Children and Lone-Mother Families: An Investigation of Factors Influencing Child Well-Being

4. Discussion

The aim of this work was to understand the magnitude of association between lone-mother status on its own with child well-being, and the magnitude of this association adjusting for other sociodemographic and personal factors. Although, on its own, lone-mother status is statistically significantly associated with all of the child outcomes examined, the magnitude of association is weak. The strong statistical significance of the results can be attributed to the large sample size.

The strength of association between lone-mother status and child outcome generally decreases when sociodemographic and/or personal variables are added to the model predicting child outcome. At the same time, the addition of these variables increases the explained variance of the outcome. The most substantial increase in explained variance in the outcome models for social impairment and psychiatric disorder comes with the addition of personal variables. The personal variable hostile parenting has the greatest magnitude of association with social and psychiatric difficulties. For the academic outcome math score, the addition of socio-demographic and personal variables does not substantially increase the explained variance.

Examination of the sociodemographic and personal variables with the greatest magnitude of association with child outcome reveals that hostile parenting is a consistent and significant predictor of child psychiatric and social difficulties. The combination of lone-mother status and hostile parenting is particularly noxious for child psychiatric and social outcomes. No single sociodemographic variable stands as being most strongly associated with child psychiatric, social and academic difficulties. Household income is the sociodemographic variable most consistently associated with child morbidity. The magnitude of association between the individual sociodemographic or personal variables does not change substantially when sociodemographic and personal variables are combined in the model.

4.1 Lone-Mother Status

The finding that lone-mother status on its own is consistently and significantly associated with all child outcomes examined is consistent with other research (Lipman and Offord, 1996; Lipman, Offord and Dooley, 1996). Even so, the size of effect (magnitude of coefficients) and the explained variance of lone-mother status, on its own or combined with other variables, on child outcome is limited. This limited effect size and explained variance are also consistent with previous work examining the effect of similar markers or risk factors on child outcome using a Canadian general population sample (Lipman and Offord, 1996).

Other researchers examining similar outcomes have found larger effect sizes and levels of explained variance for similar predictor variable (e.g., Duncan, Brooks-Gunn and Klebanov, 1994) using an infant population at high risk for developmental delay. Characteristics of the data set used for analysis, such as the characteristics of the sample (e.g., general population vs inner city), age of child at data collection, number of points of data collection (e.g. cross-sectional vs longitudinal), informant and measurement issues influence the magnitude of effect (Lipman and Offord, 1996).

While children from lone-mother families are at increased risk of child difficulties, the limited effect size and variance explained by the family status variable suggests that children from lone-mother families and two-parent families probably develop difficulties for similar reasons. However, several further comments are warranted beyond this simple explanation. First, an exception to the similar factors influencing the development of child morbidity in lone-mother and two-parent families comes with hostile parenting (see Section 4.1.1). Also, since lone-mother status remains significantly independently associated with child outcome in the presence of other variables known to be strongly associated with child morbidity (e.g., income), work needs to be done to explain mechanisms through which lone-mother status might influence child well-being.

4.1.1 Lone-Mother Status and Parenting

Hostile parenting acts as a consistent and significant predictor of child psychiatric and social outcomes. In addition, there is a significant interaction between lone-mother status and hostile parenting such that the presence of hostile parenting is linked with increased psychiatric and social difficulties not seen in two-parent families. This finding is consistent with other work done associating parental stress, parenting difficulties and poor child outcomes (e.g., Patterson, 1986) but reveals an important additional specific interaction (Boyce, Frank, Jensen, et al, 1990). These results suggest that interventions to assist with troublesome parenting practices may be helpful to all families, but that such programs aimed specifically at lone-mother families could have an even more powerful impact on child morbidity. There is evidence that positive parenting practices can have an important influence on child well-being for children in at-risk situations (Landy and Tam, 1996).

4.2 Association Between Income, Maternal Education, Maternal Depression and Child Morbidity

In both lone-mother and two-parent families, low income is significantly associated with all child outcomes examined. Both low maternal education and maternal depression are significantly associated with two of the three measures of child morbidity.

Household income is the sociodemographic variable most consistently associated with child morbidity, and it acts as a significant independent predictor of all child outcomes examined. As is the case for lone-mother status, the effect size is limited, and this is consistent with other Canadian work using a similar population (Lipman and Offord, 1996). As noted above, the characteristics of the data set, including measurement issues and variable specifications influence the magnitude of effect. In these analyses, income was measured using six categories (see Appendix A). In an attempt to tease out possible measurement issues with the specification of the income variable, several additional analyses were completed. The effect size of the income variable does not change substantially when a threshold effect or a non-linear effect is investigated.

