Objectives To investigate the mechanisms through which marital status exerts long-term effects on depressive symptoms among women and to evaluate the relative importance of psychological, social and financial resources in mediating this relationship.
Methods Data came from 6107 female respondents to a nationally representative longitudinal data set from the USA (National Survey of Families and Households 1987–1988/1992–1994). Mediation was investigated using semi-longitudinal structural equation modelling and bias-corrected bootstrapped CIs. Latent constructs with multiple indicators were used to measure depressive symptoms, primary and secondary social integration and self-esteem.
Results The total effect of marital status on subsequent depressive symptoms was statistically significant for all marital statuses relative to those in first marriages controlling for age, education, race, number of children younger than 5 in the household, T1 depressive symptoms and marital status transitions between waves; all groups experienced higher levels of depressive symptoms than those in first marriages. These effects were completely mediated for never-married women and partially mediated for separated/divorced, widowed and cohabiting women. Adjusted household income was the largest mediator for the separated/divorced, widowed and never-married, but primary social integration also played a role. Self-esteem was the only significant mediator for the remarried and cohabiting and was also important in explaining differences between the first-married and separated/divorced.
Conclusion This study demonstrates that the reasons why marital status has an influence on subsequent depressive symptoms varies depending on the specific marital status being compared with the married.
Statistics from Altmetric.com
Individuals in first marriages experience fewer depressive symptoms than the remarried, cohabiting, widowed, divorced and never-married.1 While both married men and women enjoy this mental health advantage, there is evidence that the advantage for women is greater than for men when it comes to depressive symptoms2 and that marital status is a stronger predictor of persistent depressive symptoms for women.3 Given this importance of marital status and the higher prevalence of depressive symptoms in this population,4 a closer look at the mechanisms linking marital status and depressive symptoms among women is warranted.
Many studies have found evidence of a short-term negative influence of marital dissolution on mental health.5 ,6 Strong support for the ‘crisis’ model has contributed to a neglect of the role of marital status in contributing to enduring health inequalities. Both acute and chronic stressors are associated with marital dissolution,7 indicating both temporary and long-term effects.8 ,9 In addition to the negative impact of marital dissolution through the creation of persistent secondary stressors, enduring marital status disparities may also be associated with the benefits of being married. According to the marital resource model, marriage provides access to psychological, social and economic resources, which act as intervening mechanisms in the association between marital status and health.10
This study focuses on self-esteem, primary and secondary social integration and adjusted household income as indirect mechanisms through which marital status influences depressive symptoms (figure 1). Empirical research has demonstrated marital status differences in these resources, highlighting how they are influenced positively by marriage and negatively by marital dissolution.11–15 For example, obtaining and maintaining a marital relationship can have a positive impact on an individual's sense of self-worth (self-esteem), whereas failing to obtain or losing a marital partner could contribute to negative self-appraisals. Empirical research suggests that self-esteem buffers the impact of stress and other situational conditions on mental health16 and higher levels of self-esteem are associated with lower levels of depressive symptoms.12 ,15
Primary and secondary social integration represent interactions with different types of social networks. Family, friend and neighbour (primary) relationships are considered more intimate and regularly maintained than the formal ties that arise from participation in organised groups (secondary).17 Primary integration can provide access to emotional and instrumental support, whereas secondary integration is valued for networking and informational support.17 It is generally assumed that marriage positively impacts social integration; however, marriage may also limit social integration if couples retreat to a more private dyadic social world.18 While both primary and secondary integration have been associated with better mental health,19 the extent to which they mediate the relationship between marital status and depressive symptoms is unclear.
Researchers investigating the impact of marital status on depressive symptoms often control for financial resources and fail to acknowledge its role as a potential mediator. Not only does being married promote greater material well-being as a result of combining economic resources and cost sharing but some evidence suggests that married couples, particularly married men, get paid higher wages than non-married individuals for the same work.20 Controlling for financial resources therefore underestimates the total effect of marital status on depressive symptoms.
