عشق پرزحمت: بررسی رابطه میان افسردگی مادران و رفتارهای فرزندپروری

Theoretical perspectives suggest a strong link between maternal mental health and parenting, which may facilitate the intergenerational transmission of disadvantage from depressed mothers to their children. In this paper, I extend prior research by using data from the Fragile Families and Child Wellbeing Study (N = 3659). Pooled OLS and random-effects regression models document a strong link between maternal depression and the following parenting behaviors: neglect, psychological aggression, physical assault, and engagement. Fixed-effects models, however, show little evidence that changes in maternal depression are linked to changes in parenting behaviors, suggesting some negative consequences of depression are driven by variation across individuals or unobserved time-invariant characteristics. Further, the consequences of depression for parenting behaviors do not vary by mother’s marital status, suggesting that marriage may not be protective with respect to the parenting behaviors of depressed mothers.

Introduction Depression is a common mental health condition that affects more than 10% of people in the United States annually, and lifetime prevalence rates are even higher (Kessler and Zhao, 1999). Some subgroups of the population – including the unmarried or poorly educated – are especially susceptible to depression and depressive episodes (Kessler and Zhao, 1999 and Meadows et al., 2008). Women are also more vulnerable to depression than men, and some evidence suggests that mothers with children – compared to their childless counterparts – are especially likely to report depression (Cowan and Cowan, 1992). The consequences of maternal depression are far-reaching, and a burgeoning literature documents how depression among mothers affects individuals and families. Depression and other mental illnesses, for example, impair a single mother’s likelihood of marriage (Teitler and Reichman, 2008). Depressed mothers also report lower quality relationships with their romantic partners and are more likely to report material hardship (Frech and Williams, 2007, Heflin and Iceland, 2009 and Kim and McKenry, 2002). In addition to the wide-ranging consequences that depression may have for the sufferer, a large, robust literature documents that children are particularly vulnerable to maternal depression. Children of depressed mothers, compared to their counterparts never exposed to maternal depression, have worse behavioral outcomes throughout the life course (Goodman and Gotlib, 2002 and Turney, forthcoming). One pathway through which depressed mothers may transmit disadvantages to their children may be through parenting behaviors such as discipline. Indeed, empirical research consistently finds that depressed mothers may be limited in their capacity to parent effectively (Gotlib and Goodman, 1999, Kiernan and Huerta, 2008, Lovejoy et al., 2000 and Marmorstein et al., 2004). Depressed mothers, for example, may be less empathetic, more aggressive, and less emotionally responsive to their children than their non-depressed counterparts (Feng et al., 2007, Lovejoy et al., 2000 and Silberg and Rutter, 2002). In addition, maternal depression is linked to greater negative interactions and fewer positive interactions with children (Cummings and Davies, 1994 and Lovejoy et al., 2000). Given children’s sensitivities to parenting behaviors, and the implications of parenting for children’s life course trajectories, understanding the link between maternal depression and parenting is crucial (Belsky, 1984). Though existing literature documents a strong link between maternal depression and parenting behaviors, there are several opportunities to extend this literature. Many existing conclusions, for example, come from small, non-representative samples with limited generalizability. Often times, samples are limited to clinical populations or non-clinical groups that are homogenous with respect to race, socioeconomic status, or marital status (for an exception, see Kiernan and Huerta, 2008). The fact that little research exists on the consequences of maternal depression for parenting behaviors among unmarried mothers is a particularly stark omission, given the substantial demographic changes of the past five decades (Ellwood and Jencks, 2004). Children born to unmarried parents now account for nearly 40% of all children born in the United States, and researchers are only beginning to examine how this diverse and generally disadvantaged group of children fares (Hamilton et al., 2009). In addition, there are likely important unobserved differences between depressed and non-depressed mothers that are not captured by most existing research. Many prior examinations rely on cross-sectional data or OLS regression models that make it impossible to account for unobserved heterogeneity or discern the causal ordering between maternal depression and parenting. Given theoretical perspectives and empirical research that suggest far-reaching consequences of depression, and some limitations of existing research, in this paper, I examine the link between maternal depression and parenting behaviors. I use data from the Fragile Families and Child Wellbeing Study, a longitudinal survey of nearly 5000 new and mostly unmarried parents in 20 US cities who gave birth between 1998 and 2000. Mothers were first interviewed in the hospital when their children were born, and were re-interviewed when their children were about 1, 3, and 5 years old. Using these data, I examine how maternal depression is linked to a wide range of parenting behaviors: neglect, discipline (including both psychological aggression and physical assault), and engagement. I pay particular attention to differences in this association between mothers married and unmarried at the birth of their child. Though these four indicators of parenting behaviors are not exhaustive of all ways mothers interact with their young children, they capture both positive and negative dimensions of parenting and provide a solid foundation for understanding the consequences of maternal depression for the broader family system (Lovejoy et al., 2000). Furthermore, all of these parenting behaviors have strong, robust associations with wellbeing in early childhood, a crucial period in the life course when children are placed on often static educational and socioemotional trajectories (Entwisle and Alexander, 1989 and Pianta and Cox, 1999).

