How to make a criminal

Core idea

A criminal is not born; a criminal is assembled — slowly, by adversities and choices that are mostly outside the offender's control at the moments when they matter most. The most durable finding in criminology is also the most uncomfortable one for the popular imagination: people who do harm have, on average, had more harm done to them. Childhood maltreatment, household chaos, school failure, delinquent peer networks, and limited legitimate opportunity stack into a cascade. The cascade does not force offending — protective factors regularly interrupt it — but it loads the dice.

Canter's framing: The "born criminal" idea survives because it is convenient. It absolves the rest of us of any obligation to understand the developmental pathway, and it gives the criminal justice system a target it can punish without having to fix anything upstream.

A short history of a bad idea

Cesare Lombroso, writing in 1876, claimed that criminals were a distinct biological type — atavistic throwbacks identifiable by sloping foreheads, asymmetric faces, long arms. His evidence was post-hoc measurement of prisoners with no control group; his theory was demolished within a generation by Charles Goring's matched comparison studies in 1913. And yet the idea refuses to die. Every decade reinvents it: somatotypes in the 1940s, XYY chromosomes in the 1960s, the "warrior gene" in the 2000s, predictive risk algorithms today. The form changes; the wish does not.

Why it matters

If criminality is a fixed trait of kind of person, the rational response is incapacitation — lock them up, screen them out, never let them near your neighbourhood. If criminality is a developmental outcome with identifiable upstream causes, the rational response is prevention — invest in children, families, schools, and neighbourhoods long before the police ever meet the eventual offender. The two policy worlds barely talk to each other, and they spend public money in opposite directions. Choosing between them is not a moral preference; it is an empirical question, and the evidence has been pointing the same way for fifty years.

What changes when you adopt the developmental view

You stop asking "what is wrong with that person?" and start asking "what happened to that person, and when?" You notice that the same risk factors keep appearing across different offender populations — violent, sexual, acquisitive, fraudulent — which suggests a general developmental story with offence-specific overlays, not a separate biology for each crime type. You also stop being surprised that most "criminals" desist on their own by their late twenties, because you understand offending as a phase of life rather than an identity.

Key takeaways

Mental model

Copy sourceZoom

Practical application

Read offenders developmentally, not categorically

When you encounter an offender — in a case file, a news story, a courtroom — resist the urge to ask "what kind of person does that?" Ask instead "what happened to this person, and at what age?" The categorical question produces a label that explains nothing; the developmental question produces a timeline that explains a great deal. The same instinct applies in reverse: when someone with a difficult biography has not offended, ask what protected them. Both directions teach you something the categorical view never can.

Use the age-crime curve as a sanity check

The curve also explains why long sentences for offenders in their late twenties are economically irrational — most would have desisted in the next few years anyway. The marginal incapacitation gain is small; the cost of the prison bed is large.

Distinguish risk factors from causes

Risk factors are statistical correlates. Causes are mechanisms. ACE scores predict offending across a population, but they do not predict it for any individual: most children with high ACE scores do not become offenders. Treat risk factors as populations to invest in, not individuals to monitor. The predictive-policing version of this confusion — using risk scores to target individual surveillance — has a poor empirical record and an even poorer civil-liberties one.

Build interventions around turning points

Example

Consider two boys, A and B, born the same week in the same neighbourhood.

A's parents are stable and present. School notices early that he struggles with reading; an attentive teacher refers him for support and he catches up by age 10. His friendship group skews toward kids whose parents also know each other. In early adolescence he experiments — shoplifting a chocolate bar at 13, getting drunk at a park at 15, smoking some cannabis at 16. None of it shows up in any official record. By 19 he is at a polytechnic, by 23 in a job, by 27 in a relationship. He has been, statistically, an offender, but no one will ever call him one.

B's father leaves when he is 4, and his mother's new partner is intermittently violent. The same reading problem goes unnoticed because the school is under-resourced and B is also disruptive — easier to exclude than to assess. By 11 he is in a peer group of older boys already shoplifting; by 13 he has his first caution; by 15 he has been in court three times. At 17, the modal age on the age-crime curve, he is in a youth offender institution. The system now has a thick file on him and a label.

Both boys experienced the same neighbourhood, the same age, the same year. The difference is not in their nature. It is in the sequence of small interventions that did or did not happen — the attentive teacher, the stable household, the prosocial friendship group, the timely reading support. Each was a low-cost interruption of the cascade. None of them happened for B. The criminal justice system meets B at age 17 and is then asked to undo what could have been prevented at age 7 for a fraction of the cost.

The point is not that B has no agency. He has plenty. The point is that the probability distribution of his choices was shaped long before he made any of them, and the developmental view lets us see how.

Related lessons