Previous research has been successful at partially explaining the relationship between marital status and depressive symptoms by focusing on a single type of resource21 ,22; however, these resources are inter-related, and unless other types of resources are controlled for the unique impact of any particular resource is uncertain. Recent studies have included measures of all three types of resources in an attempt to explain the link between marital status and mental health,12 ,23 ,24 but produce disparate findings, likely due to focusing on different mental health outcomes and different marital status categories. Another difficulty in interpreting their results comes from the cross-sectional analyses. While these studies provide insight into the potential long-term consequences of marital status differences in various resources for mental health, the findings are confounded with short-term processes and selection effects. Evidence of both selection and causation25 requires social selection to be accounted for in order not to overestimate the causal link. In addition, prior research has failed to take advantage of structural equation modelling (SEM) techniques that provide more reliable estimates of mediation than traditional methods.26
The purpose of this paper is to contribute to our understanding of the long-term effects of marital status on depressive symptoms among women by investigating the extent to which psychological, social and financial resources mediate these effects and whether the relative importance of these mechanisms varies across marital statuses.
Data for this study came from the National Survey of Families and Households (NSFH). The initial wave of the survey (1987–1988; T1) involved a national sample of the non-institutionalised population aged 19 years and older, living in the contiguous USA.27 Interviews with 13 007 respondents (74% response rate) were conducted with an oversampling of minorities and households containing single-parent families, step-families, recently married couples and cohabiting couples. The second wave of the NSFH (1992–1994; T2) had an 82% response rate. Of the 7752 women aged 19 years and older interviewed at T1, 6107 were interviewed at both waves and included in this study. Descriptive statistics for the sample are provided in table 1. Additional analysis (not shown) indicated that several variables predicted attrition between T1 and T2; widowed, never-married and cohabiting (compared with the first-married), age, non-white racial status and lower levels of secondary social integration were associated with higher levels of attrition.
All variables were measured at T1 unless noted.
Marital status was categorised as married (first marriage), remarried, separated/divorced, widowed, never-married and cohabiting. First marriage (first-married) was the reference category.
Three indicators from Rosenberg's self-esteem scale28 were used as indicators of a latent variable measuring self-esteem. Respondents were asked to indicate how much they agree or disagree with the following statements: On the whole I am satisfied with myself; I am able to do things as well as other people; I feel that I'm a person of worth, at least on an equal plane with others. Responses in the survey were coded from 1 (strongly agree) to 5 (strongly disagree) and were reverse coded. Higher scores denote higher levels of self-esteem.
Primary social integration
Respondents were asked how often they did the following: get together socially with relatives, get together socially with a neighbour, get together with friends who live outside the neighbourhood. Responses included never, several times a year, about once a month, about once a week and several times a week. A latent measure was created using all three indicators.
Secondary social integration
Three indicators were used for this latent variable. The first indicator represented participation in service organisations, fraternal groups or political groups; the second represented participation in work organisations, such as unions or professional societies; and the final indicator represented participation in sports or hobby groups. Frequency of participation in each set of activities was coded as never, several times a year, about once a month, about once a week or several times a week, based on the activity within each set that was most frequently participated in.
Adjusted household income (logged)
Details on different sources of income were collected from the respondent, spouses/partners and tertiary respondents. To calculate adjusted household income, a composite measure was computed that comprised all income for each member of the household. This total was then divided by the square root of the number of people in the household. This measure takes into account economies of scale.29 A constant of one was added and the natural log was used in all analyses due to the skewed distribution.30
Respondent's age was included as a continuous variable. Education indicated the number of years of completed schooling. Three racial/ethnic groups were considered: non-Hispanic white, black and other. Number of preschool children in the home was a count of the number of children under 5 years of age. Transitions to and from a partnership between waves were included as separate dummy variables with a value of 1 if a transition occurred between waves.