4. Results 4.1. Bivariate association between maternal depression and parenting behaviors Table 2 presents descriptive statistics for maternal depression and parenting behaviors over time, separately for married and unmarried mothers. Readers should keep in mind that all indicators of parenting behaviors are standardized to have a mean of 0 and a standard deviation of 1. Consistent with prior research, these descriptives demonstrate large, substantively important differences between married and unmarried mothers. At the 1-year survey, for example, about 12% of married mothers and 17% of unmarried mothers report depression (p < 0.001). The greater prevalence of depression among unmarried mothers at the 3- and 5-year surveys is also statistically significant. Additionally, at both the 3- and 5-year surveys, mothers who were unmarried at the focal child’s birth reported more neglectful behaviors (p < 0.001 at the 3-year wave, p < 0.01 at the 5-year wave), psychological aggression (p < 0.001 at both waves), and physical assault (p < 0.001 at both waves). The differences in maternal engagement by marital status at birth are not statistically significant. Table 2. Descriptive statistics of maternal depression and parenting behaviors over time. Variable Married at birth Unmarried at birth 1-year 3-year 5-year N 1-year 3-year 5-year N Mean Mean Mean Mean Mean Mean S.D. S.D. S.D. S.D. S.D. S.D. Depression 0.117 0.152 0.136 916 0.167⁎⁎⁎ 0.228⁎⁎⁎ 0.179⁎⁎ 2743 (0.321) (0.359) (0.343) (0.373) (0.420) (0.384) Neglect – −0.134 −0.096 576 – 0.044⁎⁎⁎ 0.030⁎⁎ 2364 (0.736) (0.829) (1.070) (1.046) Psychological aggression – −0.182 −0.227 576 – 0.060⁎⁎⁎ 0.071⁎⁎⁎ 2364 (0.930) (0.933) (1.015) (1.010) Physical assault – −0.205 −0.197 576 – 0.068⁎⁎⁎ 0.062⁎⁎⁎ 2364 (0.983) (0.952) (0.997) (1.007) Engagement 0.038 0.016 −0.048 915 −0.013 −0.006 0.016 3623 (0.972) (0.951) (1.000) (1.009) (1.016) (1.000) Notes: All parenting behaviors standardized (mean = 0, standard deviation = 1). Symbols compare mothers married at birth to mothers unmarried at birth. ∗p < 0.05. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.001. Table options 4.2. Maternal parenting behaviors as a function of maternal depression The first series of models in Table 3 present the pooled OLS regression models. I only present the coefficient for maternal depression in this table, for the sake of parsimony, but full random- and fixed-effects models can be found in Appendices B and C. The bivariate models demonstrate a strong association between maternal depression and all four parenting behaviors. Depressed mothers are more likely than their non-depressed counterparts to report neglect, psychological aggression, and physical assault. They also report less engagement with their children. These associations are strongly significant across all outcomes (p <0.001). Table 3. Estimating maternal parenting behaviors as a function of maternal depression. Outcome Pooled OLS regression Random-effects Fixed-effects N B S.E. B S.E. B S.E. Neglect Model 1 0.279 (0.045)⁎⁎⁎ 0.252 (0.034)⁎⁎⁎ 0.099 (0.056) 2838 Model 2 0.275 (0.044)⁎⁎⁎ 0.249 (0.034)⁎⁎⁎ 0.101 (0.056) 2838 Model 3 0.179 (0.042)⁎⁎⁎ 0.172 (0.035)⁎⁎⁎ −0.102 (0.057) 2838 Psychological aggression Model 1 0.332 (0.036)⁎⁎⁎ 0.252 (0.033)⁎⁎⁎ 0.027 (0.048) 2838 Model 2 0.286 (0.036)⁎⁎⁎ 0.221 (0.032)⁎⁎⁎ 0.028 (0.049) 2838 Model 3 0.207 (0.036)⁎⁎⁎ 0.160 (0.033)⁎⁎⁎ 0.008 (0.049) 2838 Physical assault Model 1 0.212 (0.036)⁎⁎⁎ 0.172 (0.033)⁎⁎⁎ 0.069 (0.048) 2838 Model 2 0.165 (0.035)⁎⁎⁎ 0.140 (0.032)⁎⁎⁎ 0.068 (0.048) 2838 Model 3 0.098 (0.035)⁎⁎ 0.090 (0.033)⁎⁎ 0.057 (0.049) 2838 Engagement Model 1 −0.133 (0.029)⁎⁎⁎ −0.105 (0.025)⁎⁎⁎ −0.069 (0.029)⁎ 3642 Model 2 −0.143 (0.028)⁎⁎⁎ −0.111 (0.024)⁎⁎ −0.070 (0.029)⁎ 3642 Model 3 −0.065 (0.029)⁎ −0.061 (0.025)⁎ −0.049 (0.030) 3642 Note: Coefficient for maternal depression presented. Model 1 includes no covariates. Model 2 includes the following covariates: race, immigrant status, age, lived with both biological parents at age 15, homeowner, number of children, child is male, child age, child temperament, and child born low birth weight. Model 3 includes all covariates from Model 2 and the following covariates: religiosity, education, employment status, income-to-poverty ratio, relationship status, grandmother in household, fair or poor health, and parenting stress. ⁎ p < 0.05. