Offending is part of a larger syndrome of antisocial behaviour that arises in childhood and tends to persist into adulthood. There seems to be continuity over time, since the antisocial child tends to become the antisocial teenager and then the antisocial adult, just as the antisocial adult then tends to produce another antisocial child. The main focus of this chapter is on types of antisocial behaviour classified as criminal offences, rather than on types classified for example as conduct disorder or antisocial personality disorder.

In an attempt to identify causes, this chapter reviews risk factors that influence the development of criminal careers. Literally thousands of variables differentiate significantly between official offenders and non-offenders and correlate significantly with reports of offending behaviour by young people. In this chapter, it is only possible to review briefly some of the most important risk factors for offending: individual difference factors such as high impulsivity and low intelligence, family influences such as poor child rearing and criminal parents, and social influences: socio-economic deprivation, peer, school, community, and situational factors.

I will be very selective in focussing on some of the more important and replicable findings obtained in some of the more methodologically adequate studies: especially prospective longitudinal follow-up studies of large community samples, with information from several data sources (e.g. the child, the parent, the teacher, official records) to maximize validity. The emphasis is on offending by males; most research on offending has concentrated on males, because they commit most of the serious predatory and violent offences. The review is limited to research carried out in the United Kingdom, the United States, and similar Western industrialized democracies. More extensive book length reviews of antisocial behaviour and offending are available elsewhere.⁽¹⁾

I will refer especially to knowledge gained in the Cambridge Study in Delinquent Development,⁽²⁾ which is a prospective longitudinal survey of over 400 London males from age 8 to age 40. Fortunately, results obtained in British longitudinal surveys of delinquency are highly concordant with those obtained in comparable surveys in North America, the Scandinavian countries, and New Zealand and indeed with results obtained in British cross-sectional surveys. A systematic comparison of the Cambridge Study with the Pittsburgh Youth Study showed numerous replicable predictors of offending over time and place, including impulsivity, attention problems, low school attainment, poor parental supervision, parental conflict, an antisocial parent, a young mother, large family size, low family income, and coming from a broken family.

Offending is defined as acts prohibited by the criminal law, such as theft, burglary, robbery, violence, vandalism, and drug use. It is commonly measured using either official records of arrests or convictions or self-reports of delinquency. The advantages and disadvantages of official records and self-reports are to some extent complementary. In general, official records identify the worst offenders and the worst offences, while self-reports include more of the normal range of delinquent activity. The worst offenders may be missing from samples interviewed in self-report studies. Self-reports have the advantage of including undetected offences, but the disadvantages of concealment and forgetting. By normally accepted psychometric criteria of validity, self-reports are valid. Self-reported delinquency predicted later convictions in the Cambridge Study. In the Pittsburgh Youth Study,⁽³⁾ the seriousness of self-reported delinquency predicted later court referrals. However, predictive validity was enhanced by combining self-report, parent, and teacher information about offending.

The key issue is whether the same results are obtained with both methods. For example, if official records and self-reports both show a link between parental supervision and delinquency, it is likely that supervision is related to delinquent behaviour (rather than to any biases in measurement). Generally, the worst offenders according to self-reports (taking account of frequency and seriousness) tend also to be the worst offenders according to official records. In the Cambridge Study between ages 15 and 18, 11 per cent of the males admitted burglary, and 62 per cent of these males were convicted of burglary. The predictors and correlates of official and self-reported delinquency were very similar.

Much is known about the epidemiology and development of offending and criminal careers, but there is not space to review these topics here.⁽⁴⁾ For example, the prevalence of offending tends
to peak in the teenage years, and an early onset of offending predicts a long criminal career. Offenders tend to be versatile rather than specialized, in committing not only different types of offences but also different types of other antisocial acts. While there is considerable continuity over time, in the sense that the most antisocial people at one age tend also to be the most antisocial at another, only about half of antisocial juveniles tend to become antisocial adults.

A risk factor is defined as a variable that predicts an increased risk of offending. For example, children who experience poor parental supervision have an increased risk of committing offences later on. Since risk factors are defined by their ability to predict later offending, it follows that longitudinal data are required to discover them. Risk factors tend to be similar for many different outcomes, including violent and non-violent offending, mental health problems, alcohol and drug problems, school failure, and unemployment. Protective factors are also important. They are defined as factors that predict a low risk of offending or that counteract risk factors.

An obvious problem is that it is not clear to what extent any risk factor is a cause of offending. It is important to investigate causal mechanisms linking risk factors and offending. The best way of establishing a cause is to carry out a prevention experiment tackling that risk factor; preferably a randomized experiment, because the random assignment of people to conditions in principle controls for all other influences on offending. If a prevention experiment was carried out in which parental supervision was improved, and if offending was reduced as a consequence, this would be powerful evidence that the risk factor of parental supervision truly had a causal effect on offending. However, most knowledge about causes comes from quasi-experimental analyses.

