Abstract
In this study we explore the gap between preferred and actual work situations for mothers of young children. We further examine the effects of not meeting desired work situation preferences on mothers’ depression and parenting stress over time. Using data from the National Institute of Child Health and Human Development’s Study of Early Child Care (1141 families) our findings indicate up to 71 percent of mothers in this sample did not achieve their preference. After controlling for marital status, ethnicity, education, income-to-needs, perceived costs of employment, and social support, both intra-individual and inter-individual assessments of the gap between a mother’s ideal and actual employment significantly predicted depression across the first three years of parenthood. Mothers’ employment hours and preferences for employment predicted parenting stress, but social support buffered this effect. We tested preliminary distinctions between maternal under-employment and over-employment and discuss implications for home and work.
Recent studies suggest that many adults, particularly women, are not in their desired work situations (employed full time, employed part time, or not employed) (Drago et al., 2009; Reynolds, 2003). For example, results from a US nationally representative sample of mothers indicated that only 36 percent were in their preferred employment circumstance (Jacob, 2008). From a ‘person-environment fit’ perspective, the prevalence of employment mismatches may have negative implications for maternal wellbeing and employers who attempt to retain talented mothers with young children (Spokane et al., 2000).
Further research is needed to better understand the realities of a maternal employment gap and its implications for maternal wellbeing at home and at work. The current study draws on data from the Eunice Kennedy Shriver NICHD Study of Early Child Care (SECC) to explore the extent and implications of a gap between desired and actual work situations for mothers of young children. In doing so, the study attempts to clarify findings from competing hypotheses regarding the beneficial or harmful effects of maternal employment on maternal wellbeing. The current study builds on efforts to sort out these contradictory assumptions by focusing on a mother’s own perception of her work and family roles during the early period of parenting (Barnett, 2004).
Potentially significant implications for mothers, children, and employers underscore the importance of exploring the relationship between work and family decisions and maternal wellbeing during this period. Transformations in personal identities as well as shifts in the priorities attributed to employment during the early stages of motherhood highlight the unique importance of this period for mothers (Hilfinger Messias and DeJoseph, 2007). Further, maternal psychological wellbeing during the early years of a child’s life has been linked to a host of significant child outcomes including language, cognitive, and social-emotional difficulties beginning in infancy (see Murray and Cooper, 2003). In addition, depressed maternal affect may negatively influence productivity on the job or contribute to attrition of key talent away from the organization and into the home. When mothers of young children choose to leave the organization there are significant costs to employers as they need to attract, recruit and train replacements (Hill and Weiner, 2003).
Conceptual framework
This study is guided by social cognitive theories of maternal wellbeing that link a mother’s perceived ability to enact desired choices with her psychological wellbeing (Bandura, 2001). Social Cognitive Career Theory (SSCT) (Lent et al., 1994) identifies self-efficacy as a key factor in determining a person’s willingness to initiate career behaviors as well as one’s emotional reactions to perceived barriers and conflicts. Personal beliefs about the inability to control one’s environment and life circumstances have been linked with the likelihood of depression and parenting stress (Bandura, 2001). SCCT also recognizes the influence of gender-based socialization experiences on the formation of self-efficacy beliefs, which in turn influence career preferences and decisions. The dual expectation for women to pursue a career and invest in family with its associated heavy demands and acute responsibility is likely to influence efficacy, and in turn, career-related decisions (Cinamon, 2006).
Attainment of desired work situations: Review of related research
Research exploring mothers’ work situation preferences suggests that attaining desired work situations may be difficult, particularly during the early period of parenting. Mothers’ work and family preferences generally reflect polarization into three distinct groups: those giving priority to marriage and child-rearing as their central life activity, those giving priority to market work or equivalent activities, and those characterized by women who want to combine work and family, many of whom enter and exit the workforce, or have unplanned careers (Hakim, 2000). These preferences often become more complex as mothers make career adjustments to meet family goals. Mainiero and Sullivan’s (2005) Kaleidoscope model illustrates how women alter career patterns by ‘rotating different aspects in their lives to arrange their roles and relationships’ (p. 111). While women tended to factor in the needs of children, spouses, aging parents, friends, and even coworkers when making career decisions, men tended to keep work and non-work lives separate, orienting career decisions to individual goals.
This relational orientation towards career suggests complexity in how women experience and attain work preferences. Women’s employment preferences and trajectories reflect an ongoing decision to work for pay that is ‘responsive to highly particularistic, complex, and unfolding circumstances’ (Hynes and Clarkberg, 2005: 225). Events that alter life stage, such as the birth of a child, trigger changes in a woman’s orientation towards work and family roles (Hostetler et al., 2007). Yet, pay, benefit, retirement systems, and work hour arrangements were built to accommodate a male breadwinner role that did not require such flexibility (Drago et al., 2009). The gap between the perceived need for open-ended career adjustments and the institutional changes needed to accommodate such flexibility suggest the likelihood of a gap between preferred and actual work situations for women.