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.001. Table options These bivariate models, however, do not account for additional factors that may alter the relationship between maternal depression and parenting. When I better isolate this relationship by including a host of individual-level covariates in Model 2, maternal depression is still strongly predictive of parenting behaviors. Across each outcome, the inclusion of the covariates reduces the magnitude of the maternal depression coefficients, but depression is still strongly associated with impairments in parenting. Model 3, the most conservative model that includes additional individual-level covariates, also shows a strong association between maternal depression and parenting behaviors. The magnitude of the association between maternal depression and parenting behavior attenuates across all outcomes, with percentage changes ranging from 28% for psychological aggression to 55% for engagement. However, maternal depression is still associated with nearly one-fifth of a standard deviation increase in neglectful behaviors (p < 0.001). Additionally, when mothers are depressed, they report levels of psychological aggression that is about one-fourth of a standard deviation higher than their non-depressed counterparts (p < 0.001). These full models also show that maternal depression is associated with greater levels of physical assault and less engagement. These estimates are limited because they do not take into account time-varying characteristics of individuals and, thus, may overestimate the link between maternal depression and parenting behaviors. An alternative way to model maternal parenting behaviors as a function of maternal depression is to use random-effects models, which account for both between- and within-person effects. The random-effects models produce results consistent with the pooled OLS models: Depressed mothers engage in less favorable parenting practices with their young children compared to their non-depressed counterparts. Because the findings are slightly smaller in magnitude across all four outcomes, they suggest a weaker association between maternal depression and parenting behaviors. For example, in the full models predicting psychological aggression, the maternal depression coefficient is strongly significant in both the pooled OLS models and the random-effects models, but about one-fourth smaller in the final random-effects model (0.207, compared to 0.160). Lastly, I use fixed-effects models to estimate the link between maternal depression and parenting behaviors. By only taking into account within-person changes in depression and parenting behaviors, these models provide the most conservative estimate of this relationship. The results from the fixed-effects models are strikingly different from the random-effects models in that they show a much weaker – and mostly non-existent – association between maternal depression and parenting behaviors. To begin with, the fixed-effects models show that, even at the bivariate level, there are no discernable differences between reports of neglect, psychological aggression, or physical assault when mothers experience a change in depression between waves. Model 1, however, shows that a change in maternal depression is associated with less engagement. The association persists in Model 2, but disappears when the full set of covariates are taken into account in Model 3. Thus, the link between maternal depression and parenting behaviors may be driven by variation across individuals or unobserved time-invariant