While the correlation between the income and parenting variables used in these analyses is small (see Appendix E), this may reflect the cross-sectional "snapshot" nature of the data. Others have demonstrated that low income and its associated stresses can decrease the capacity for poor mothers or fathers to provide optimal parenting that is supportive and consistent due to their level of psychological distress, and this effects child well-being (McLoyd, 1990; Elder, Nguygen and Capsi, 1985; Conger, Ge, Elder et al, 1994; Lempers, Clark-Lempers and Simons, 1989; Patterson, 1986). Interventions to decrease the stresses associated with economic disadvantage could be helpful to children. There is persuasive evidence that programs for socioeconomically disadvantaged children lead to improvements in school performance and cognitive functioning (Lazar, Darlington, Murray et al, 1982).

Maternal education is significantly associated with psychiatric disorder and math score. Increases in maternal educational status could potentially help families in a variety of ways. Interventions to increase maternal education could assist with later maternal employment and family income. Increased maternal education could also positively influence children's academic performance as documented by the association between maternal education and child academic performance demonstrated here and elsewhere (e.g., Lipman, Offord and Boyle, 1995). The mechanisms through which maternal education influences child academic performance could be biological (e.g., mothers who attain higher educational levels are smarter and have smarter children) and/or environmental (e.g., mothers who attain higher education pay more attention to providing stimulating books, toys and activities for their children which enriches a child's capacity to do well in school).

Maternal depression was found to be significantly associated with child social difficulties and child psychiatric difficulties. The finding of increased risk of a range of child difficulties associated with maternal depression has been demonstrated elsewhere (e.g., Downey and Coyne, 1990). The mechanisms through which maternal depression influences child morbidity could be biological (e.g., increased risk of depression in families where a parent is depressed) and/or environmental (e.g., mothers who are depressed are unavailable to interact appropriately with their child which may lead to poor self-esteem and social difficulties for the child). Adequate treatment of maternal mood disorders would be expected to be beneficial both to the mothers themselves and their children.

4.3 Policy Implications

The implications for policy analysts that arise from these analyses suggest that policies aimed at healthy child development should be aimed at all families and not specifically at lone-mother families. Given equal access to programs (e.g., transportation, money for registration fees, etc.), services aimed at alleviating factors consistently and significantly associated with child morbidity should be aimed at the whole population and should be helpful to both lone-mother and two-parent families. The exception to this statement is in the area of parenting, where programs aimed to encourage consistent, supportive and appropriate parenting among lone-mother families could have a specific effect on child morbidity over and above that seen in two-parent families.

The variables that are most consistently associated with child morbidity differ somewhat in quality and quantity for social or psychiatric outcomes and the academic outcome. This finding of distinct correlates for different forms of child morbidity is supported by the literature (Offord, Boyle and Jones, 1987; Rutter, 1985a, 1985b). However, even given these differences, parenting problems, low income, low maternal education and maternal depression represent difficulties associated with child morbidity on a significant and consistent basis, and potential areas of policy impact if sound interventions for these difficulties exist. With the exception of specific parenting programs for lone mothers, income support programs, programs that allow mothers to increase their education and programs aimed at alleviating depression all have the potential to increase child well-being in lone-mother and two-parent families alike. Endeavours to ensure equal access to such programs for lone-mothers and mothers from two-parent families are important.

4.3.1 Difficulties with Policy Planning

It is not possible to provide more specific suggestions for policy planning due to a number of limitations of the NLSCY data available. These difficulties include sorting out temporal relationships, sorting out causal relationships, and other measurement issues.

In terms of sorting out temporal relationships, the data used in this paper are cross-sectional, allowing only a "snapshot" view of the impact of family status, sociodemographic and personal variables on child outcome. Associations between variables such as lone-mother status and child behaviour problems are demonstrated but it is impossible to determine whether lone-parent family status preceded child behaviour problems or not. It is only when a factor associated with a specific outcome precedes an outcome that it can be called a risk factor for the outcome (Kraemer, Kazdin, Offord, et al., 1997). Further, it is only when a risk factor can be changed, and that manipulation of the factor changes the outcome can it be called a causal risk factor (Kraemer, Kazdin, Offord, et al., 1997). It is also important to understand the mechanism(s) through which risk factors acts to influence outcome.

