Parental Depression and their Children’s Marriage Timing: The Long-term Consequences of Parental Mental Disorders

Data

To examine the relationship between parental depression and children’s marriage timing, we use data from a comprehensive psychiatric epidemiological survey conducted as part of the CVFS. The CVFS studies whole families and follows all family members across time, including migrants who leave Nepal, which allows us to link parental MDD measures to data on their children. From 2016 to 2018, the CVFS added new clinically validated measures of mental disorders for a general population representative sample of all CVFS family members aged 15–59 (described below). We link these mental disorder measures from parents to 240 months of their children’s CVFS household registry data, which spans from 1997 to 2016. The registry tracks every child born to CVFS members, following the children forward in time, and measuring their marital events. As such, our analytic sample consists of unique mother-father-child triads where parents have participated in the diagnostic mental health survey.

Using the CVFS registry and mental disorder measures, we integrate mothers’ and fathers’ mental health histories with their children’s marriage timing. Also important, we link these parents to their 2008 individual interviews, which include measures of attitudes and marital relationship quality that have been found to shape both children’s marriage timing (Brauner-Otto et al., 2020; Jennings et al., 2012) and mental health in this setting (Zhang & Axinn, 2021). Finally, we also link 2006 household-level measures of income, wealth, and assets to ensure we can assess the influence of parental mental disorders independent of household predictors of children’s motivations to leave the parental home and attractiveness on the marriage market. Our final analytic sample includes 1,069 sons and 982 daughters, the children of 1,004 unique mothers and 998 fathers.

Measures

Analysis

We modelled individual children’s rates of transitioning into marriage. Specifically, we modelled the monthly hazard (h) of marrying, where the hazard at month t is defined as the probability of making the transition within that month, given that a person has not made the transition before month t. Note that instead of analyzing the transition rate (h) directly, we use a logit transformation, log[h/(1-h)], and model this transformed outcome as a function of the predictors. Each individual respondent is identified as being at risk for marriage beginning on the date when she/he reaches age 15, and then is followed until she/he experiences marriage or is censored. Note, beginning these models at age 10 yields exactly the same results.

We analyzed the hazard rate of marriage using multivariable discrete-time hazard models (equivalently known as event history models), with the person-month of exposure to marriage as the unit of analysis. This approach has been used extensively in previous research (Jennings et al., 2012; Zhang & Axinn, 2021). Note that while using person-months of exposure to risk as the unit of analysis in discrete-time models substantially increases the sample size, it does not deflate standard errors (which are clustered at the household level to account for the potential non-independence of sons); thus, the models provide appropriate tests of statistical significance (Allison, 1984; Petersen, 1992). The probability of marriage events is small within each one-month interval, so odds ratio estimates from discrete-time models are similar to hazard ratios obtained using comparable continuous-time models (Petersen, 1992; Yamaguchi, 1987).

In this study, we estimate the relationship between parental MDD (of mothers and fathers considered separately and then jointly) and children’s hazard of marriage using a random-effects model that appropriately accounts for the multilevel nature of the data. Using paternal MDD as an example, we tested a mental disorder predictor coded ‘0’ in all person-months before a father’s onset of first MDD and coded ‘1’ in the first month of the father’s age of onset and then stayed on for all subsequent months (‘father ever had MDD’). Mental disorder predictor variables incorporated a time lag of one year to ensure that first onsets of a mental disorder in the same year as the child’s marriage were not included in the predictor set. The survival coefficients in our tables are interpreted as the odds of transition to marriage for a child with a father who has experienced MDD compared to children whose fathers have not experienced MDD. Successive models include controls for fathers’ other mental disorders as well as mothers’ MDD and other mental disorders.

Descriptive Statistics

Table 1 shows the sample characteristics and distribution of the measures, presented separately for sons and daughters. Around a third (32%) of sons and 43% of daughters got married during the period of observation. The lifetime prevalence of MDD for fathers is 5%, with a lifetime prevalence of any other disorder (predominantly alcohol use disorders) of 16%. For mothers, lifetime prevalence of MDD is 26%, with 17% for any other disorder (predominantly anxiety disorders). These prevalence rates differ slightly from what we observed for the full 2016 CIDI sample, because the latter spans a wider age range and is not restricted to parents (Scott et al., 2021). Measures of other dimensions of the parental family background are consistent with previously published research on Nepal (Brauner-Otto et al., 2020).

Parental Depression and Children’s Marital Timing

Table 2 shows associations between fathers’ and mothers’ depression and the timing of the transition to marriage for sons and daughters. Model 1 shows the association of fathers’ MDD with sons’ marriage timing, controlling only for any other paternal mental disorder, age, and ethnicity. The odds ratio of 1.73 for fathers’ MDD indicates that sons of fathers who have had MDD experience a statistically significant speeding up of the transition into marriage (a 73% higher rate of marriage), compared to sons whose fathers did not suffer from MDD. In Model 2, we add controls for parents’ education, parents’ marital experiences, and household income and assets and find that the association between fathers’ MDD and sons’ marriage timing persists (OR = 1.56). In contrast to paternal MDD, there is no association between maternal MDD and son’s marital timing (models 3 and 4). Models 5–8 of Table 2 show results from the identical analyses of Models 1–4, but for daughters. Consistent with prior CVFS studies of parental social psychological factors predicting children’s marriage timing (Brauner-Otto et al., 2020; Jennings et al., 2012), there is no evidence of an association between either fathers’ or mothers’ lifetime MDD and daughters’ marital timing.

