Modelling the contribution of changes in family life to time trends in adolescent conduct problems

Abstract

The past half-century has seen significant changes in family life, including an increase in parental divorce, increases in the numbers of lone parent and stepfamilies, changes in socioeconomic well being, and a decrease in family size. Evidence also shows substantial time trends in adolescent mental health, including a marked increase in conduct problems over the last 25 years of the 20th Century in the UK. The aim of this study was to examine how these two sets of trends may be related. To illustrate the complexity of the issues involved, we focused on three well-established family risks for conduct problems: family type, income and family size. Three community samples of adolescents from England, Scotland and Wales were compared: 10,348 16-year olds assessed in 1974 as part of the National Child Development Study, 7234 16-year olds assessed in 1986 as part of the British Cohort Study, and 860 15-year olds assessed in the 1999 British Child and Adolescent Mental Health Survey. Parents completed comparable ratings of conduct problems in each survey and provided information on family type, income and size. Findings highlight important variations in both the prevalence of these family variables and their associations with conduct problems over time, underscoring the complex conceptual issues involved in testing causes of trends in mental health.

Introduction

Recent decades have seen substantial changes in family life—including rising divorce rates, changes in socio-economic well being, and reductions in family size. The same period has also been marked by substantial increases in psychosocial disorders of youth, with increases in rates of crime, depression, suicide, and drug use in most Western societies since the second world war (Fombonne, 1998; Maughan, Iervolino, & Collishaw, 2005; Rutter & Smith, 1995). A key question with obvious policy and theoretical implications is whether changes in the family are responsible for the rise in young people's mental health problems.

To address this it is necessary to consider a number of related issues and questions. First, there is a wealth of evidence on risk factors that contribute to individual differences in mental health. Do changes in these same factors also account for overall level differences in mental health? Second, has the correlation among family risk factors changed? Third, does the association of family risk factors with child outcomes change as the prevalence of these factors in the population varies? Fourth, to what extent have reductions in prevalence or risk of some family factors offset any deleterious impact on child outcomes of increases in other risk domains? This study explores these issues by focusing on three well-established family risk indicators: family type, family income, and family size (Hill, 2002). The study examines the extent to which changes in their prevalence, patterning or association with conduct problems account for the documented increase in the prevalence of adolescent conduct problems (Collishaw, Maughan, Goodman, & Pickles, 2004).

The study will consider these issues in the context of trends in the UK over the last quarter of the twentieth century. The scope for generalising from findings about UK trends in adolescent experiences to other countries is an important issue. Many of the social trends discussed here (e.g. increasing rates of parental divorce, decreasing family size) affect not just families in the UK, but families in many countries and societies. Similarly, evidence for a deterioration in adolescent mental health comes from a number of different countries (Maughan et al., 2005; Rutter & Smith, 1995). Nevertheless, most explanations of secular change implicate society-level phenomena, and these may of course be relatively culture specific. There are significant cross-cultural differences in the rates of child and adolescent mental health problems (Crijnen, Achenbach, & Verhulst, 1997), and it would be no surprise if specific cohort differences also varied from country to country (e.g. in terms of timing, duration and extent of historical change). The current study examined the links between changes in family structure and income and trends in adolescent conduct problems in the UK. In doing so we hoped to highlight a number of important conceptual and methodological complexities that are of international relevance to research on trends in mental health.

In the UK, the annual divorce rate increased around seven-fold between 1958 and 2003 (Office for National Statistics, 2006). In the short term, almost all children of divorced and separated families experience a period of living in a single-parent family, while in the longer term many children of divorced parents live in stepfamilies. Population statistics show that rates of single parenthood in the UK increased sharply throughout the 1970s and 1980s – 7% of children lived in single-parent families in 1971, compared with 24% in 2005. In addition, a further 10% of children in 2005 lived in stepfamilies, primarily with their mother and stepfather (Office for National Statistics, 2006). One aim of the present study was to assess the extent to which changes in family type have contributed to trends in adolescent adjustment.

The socio-economic conditions in which teenagers have grown up have also changed. In particular, there has been a substantial increase in UK household disposable income with average income twice as high in 1999, compared with, and adjusted to, 1970 levels (Office for National Statistics, 2004). At the same time, however, there has been an increase in the number of families with children living in relative poverty, with the proportion with incomes below 60% of the median increasing from around 15% in the 1960s and 1970s to around 25% in the early 1990s, and since reducing to around 20% (Office for National Statistics, 2006). It has been suggested that population income inequalities are a more pertinent predictor of mental health problems and criminality than absolute disposable income (Kawachi, Kennedy, & Wilkinson, 1999).