Because of the difficulty of establishing causal effects of factors that vary only between individuals (e.g. gender and ethnicity), and because such factors have no practical implications for prevention (e.g. it is not practicable to change males into females), unchanging variables will not be reviewed here. In any case, their effects on offending are usually explained by reference to other, modifiable, factors. For example, gender differences in offending have been explained on the basis of different socialization methods used by parents with boys and girls, or different opportunities for offending by men and women. Similarly, risk factors that are or might be measuring the same underlying construct as delinquency (e.g. physical aggression) will not be reviewed; the focus is on risk factors that might be causes. For simplicity, risk factors are reviewed one by one. Biological factors are not reviewed. I will not attempt to review additive, interactive, independent, or sequential effects of risk factors, although these are important issues. Nor will I review developmental theories of offending.

Hyperactivity and impulsivity are among the most important personality or individual difference factors that predict later offending.⁽⁵⁾ Hyperactivity usually begins before age 5 and often before age 2, and it tends to persist into adolescence. It is associated with restlessness, impulsivity and a short attention span, and for that reason has been termed the ‘hyperactivity-impulsivity-attention deficit’ or HIA syndrome. Related concepts include a poor ability to defer gratification and a short future time perspective.

Many investigators have reported a link between hyperactivity or impulsivity and offending. For example, in the Orebro (Sweden) longitudinal survey,⁽⁶⁾ hyperactivity at age 13 (rated by teachers) predicted violent offending up to age 26. The highest rate of violence was among males with both motor restlessness and concentration difficulties. The most extensive research on different measures of impulsivity was carried out by Jennifer White and her colleagues in the Pittsburgh Youth Study. This showed that cognitive or verbal impulsivity (e.g. acts without thinking, unable to defer gratification) was more strongly related to delinquency than was behavioural impulsivity (e.g. clumsiness in psychomotor tests).

In the Cambridge Study, a combined measure of hyperactivityimpulsivity- attention deficit was developed at age 8–10, and it significantly predicted juvenile convictions independently of conduct problems at age 8–10. Hence, HIA is not merely another measure of antisocial personality, but it is a possible cause, or an earlier stage in a developmental sequence leading to offending. Similar constructs to hyperactivity, such as sensation seeking, are also related to delinquency. In the Cambridge Study, the extent to which the boy was daring or took risks at age 8–10, as well as restlessness and poor concentration, significantly predicted convictions and high self-reported offending. Daring was consistently one of the strongest independent predictors of offending.

Low intelligence is an important predictor of offending, and it can be measured very early in life. In a prospective longitudinal survey of about 120 Stockholm males,⁽⁷⁾ low IQ measured at age 3, significantly predicted officially recorded offending up to age 30. Frequent offenders (with 4 or more offences) had an average IQ of 88 at age 3, whereas non-offenders had an average IQ of 101. All of these results held up after controlling for social class. Similarly, low IQ at age 4 predicted arrests up to age 27 in the Perry preschool project.⁽⁸⁾

In the Cambridge Study, twice as many of the boys scoring 90 or less on a non-verbal IQ test (Raven’s Progressive Matrices) at age 8–10 were convicted as juveniles as of the remainder. However, it was difficult to disentangle low intelligence and low school attainment. Low non-verbal intelligence was highly correlated with low verbal intelligence (vocabulary, word comprehension, verbal reasoning) and with low school attainment, and all of these measures predicted juvenile convictions to much the same extent. In addition to their poor school performance, delinquents tended to leave school at the earliest possible age (which was then 15) and to take no school examinations.

Low non-verbal intelligence predicted juvenile self-reported offending to almost exactly the same degree as juvenile convictions, suggesting that the link between low intelligence and delinquency was not caused by the less intelligent boys having a greater probability of being caught. Also, measures of intelligence and attainment predicted measures of offending independently of other variables such as family income and family size. Delinquents often do better on non-verbal performance tests, such as object assembly
and block design, than on verbal tests, suggesting that they find it easier to deal with concrete objects than with abstract concepts.

Low IQ may lead to delinquency through the intervening factor of school failure; the association between school failure and delinquency has been demonstrated consistently in longitudinal surveys. In the Pittsburgh Youth Study, Donald Lynam and his colleagues concluded that low verbal IQ led to school failure and subsequently to self-reported delinquency, but only for African-American boys. Another plausible explanatory factor underlying the link between low IQ and delinquency is the ability to manipulate abstract concepts. Children who are poor at this tend to do badly in IQ tests and in school attainment and they also tend to commit offences, mainly because of their poor ability to foresee the consequences of their offending and to appreciate the feelings of victims. Low IQ may be one aspect of cognitive and neuropsychological deficits in the executive functions of the brain.

Many different types of child-rearing methods predict offending. The most important dimensions of child rearing are supervision or monitoring of children, discipline or parental reinforcement, warmth or coldness of emotional relationships, and parental involvement with children. Parental supervision refers to the degree of monitoring by parents of the child’s activities, and their degree of watchfulness or vigilance. Of all these child-rearing methods, poor parental supervision is usually the strongest and most replicable predictor of offending. Many studies show that parents who do not know where their children are when they are out, and parents who let their children roam the streets unsupervised from an early age, tend to have delinquent children. For example, in Joan McCord’s classic Cambridge–Somerville Study in Boston,⁽⁹⁾ poor parental supervision in childhood was the best predictor of both violent and property crimes up to age 45.