As Table 2 also shows, some of the controls are powerful independent predictors of marriage timing, especially both mothers’ and fathers’ educational attainment, which slows children’s marital timing, as does mothers’ arranged marriage. Ethnicity is also influential, with membership of all ethnic groups relative to the highest caste Brahmin group being associated with faster marital transitions for sons, but a slower rate of marriage for daughters. Our key takeaway from Table 2, however, is that fathers’ MDD is significantly associated with sons’ marital timing independent of other parental and household characteristics.

The models in Table 2 considers fathers and mothers separately, but Table 3 considers them jointly, so that we can observe whether the association between fathers’ MDD and sons marital timing remains once the mental health experience of mothers is also considered. Indeed, this is what we observe, with the association of fathers’ MDD with sons’ marital timing remaining significant, albeit slightly reduced with an OR of 1.53. This association is therefore independent of fathers’ other mental disorders, independent of mothers’ mental disorders, and independent of the full range of other parental and household characteristics included in Table 2.

Limitations

Like all data used to study the social world, these data have important limitations, several of which help to shape the priorities for future research on these topics. First, here we focus on a categorization of mental health – a diagnosis of mental disorder – but as multiple scientists have pointed out, mental health and ill-health exist across a spectrum (Kessler, 2002). Mental disorder diagnostic thresholds define syndromes, the boundaries of which vary as more research and clinical experience accrues, leading to changing definitions over time. Our use of a diagnostic threshold is in some respects a strength, because, by definition, it captures clusters of symptoms that are clinically significant in their duration and impact; however, it represents an average across individuals of a wide range of MDD experiences. By not taking the severity or persistence of MDD episodes into account, the associations we document here likely underestimate the association between the most severe and enduring parental depression and children’s lives. More research that captures the effects of this variation in severity remains a high priority for future studies.

Second, the retrospective assessments we use to measure mental health and the timing of mental disorder onset are known to underestimate occurrence of mental disorders relative to frequent longitudinal measures (Moffitt et al., 2010). To address this limitation, the CVFS used a setting-specific, life history calendar enhanced set of measures of mental disorders to improve retrospective recall. Experimental evidence shows that this approach significantly improves retrospective reporting and produces high levels of clinical validity (Axinn et al., 2020), but it is likely that some recall bias remains.

Third, although we identify multiple possible mechanisms, we do not test them in this initial exploratory study. This will be an important focus for future research to gradually unpack the many possible factors in parents and children’s lives that could contribute to the intergenerational associations we document. In particular, one of the important potential mechanisms linking fathers’ depression to children’s life experiences is children’s own mental health (Dean et al., 2018; S. H. Goodman et al., 2011). Integrating measures of children’s mental health into intergenerational models of family formation will be an important focus for future research. Likewise, additional research on mothers’ MDD, and specific attention to how the timing of maternal MDD intersects with different stages of children’s life course, may yield greater insight into mechanisms linking mothers’ mental health to their children’s transitions to adulthood, even though they are not in evidence here (Banchoff et al., 2022).

Fourth, this research was done in Nepal, which raises the issue of generalizability, yet multiple studies of marriage timing using the CVFS demonstrate associations consistent with European diaspora settings. For example, careful research shows that individual educational enrollment delays marriage while attainment speeds marriage in Nepal (Yabiku, 2005), just as in the U.S. and Europe (Blossfeld & Huinink, 1991; Thornton et al., 2007). The key reason for testing this association in Nepal is the special strengths of the CVFS: the whole-family panel design provides a rare opportunity to investigate how parental mental health prospectively influences children’s life trajectories. Prior studies using these design features from the CVFS also reveal high consistency of intergenerational associations with studies from the European diaspora, but new insights into the independent intergenerational consequences of mothers and fathers (Brauner-Otto et al., 2020; Jennings et al., 2012). The results we document here expand those findings to fathers’ depression, a crucial advance in intergenerational studies. Also relevant, the Nepalese setting is characterized by exceptionally low levels of mental healthcare resources (Jordans et al., 2013). This means that, in contrast to many European diaspora settings, selective access to medical or therapeutic treatments is not likely to bias the results we observe. Nevertheless, more research across a broader range of settings is urgently needed.

Carefully constructed data that reveal new insights into the social world frequently raise more questions than they answer. The work here is a crucial first step toward a new science, with many unanswered questions about the potential breadth and depth of the intergenerational influences of parental mental disorders. The most important result of the research documented here is the motivation for new research investigating the powerful consequences of parental (and perhaps particularly paternal) mental disorders for their children’s life course. The strong associations between fathers’ depression and sons’ marriage timing point to the general importance of this intergenerational influence on children’s transition into adulthood. We study marriage timing because it is of particular significance as a marker for transition into adulthood in Nepal. But in principle this finding demonstrates that parental mental health problems can influence the next generation in ways not previously studied, and later in the child’s life than is usually studied. As such, this understudied dimension of the parental family of origin may exert substantial intergenerational influences on a wide range of adult social life.