Poverty and Child Mental Health: Title and subTitle BreakNatural Experiments and Social Causation

Most social scientists have tended to assume that poverty predisposes to mental disorder, on the grounds that numerous studies have shown statistical associations between the two.^{1 - 2} However, as Costello et al³ point out in this issue of THE JOURNAL, causal inference is problematic in the absence of experimental evidence because of the crucial need to differentiate between social selection and social causation.^{4 - 5} In other words, does the statistical association reflect the tendency for individuals with mental disorder to drift into or remain in poverty, or does the experience of poverty itself predispose to mental disorder?

The problem arises when both poverty and the mental disorder apply to the same individuals, as would be the case with studies of adults, but it is especially a concern when poverty derives from the parents, with the effects being considered with respect to mental disorder in the children. Sometimes social selection vs social causation is viewed as a choice between genetic mediation (leading to social selection) and environmental mediation (leading to social causation). However, the issues are much broader than that, particularly when dealing with effects on children. Families in poverty are far from a random selection of the population, and parents who are poor may be so because of either genetic predisposition or environmental adversities or, more usually, a mixture of the two.

It is rare that opportunities arise to put the social selection vs social causation dichotomy under rigorous test. Costello and colleagues³ seized such an opportunity that occurred when a casino was established on an American Indian reservation during the course of their own longitudinal epidemiological study of child psychopathology. The natural experiment derives from the agreement that every adult and child tribe member would receive income from the proceeds of the casino and that this would happen without regard for what the families did. Compared with the 4 years before the casino was opened, the 4 years after the opening were accompanied by a reduction in the number of Indian families with incomes below the federal poverty line, whereas there was no effect on income levels in non-Indian families. The findings from the study showed that the Indian children whose family income was no longer below the poverty line showed a significant reduction in behavioral symptoms of oppositional/defiant and conduct disorder, although there was no effect on anxiety and depression symptoms. The study allows a reasonably strong inference that the effect represented social causation.

It was also important to determine why the relief of poverty had this beneficial effect on behavioral symptoms in the children. Further analysis revealed that improved parental supervision of the children was associated with a reduction in the proportion of single-parent households, an increase in the number of households in which both parents worked, but a decrease in the time demands that were placed on the key parent. Improved parental supervision accounted for about 77% of the effect of changing poverty level on the number of psychiatric symptoms among the children.³ The inference, therefore, is that relief of poverty had benefits for the children when it led to improved parental supervision. In other words, the benefits deriving from relief of poverty were mediated by changes in parenting.

The finding that the effects of poverty, and the benefits associated with its relief, are mediated by the impact on the quality of parenting, and especially on parental supervision, is broadly in keeping with the findings from most nonexperimental studies^{6 - 8} as well as with those from an experimental welfare program in Minnesota.⁹ The implication is that poverty constitutes a distal risk factor (because it makes good parenting more difficult) but that the more direct proximal risk mechanism concerns aspects of parenting. Some caution is needed before concluding that it is parental supervision, as such, that matters. Although that may be the case, Kerr and Stattin¹⁰ suggested that much of the apparent beneficial effect of good parental supervision actually derived from the children letting parents know where they were and what they were doing. Doubtless there are 2-way effects, but that cannot be the whole explanation in the study by Costello et al because the benefits of relief of poverty were associated with a reduction in single-parent households, an increase in both parents working, and gains in the time available to the parents. These findings, however, do suggest that part of the benefit of the casino derived from the increase in the jobs available as well as from the income provided as a direct supplement.

Two further issues arise from the authors' findings. First, it is not clear why the income supplement and the greater availability of jobs enabled only 14% of the Indian families to rise above the federal poverty line. Was this because those who failed to rise out of poverty had greater financial needs (perhaps because of having a greater number of children), because they wasted the money gained, or because they failed to take the job opportunities or were not able to use the increased resources to improve their parenting? The policy implication is that because the benefits of relief of poverty are indirect, it cannot be expected that more money will, in itself, provide the hoped-for benefits. Rather, steps need to be taken to make it more likely that economic gains will be translated into benefits in family functioning.

Second, it is necessary to consider the effects on mental disorders in childhood of the increase in affluence, and the reduction in poverty, that has taken place over the last half century in most, but not all, developed countries.¹¹ If the relief of poverty truly has behavioral benefits for children, it might be expected that the economic gains that have taken place over time would have brought about reductions in children's behavioral disturbance. It is striking that the evidence shows that, to the contrary, the rates of antisocial behavior have increased over the same period that income levels have increased. This contrasts with the evident benefits of higher income levels on improved physical health.¹¹

Of course, time trend data have important limitations. The aggregated data do not allow an accurate determination of the effects of changed economic resources on individual children. It is commonly assumed that inequalities in economic resources, rather than absolute levels, are the key factor with respect to physical health. Nevertheless, this distinction does not seem to help with regard to the lack of evident benefits on behavioral disturbance in young people from economic gains in the population as a whole. Time trend data do not necessarily contradict the findings of Costello et al because the benefits of increased income were indirect and were mediated by changes in family function. The apparent lack of benefits for child mental health from improvements in a population's overall economic situation may mean either that, for some reason, the economic benefits did not translate into improved family functioning or that such benefits were offset by other negative changes in the society as a whole. Moreover, time trend data largely concern levels of crime as reflected in official statistics, rather than clinically significant mental disorder. Conversely, the findings of Costello et al involve self-reported information from parents and children and lack external verification by reports from teachers or police.

Clearly, many questions remain about the ways in which changes in economic level affect family functioning and the ways in which such effects do or do not have benefits for children. In the meantime, it is important to recognize the great value of this carefully assessed natural experiment regarding the behavioral benefits of relief of poverty in American Indian children living on a reservation. The findings go a long way in demonstrating the reality of social causation, and the findings on mediation draw attention to the features of family functioning that particularly merit further attention. Societies need to recognize that economic levels do have important implications for both family functioning and child mental health^{12 - 13} but, equally, policy makers need to ensure that economic benefits actually have the intended psychological benefits. Much remains to be learned about what makes the difference in translating economic gains into psychological improvements for parents and children.