Finally, the second-half of the 20th Century saw a considerable reduction in family size (Office for National Statistics, 2006). The proportion of families in the UK with three or more children declined from 43% in 1972 to 29% in 2005. Large family size has been highlighted as a risk factor for conduct problems both in historical and contemporary studies (Loeber & Farrington, 1998; Rutter, Tizard, & Whitmore, 1970), so it is plausible that there have been benefits for children's adjustment as a result of these reductions in family size.

Alongside a general consensus on the existence of adolescent mental health trends, reviews have also highlighted major methodological complexities in the field (Rutter & Smith, 1995). Changes in the official recording of crime and in the diagnosis and recognition of psychiatric disorders (e.g. ADHD and autism) have made it difficult to provide a clear picture of trends in youth mental disorder and criminality. Only a few studies have examined trends in adolescent emotional and behavioural outcomes using the same instruments at each time point (e.g. Achenbach, Dumenci, & Rescorla, 2003; Collishaw et al., 2004; Verhulst, van der Ende, & Rietbergen, 1997). Collishaw et al. (2004) assessed the extent to which conduct, hyperactive and emotional problems became more common over a 25-year period in three general population samples of UK adolescents. Results showed a substantial increase in adolescent conduct problems that affected males and females, all social classes and all family types. Further analyses using longitudinal data from the first two cohorts showed that long-term psychosocial outcomes for adolescents with conduct problems were equally poor, suggesting that observed trends in conduct problems were not simply the result of changes in reporting thresholds.

Establishing trends in adolescent mental health is complicated by a number of methodological difficulties—most obviously the need to use comparable measures at each time point. Testing explanations for such trends is even more difficult. To date, most people have looked at aggregate trends in disorder and hypothesised explanatory factors. Such comparisons may be helpful, but they are not conclusive. As described above, some family-level risk factors for conduct problems have become more prevalent. In particular, much attention has been paid to the increasing prevalence of parental divorce. Given that there is an association between parental divorce and adolescent adjustment (e.g. Fergusson, Horwood, & Lynskey, 1994; Hill & Maughan, 2001), it is plausible that trends in divorce have contributed to trends in adolescent outcomes. Taken alone, however, the coincidence of changes in adolescent adjustment and in family structure cannot prove a causal connection. In particular, it is important to consider the possibility that as risk factors change in their prevalence, so their associations with child adjustment may also vary. Evidence on parental divorce provides some interesting pointers.

Parental divorce is likely to be stressful and difficult for most parents and children. Nonetheless, commentators have argued that the dramatic increases in the divorce rate in recent decades is likely to mean that the conditions surrounding divorce will have changed and, as a result, that its negative impact on child outcomes may have diminished (Amato & Keith, 1991). As divorce becomes more common, these arguments suggest, associated stigma may be reduced; couples may separate at lower thresholds of marital discord than in the past; and families and society as a whole may be better prepared for managing the negative consequences of divorce.

To date, evidence on changes in associations between divorce and child outcomes is mixed. The limited UK evidence suggests little if any diminution of effects on educational (Ely, Richards, Wadsworth, & Elliott, 1999) or later mental health outcomes (Sigle-Rushton, Hobcraft, & Kiernan, 2005). A meta-analysis of the international literature about effects on children's academic, social and emotional outcomes published in the early 1990s initially supported the view that divorce had become less ‘toxic’ as it became more common, with lower effect sizes reported in studies published in the 1980s than in those conducted in the 1960s and 1970s (Amato & Keith, 1991). However, a recent update to this meta-analysis (Amato, 2001) has shown that effect sizes for studies published in the 1990s were significantly higher than those conducted in the 1980s, though not quite as high as those reported in the earliest studies. The issue is clearly a complex one, but current evidence lends some support to the view that as the prevalence of divorce has changed, its effects may also have moderated.

While a dilution of risk may be associated with some aspects of recent demographic change, other trends may be associated with increasing associations with child psychopathology. In relation to socio-economic disadvantage, for example, an increasing emphasis on individualist and materialist values in recent decades (Eckersley, 2006) may have resulted in the increased impact of socio-economic disadvantage on individuals’ mental well being. Again there is only scant empirical evidence with which to test this hypothesis. However, comparisons of self-reported depression and anxiety among adults in their thirties in the three British birth cohorts indicate the existence of social gradients in affective problems that have become more prominent over time (Ferri, Bynner, & Wadsworth, 2003).

There have been many social changes affecting children's families over recent decades. It is unclear whether and to what extent these changes are associated with trends in child well being. Our study addresses how far trends in three family demographic indicators can account for trends in adolescent conduct problems. Specifically, the study focuses on family type, family income and family size, and parent-rated conduct problems, assessed in comparable ways in nationally representative UK studies conducted in 1974, 1986 and 1999. First, changes in the prevalence and inter-correlation of the family indicators are summarised. Second, analyses test for changes in levels of association between these family indicators and adolescent conduct problems. Third, the study tests the extent to which trends in family life have contributed to trends in conduct problems.