T1 and T2 depressive symptoms were modelled as latent variables, each with 12 indicators from the Center for Epidemiological Studies Depression Scale (CES-D). The CES-D has high construct validity and internal consistency.31 At each wave, respondents were asked how many days in the past week: “You were bothered by things that usually don't bother you?”; “You felt lonely?”; “You felt that you could not shake off the blues, even with help from your family or friends?”; “Your sleep was restless?”; “You felt depressed?”; “You felt that everything you did was an effort?”; “You felt fearful?”; “You had trouble keeping your mind on what you were doing?”; “You talked less than usual?”; “You did not feel like eating, your appetite was poor?”; “You felt sad?”; “You could not get going?”
Method of analysis
Mplus 6.11 was used to conduct SEM. A simplified diagram of the structural model is provided in figure 2. SEM provides more reliable estimates of mediation than traditional path analysis because the use of latent variables can overcome the biasing effects of measurement error, and all equations are estimated simultaneously.26 Bias-corrected bootstrap CIs from 1000 resamples were used as this technique provides the most accurate estimates of mediation by allowing for the non-normal distribution of indirect effects.32 Weighted least square parameter estimates using a diagonal weight matrix with SEs and mean-adjusted χ2 test statistic that use a full weight matrix, and THETA parameterisation were used. Missing data were handled using full information maximum likelihood.33
The measurement model (CFI 0.993, TLI 0.992, RMSEA 0.019) and the structural model (CFI 0.981, TLI 0.978, RMSEA 0.024) displayed good model fit.34 Each arrow in figure 2 represents a parameter or a set of parameters that were estimated and reported in table 2. These parameter estimates can be interpreted as regular ordinary least squares coefficients.
The direct, indirect and total effects of marital status on subsequent depressive symptoms, net of control variables, were calculated by Mplus using the IND command and are presented in table 3. The total effect of marital status on later depressive symptoms was statistically significant for all marital statuses relative to those in first marriages (p<0.01; never-married p<0.05); each of these groups experienced higher levels of depressive symptoms than those in first marriages. Complete mediation was observed for the never-married (indirect and total effects were significant but direct effect was not). Partial mediation was observed for separated/divorced (p<0.01), widowed (p<0.01) and cohabiting (p<0.05) women, as the overall indirect effects were significant but a significant direct effect remained (p<0.01). Indirect effects accounted for 21.7% of the total effect of being separated/divorced, 35.8% of the total effect of being widowed and 15.5% of the total effect of cohabiting compared with being first-married. Overall, indirect effects were not significant in accounting for the higher levels of depressive symptoms among remarried women.
The effect of marital status through each individual pathway is provided in table 3 under the specific indirect paths. The parameter estimates for the indirect effects were calculated by multiplying the direct effect of a marital status (relative to first-married) on a mediator by the direct effect of the mediator on depressive symptoms.26 Overall, self-esteem was a significant pathway through which marital status influenced T2 depressive symptoms for remarried (p<0.05), separated/divorced (p<0.05) and cohabiting women (p<0.01). For cohabiting women, self-esteem was the largest indirect pathway. Separated/divorced (p<0.01), widowed (p<0.05) and never-married (p<0.05) women experienced higher levels of depressive symptoms at T2 than first-married women because they had higher levels of primary integration; this pathway was the second largest in magnitude for these three groups. Secondary integration was not a significant mediator, although it reached marginal significance (p<0.10) for the separated/divorced, widowed and cohabiting. Adjusted household income was a significant mediator for separated/divorced, widowed and never-married women (p<0.05); this pathway was the largest in magnitude for these three groups. The magnitude of the coefficients for the specific indirect effects are comparable despite the different metrics of the mediating variables as they represent the difference in the latent measure of depressive symptoms between those in first marriages and each marital status category that can be attributed to each indirect pathway.