characteristics of the mothers. Taken together, these models suggest that maternal depression is associated with less favorable parenting behaviors but that it does not necessarily lead to impairments in parenting. Additional covariates are generally consistent with prior literature (see Appendices B and C). Holding constant a host of individual-level characteristics, the random-effects models show that minority mothers generally report less positive parenting behaviors than their white counterparts. Employment is associated with less physical assault and less engagement, and a greater income-to-poverty ratio is protective against less optimal parenting behaviors. Also, mothers married or cohabiting with the focal child’s father, compared to mothers not in a romantic relationship, report fewer neglectful behaviors. Consistent with expectations, greater numbers of children in the household, having a temperamentally difficult child, and having a male child are linked to parenting difficulties. Parenting stress is one of the most robust predictors of parenting behaviors. Greater parenting stress is associated with less favorable parenting, and this association holds up in all but one of the fixed-effects models. 4.3. Association between maternal depression and parenting behaviors, by marital status at birth By using longitudinal data from a large and recent sample of mothers, prior findings extend much of what we know about the consequences of maternal depression for parenting behaviors. Though it is important to understand the average effects of maternal depression, it is equally important to understand how this relationship may vary among subgroups of the population. Given the voluminous literature that documents the emotional benefits of marriage, as well as other work that demonstrates stark differences between married and unmarried mothers, it is likely that maternal depression is not an equal opportunity risk factor for impaired parenting. Thus, in Table 4, I predict maternal parenting behaviors, paying particular attention to the interaction between maternal depression and marital status at birth. Turning first to the random-effects models shown in Table 4, Model 1 extends the final model from Table 3 to include marital status at birth. The interaction between maternal depression and marital status at birth, presented in Model 2, is statistically insignificant across all four outcomes. Regardless of marital status at birth, maternal depression has similar consequences for maternal parenting behaviors. The models presented in Table 4 include all covariates from Model 3 of Table 3. Some of these variables may be endogenous to maternal depression and, thus, the effects of maternal depression on parenting may be underestimated. However, similar findings persist when I examine the interaction between maternal depression and marital status at birth with a limited set of covariates. Additionally, the fixed-effects models (not presented) show that the relationship between maternal depression and parenting behaviors is similar for mothers married and unmarried at the birth of the focal child. Table 4. Random-effects models estimating association between maternal depression and maternal parenting behaviors by marital status at birth. Neglect Psychological aggression Physical assault Engagement Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Depression 0.171⁎⁎⁎ 0.202⁎⁎⁎ 0.160⁎⁎⁎ 0.201⁎⁎ 0.089⁎⁎ 0.163⁎ −0.062⁎ −0.014 (0.035) (0.077) (0.033) (0.073) (0.033) (0.072) (0.025) (0.054) Unmarried at baseline 0.089⁎ 0.097 0.057 0.066 0.030 0.047 0.110⁎⁎ 0.120 (0.044) (0.050) (0.048) (0.050) (0.047) (0.050) (0.039) (0.041) Depression * unmarried at baseline −0.038 −0.052 −0.092 −0.060 (0.085) (0.081) (0.080) −0.061 Constant −0.815 −0.523 0.012 1.577 N 2838 2838 2838 2838 2838 2838 3642 3642 Note: Models include all covariates from Model 3 of Table 3. ⁎ p < 0.05. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.001.