal benefit, adult-taught tradecraft, offending well outside the young person’s usual area, third-party funding of legal or bail support, and silence driven by fear rather than ordinary reluctance.

Volume offending and organised-crime contracting are not the same. Intelligence systems should flag “prolific”, “tasked” and “at risk of recruitment” separately. Conflating them produces poor decisions at both ends: unnecessarily harsh treatment for low-risk young people and inadequate disruption of adults who deliberately recruit children.

A better model for Victoria

Victoria does not need to choose between prevention and enforcement. It needs a model that uses each where it is most effective.

That model would include:

light-touch warnings and cautions for genuinely low-risk offending;
a statutory and measurable community-safety objective;
a harm-ranked prolific-offender cohort with intensive multi-agency management;
separate intelligence flags for organised-crime tasking and recruitment vulnerability;
enforcement focused on the adults who commission and profit from youth offending; and
public funding tied to independent, counterfactual program evaluation.

Diversion should remain available because proportionate justice matters. But proportionality works in both directions. A system that under-responds to repeated serious harm is no more balanced than one that over-responds to a first mistake.

The enduring lesson is that organised crime adapts. If criminal enterprises can obtain cheap, replaceable and legally vulnerable labour, they will use it. Victoria’s response must disrupt that labour market while preserving a sensible exit for young people who present little ongoing risk.

That adaptive mindset is central to my book, The Enterprise. The person who becomes a career criminal often comes to see crime as precisely that: a career and a business. To remain successful, they evolve the offences they commit, the commodities they trade and the methods they use. A security measure, policing strategy or market disruption may close one opportunity, but the committed criminal entrepreneur looks for the next one.

In the coming days, I will explore that proposition further through Victoria’s tobacco wars—an illicit market that shows how criminal enterprises change products, tactics and alliances while continuing to pursue profit. The Enterprise examines the structures and thinking that allow this adaptation. The book is available in Kindle and print editions.

If your organisation is reviewing its exposure to organised crime, financial crime or integrity risk, contact daniel@integritysolve.com and we will be in contact.

Youth crime in Victoria: diversion is a tool, not a strategy