While the results in table 3 tell us whether or not the mediating effects are significantly different than zero, they do not tell us if the magnitude of specific indirect effects are significantly different from each other. Additional analyses (not shown) comparing pathways within each marital status found that the magnitude of the mediating effect of adjusted household income was significantly different than the mediating effect of self-esteem for the remarried, never-married and cohabiting (p<0.05) and was marginally significant for the separated/divorced and widowed. The magnitude of the mediating impact of self-esteem was significantly different than the mediating impact of primary integration for the remarried (p<0.01), never-married (p<0.05) and cohabiting (p<0.05). Among the never-married, the magnitude of the mediating effects of adjusted household income (p<0.01) and primary integration (p<0.05) were greater than secondary integration.
Findings from this study demonstrate that marital status has long-term implications for levels of depressive symptoms among women and enhance our understanding of the relative importance of self-esteem, primary and secondary social integration and adjusted household income in explaining these differences. Even after controlling for age, education, race, preschool children in the household, baseline depressive symptoms and transitions between waves, marital status is associated with levels of depressive symptoms approximately 6 years later. Collectively, the psychological, social and financial resources captured in this study completely mediated the relationship between marital status and T2 depressive symptoms among never-married women relative to their first-married counterparts and partially mediated the differences between the first-married and the separated/divorced, widowed and cohabiting. Studies that report marital status differences in depressive symptoms after controlling for these mediators underestimate the total effect of marital status on mental health by ignoring the indirect effects of marital status through resources.
Self-esteem did not factor into later differences in depressive symptoms between the first-married and the never-married or widowed. However, the mental health disadvantage of the remarried and separated/divorced via self-esteem suggests a damaging effect of becoming separated/divorced on self-esteem that is not completely repaired by re-partnering. The disadvantage of cohabiters attributable to self-esteem may also reflect the negative impact of marital dissolution, as many were previously married, although it could also be related to the marginal status of cohabiting unions in the USA. Future research should directly measure individual perceptions of social approval or stigma associated with marital status24 in order to more clearly articulate how marital status influences self-esteem and subsequent mental health.
Part of the disadvantage in T2 depressive symptoms observed among separated/divorced, widowed and never-married women relative to those in first marriages was a result of having higher levels of primary integration at T1. The higher levels of primary integration among the non-married/non-cohabiting suggest a privatising effect of marriage, where couples restrict their social contact with other family, friends and neighbours.18 The finding that higher levels of primary integration were associated with higher levels of depressive symptoms is contrary to expectations, although not without precedent. For example, higher levels of perceived social support have been associated with subsequent declines in mental health.21 While primary integration can be a source of emotional support, it can also be an emotional burden.17 This finding challenges the pervasive tendency to focus on the positive aspects of social ties and calls for greater attention to the qualitative aspects of social relationships that may be harmful for mental health.
Although only significant at the trend level, secondary integration contributes to the mediation of differences in subsequent depressive symptoms between the first-married and the separated/divorced, widowed and cohabiting. Consistent with previous research,13 the remarried and never-married did not significantly differ from the first-married on secondary integration, suggesting that while a first marriage may not confer a secondary integration advantage over the never-married, marital dissolution and cohabiting are associated with a secondary integration disadvantage that influences subsequent depressive symptoms. More research is needed to identify why the cohabiting and previously married have lower levels of secondary integration.
Despite previous studies that have found remarried and cohabiting individuals to be at an economic disadvantage relative to the first-married,14 ,35 this study did not find that financial resources mediated subsequent differences in depressive symptoms for these groups. However, adjusted household income was the largest mediator of later differences in depressive symptoms between the first-married and the separated/divorced, widowed and never-married. These findings demonstrate the importance of sharing a household with a significant other for providing access to financial resources that have an enduring effect on women's mental health.