An example of such difficulties follows. Low income, hostile parenting and maternal depression are all significantly associated with the child outcome psychiatric disorder. While the regression analyses suggest that each of the sociodemographic and personal variables are relatively independent of each other (since the standard ß coefficients do not decrease much when all variables are combined in the model), one might imagine that some relationship does exist between each of these three variables. Without understanding the temporal relationship between these variables, whether some variables act as risk factors or causal risk factors on others, and the mechanisms through which these factors have their effect, policy planning is difficult. One is left with multiple possible interpretations of a single result, and a given interpretation may lead to misallocation of resources.

Returning to the example, one possible interpretation of the findings is that low income influences child well-being directly (e.g., through inability to afford an adequate level of consumption) and/or indirectly (e.g., through increasing levels of parental stress leading to maternal depression). The latter relationship between low income and maternal depression is supported by the correlation between variables (see Appendix E). The policy implications arising from this interpretation would suggest that increasing income levels would be helpful. An alternative interpretation of the findings could be that hostile parenting and maternal depression are more strongly influenced by other unmeasured factors (e.g., experiences in one's family of origin such as child maltreatment or how one was parented). If this is the case, interventions other than cash transfers may be more helpful. Yet another interpretation is that income insecurity is due to parental depression since depressed parents may be unable to work. The best intervention in this case may be treatment of parental depression, which would indirectly have the likely effect of increasing income and increasing appropriate parenting behaviour. Unfortunately, available data are inadequate to assess the relative importance of each (Dooley, Curtis, Lipman et al., 1998).

Other measurement issues are important to note, such as using a single point of data collection, the informant used, and the questions included in the NLSCY. A single point of measurement for some variables may not reflect the longitudinal status of the same variable. For example, children's families may move in and out of poverty or children may experience transitions between lone-mother families and families with two adult partners. Other variables not available for these analyses, such as degree of martial conflict, may act as important predictors of child outcome. With respect to informant issues, the outcome measures are largely dependent on maternal ratings (for two of three child outcomes). The same informant provides information on the sociodemographic and personal variables used. None of the measures, except math score, use alternate methods (such as standard tests or observations). No single variable is assessed using multiple methods. Finally, no policy-oriented questions were included in the NLSCY. Availability of such information will be of assistance to policy makers.

4.3.2 Study Strengths

The strengths of this study must also be acknowledged. The NLSCY data base provides a large nationally representative longitudinal sample of Canadian children. While the data for this paper is cross-sectional and derived from the first release of Cycle 1 of the NLSCY, information collected in future cycles of the NLSCY will allow monitoring of more refined poverty and lone parent variables, by including information about long-term vs. transient economic disadvantage and duration of time or child age in lone-parent families. In fact, although Cycle 1 of the NLSCY is a cross-sectional sample, the pending release of the rest of Cycle 1 data includes variables which may provide a historical perspective on the child and family (e.g., custody and access). Increasing refined specifications of family status, and the personal and socio-demographic variables affecting child outcome have the potential to help plan rational targeted intervention and prevention initiatives, and to support healthy child development.

4.4 Summary

Lone-mother status on its own acts as a statistically significant predictor of all child outcomes examined, although the size of effect is limited. This suggests that children from lone-mother families probably develop difficulties for the same reasons that children from two-parent families develop difficulties. Two caveats to this general statement must be acknowledged. First, the presence of hostile parenting in a lone-mother family significantly increases the risk of child morbidity in a way that is not seen for two-parent families, so specific interventions for lone-mother families may be warranted. Second, since lone-mother family status remains a significant independent predictor of child outcome even in the presence of other stronger predictors, work to further understanding of the mechanisms through which lone-mother status might influence child well-being remain important.

The policy implications that arise from these analyses suggest that policies aimed at healthy child development should be aimed at all families, and not specifically at lone-mother families. Given equal access to programs, services aimed at alleviating factors strongly associated with morbidity should be aimed at the whole population and should be helpful to both lone-mother and two-parent families. The exception is in the area of parenting, where specific programs aimed at lone mothers may be warranted.

Low income, parenting problems, low maternal education and maternal depression are consistently and significantly associated with child morbidity, and represent potential areas of policy impact if sound interventions exist. More specific policy suggestions are difficult to justify, given difficulties sorting out the temporal and causal relationships between variables and other measurement issues.

Contents

Children and Lone-Mother Families: An Investigation of Factors Influencing Child Well-Being - October 1998