Parental discipline refers to how parents react to a child’s behaviour. It is clear that harsh or punitive discipline (involving physical punishment) predicts offending. In their follow-up study of nearly 700 Nottingham children, John and Elizabeth Newson⁽¹⁰⁾ found that physical punishment at ages 7 and 11 predicted later convictions; 40 per cent of offenders had been smacked or beaten at age 11, compared with 14 per cent of non-offenders. Erratic or inconsistent discipline also predicts delinquency. This can involve either erratic discipline by one parent, sometimes turning a blind eye to bad behaviour and sometimes punishing it severely, or inconsistency between two parents, with one parent being tolerant or indulgent and the other being harshly punitive.

Cold, rejecting parents tend to have delinquent children, as Joan McCord found in the Cambridge–Somerville Study. More recently, she concluded that parental warmth could act as a protective factor against the effects of physical punishment. Whereas 51 per cent of boys with cold physically punishing mothers were convicted in her study, only 21 per cent of boys with warm physically punishing mothers were convicted, similar to the 23 per cent of boys with warm non-punitive mothers who were convicted. The father’s warmth was also a protective factor against the father’s physical punishment.

Most explanations of the link between child-rearing methods and delinquency focus on attachment or social learning theories. Attachment theory was inspired by the work of John Bowlby, and suggests that children who are not emotionally attached to warm, loving, and law-abiding parents tend to become offenders. Social learning theories suggest that children’s behaviour depends on parental rewards and punishments and on the models of behaviour that parents represent. Children will tend to become offenders if parents do not respond consistently and contingently to their antisocial behaviour and if parents themselves behave in an antisocial manner.

At least in Western industrialized countries, early child-bearing, or teenage pregnancy, predicts many undesirable outcomes for the children, including low school attainment, antisocial school behaviour, substance use, and early sexual intercourse. The children of teenage mothers are also more likely to become offenders. For example, Morash and Rucker⁽¹¹⁾ analysed results from four surveys in the United States and the United Kingdom (including the Cambridge Study) and found that teenage mothers were associated with low income families, welfare support, and absent biological fathers, that they used poor child-rearing methods, and that their children were characterized by low school attainment and delinquency. However, the presence of the biological father mitigated many of these adverse factors and generally seemed to have a protective effect. In the Cambridge Study, teenage mothers who went on to have large numbers of children were especially likely to have convicted children. In the Newcastle Thousand-Family Study⁽¹²⁾ mothers who married as teenagers (a factor strongly related to teenage childbearing) were twice as likely as others to have sons who became offenders by age 32.

There is considerable intergenerational transmission of aggressive and violent behaviour from parents to children, as Maxfield and Widom⁽¹³⁾ found in a retrospective study of over 900 abused children in Indianapolis. Children who were physically abused up to age 11 were significantly likely to become violent offenders in the next 15 years. In the Cambridge–Somerville Study in Boston, Joan McCord found that about half of the abused or neglected boys were convicted for serious crimes, became alcoholics or mentally ill, or died before age 35. In the Rochester Youth Development Study,⁽¹⁴⁾ child maltreatment under age 12 (physical, sexual or emotional abuse or neglect) predicted later self-reported and official offending. Furthermore, these results held up after controlling for gender, race, socio-economic status, and family structure.

Numerous theories have been put forward to explain the link between child abuse and later offending. Timothy Brezina described three of the main ones.⁽¹⁵⁾ Social learning theory suggests that children learn to adopt the abusive behaviour patterns of their parents through imitation, modelling, and reinforcement. Attachment or social bonding theory proposes that child maltreatment results in low attachment to parents and hence to low self-control. Strain theory posits that negative treatment by others generates negative emotions such as anger and frustration, which in turn lead to a desire for revenge and increased aggression. Based on analyses of the Youth in Transition Study, Brezina found limited support for all three theories.

Many studies show that broken homes or disrupted families predict offending. In the Newcastle Thousand-Family Study, marital disruption (divorce or separation) in a boy’s first 5 years doubled his risk of later convictions up to age 32. Similarly, in the Dunedin Study in New Zealand,⁽¹⁶⁾ children who were exposed
to parental discord and many changes of the primary caretaker tended to become antisocial and delinquent. The same study showed that single parent families disproportionally tended to have convicted sons; 28 per cent of violent offenders were from single parent families, compared with 17 per cent of non-violent offenders and 9 per cent of unconvicted boys.

The importance of the cause of the broken home is shown in the UK National Survey of Health and Development.⁽¹⁷⁾ Boys from homes broken by divorce or separation had an increased likelihood of being convicted or officially cautioned up to age 21, in comparison with those from homes broken by death or from unbroken homes. Homes broken while the boy was under age 5 especially predicted offending, whereas homes broken while the boy was between ages 11 and 15 were not particularly criminogenic. Remarriage (which happened more often after divorce or separation than after death) was also associated with an increased risk of offending, suggesting a possible negative effect of step-parents. The meta-analysis by Wells and Rankin⁽¹⁸⁾ also shows that broken homes are more strongly related to delinquency when they are caused by parental separation or divorce rather than by death.