The resources captured in this study failed to account for the relative disadvantage of remarried women compared with those in first marriages and left a significant portion of the difference between the first-married and the separated/divorced, widowed and cohabiting unexplained. Part of these unexplained differences could be related to the presence/quality of a confidant, which is confounded with marital status in this study. Although friends and relatives can also function as confidants, it is argued that the intimacy of the marital relationship promotes a unique form of social support that is of higher quality and therefore more powerful than support provided by others.36 ,37 Compared with those in first marriages, remarried and cohabiting individuals may experience lower levels of relationship quality and subsequent mental health than those in first marriages as a result of step parenting, role ambiguity or perceptions about the long-term stability of the relationship.24 ,38 Relationship quality in general deserves closer scrutiny as not all marriages are healthy, and those of poor quality can negatively impact mental health. The results in this study demonstrate the average impact of marital status on depressive symptoms, but the magnitude and direction of these relationships may vary by marital quality and require further examination.
Further comprehensive research needs to be done to determine what other factors mediate the lasting influence of marital status on depressive symptoms for women. Personal mastery may be important, as well as religious participation and perceived social support.21–24 Preliminary analyses (not shown) did not find religious participation or perceived social support (from people other than a spouse/partner) to be significant mediators and were dropped from the model. However, future research should give these factors more consideration, as the variables available in the NSFH may not have been operationalised in a way that captures their potential mediating effect.
Given the complex direct and indirect pathways through which marital status influences depressive symptoms demonstrated in this study, policies and programmes that promote healthy marriages could be considered as a strategy to improve public mental health. The US government-sponsored Healthy Marriage Initiative provides funding for programmes supporting couples in forming and sustaining healthy marriages, and research confirms the effectiveness of these programmes in diverse populations.39 In addition, removing disincentives for marriage embedded in policies and programmes, such as assistance programmes that provide benefits on the basis of the family income or a marriage penalty in the tax code, could reduce barriers to marriage, particularly for low-income couples.40
Findings from this study should be interpreted with certain caveats in mind. While controlling for baseline depressive symptoms reduced the likelihood of biasing the influence of marital status on subsequent depressive symptoms upwards due to selection effects, it is likely that this control variable contributed to an underestimation of these relationships. This is because the T1 measure of depressive symptoms was recorded after marital status transitions occurred and its effect includes selection effects as well as any causation effects that occurred prior to T1. Furthermore, this study does not control for selection into or out of marriage based on any of the mediating mechanisms. This weakens the causal argument that can be made about the impact of marital status on the intervening mechanisms. However, regardless of selection into marriage based on these factors, this study demonstrates the extent to which self-esteem, primary and secondary social integration and adjusted household income account for inequalities in depressive symptoms across marital statuses and provide a direction for targeted interventions to reduce these inequalities.
Another noteworthy limitation is the age of the data. To my knowledge, this is the most recent nationally representative longitudinal data set in the USA that includes detailed measures of marital status and household income, as well as multiple indicators of depressive symptoms, social integration and self-esteem. This demonstrates a strong need for more comprehensive contemporary data collection. Historical trends in the relationship between marital status and self-rated health suggest that marital status inequalities in health have increased since the data used in this study were collected.41 The degree to which these increasing marital status disparities can be tied to changes in the distribution of resources across marital statuses is an empirical question that remains to be answered. Lastly, while the pathways through which marital status influences depressive symptoms are likely similar across nations, differing levels of inequality, marriage patterns, social policies and social norms related to marriage in other countries may produce unique results that should be explored more fully.
What is already known on this subject
The mental health advantage of the married over other marital statuses is well established in the literature.
Most research examining how marital status influences mental health focuses on the short-term ‘crisis’ effects of marital dissolution.
Less is known about the long-term contributions of marital status to mental health inequalities and the mechanisms through which marital status has an enduring effect.
What this study adds
This study demonstrates the unique effects of self-esteem, primary and secondary social integration and adjusted household income in mediating the relationship between marital status and subsequent depressive symptoms using state of the art mediation analysis techniques.
Findings from this study allow for the development of targeted interventions to reduce marital status inequalities in depressive symptoms among women, as mediation pathways varied depending on the specific marital status being compared with those in first marriages.
Competing interests None.
Ethics approval This study was exempt from institutional review board approval because it is based on secondary data analysis of publically available data.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.