Drago et al.’s (2009) analysis indicated that the arrival of a new child was one of three key predictors of a mismatch between preferred and actual work hours. Stone’s (2007) earlier ethnographic work found that many women would have preferred reduced-hour arrangements after the birth of their child, but they opted out of the labor market because reduced-hour arrangements were either unavailable or involved poor pay or working conditions.
Attaining work situation preferences is further complicated by unexpected challenges. Marital instability (Cherlin, 2009), single parenthood (Mischel et al., 1999), job instability and declining real wages led to increased dependence on women’s contributions to family incomes (Mischel, et al., 1999; US Department of Labor, 2005). The recessions of 2001 and 2008 contributed to sustained periods of job loss, with more dramatic losses for men and increased reliance on women’s incomes (Hartmann, 2008; US Department of Labor, 2005).
Even in better economic times, women experience strong pressures at both macro and micro levels to conform to societal expectations for the ‘good mother’ or ‘good worker’ role (Blair-Loy, 2003). Both ideologies suggest undivided loyalty to attain these ideals, increasing the difficulty of reaching high expectations in both roles. The result of these opposing logics is a context of dilemmas in the emotional and practical experience of work and family life (Moen and Roehling, 2004). Women may feel they are left in an untenable position. In the words of Mary Still (2006): ‘The choice is between two undesirable options: quit work to become an “intensive” mother, or become an ideal worker who is regarded as cold and a bad mother’ (p. 168).
Implications of a maternal employment gap on maternal wellbeing
Previous findings suggest that the difficulty mothers face as they attempt to enact choices around work and family goals may negatively influence psychological wellbeing. For mothers of young children, maternal employment may touch the most intimate maternal concerns – the development of children, and the pursuit of individual and family interests and needs. Evaluations of the relationship between length of maternity leave and emotional wellbeing found that after a child’s first year, depression was greatest among mothers who preferred full-time employment but had long maternity leaves (Klein et al., 1998). On the other hand, working mothers who felt they should be at home experienced maternal guilt, and these perceptions of themselves as ‘bad mothers’ contributed to poorer wellbeing (Elvin-Nowak, 1999; see also Judge et al., 2006). Although societal norms may have supported the caregiving focus for each group of mothers, the mothers themselves seemed to identify a different choice than they could attain. When the desired choice did not match their employment circumstances, both sets of mothers experienced poor psychological wellbeing.
Mistry et al.’s (2008) recent expansion of the family economic stress model drawing on data from 516 low-income families found that mothers commonly described feeling stressed when their income was ‘inadequate to meet their needed and desired expenditures’ (p. 199). Inability to enact choices that they believed important to their families, whether or not they were essential to survival, contributed to maternal stress and negatively impacted parenting. Dooley et al. (1996) concluded that being unemployed when one desires to be employed was the most consistent predictor of lower emotional wellbeing for mothers in poverty. The current economic recession may have increased the number of mothers who find themselves in this situation. Nearly 75 percent of those who have lost jobs have been men, requiring many mothers to become major breadwinners (Lovell et al., 2008).
Similarly, mothers who perceive their employment to have negative consequences for child development report greater role strain (Nomaguchi et al., 2005). Mothers of young children are likely to be more vulnerable to role strain between work and family because of caregiving demands (Noor, 2004). Discrepancies between what mothers think they should do for their child, and what they are actually able to do, may also contribute to role strain. Employed mothers who reported more positive outcomes from maternal employment reported less parenting stress and role strain than employed mothers who believed maternal employment might be detrimental for children’s development (Chang and Huston, 2001).
Implications of maternal wellbeing on workplace effectiveness
Research on the relationship between emotional wellbeing and workplace effectiveness suggest that maternal depression and stress could have negative implications for the workplace (Martin, 2010). The economic costs and management challenges associated with depression have led to its label as ‘the occupational disability of the new millennium’ (Raderstorf, 2001: 57). Stress and depression have been linked to increasing absenteeism and early retirement across the European Union (McDaid et al., 2005). Additional findings using focus groups with British workers found a strong link between depression and impaired work performance and decreased workplace safety (Haslam et al., 2005). Employers aware of gaps between societal norms and a mother’s desires may benefit from research findings that employers who offer workplace flexibility in the form of gradual return to full-time work hours are likely to retain these mothers in productive work (Hill et al., 2004).
Research questions and hypotheses
To enhance the current literature, we respond to two basic research questions: What is the likelihood of attaining work situation preferences for mothers during the first 36 months of parenthood? How does not attaining her work situation preference influence a mother’s subsequent depression and parenting stress levels?
Guided by social cognitive theories we hypothesize that mothers whose actual and ideal work situations are incongruent across the first 36 months of their child’s life are likely to experience higher levels of maternal depression and stress than mothers whose actual and ideal situations are congruent. We expand the current literature by studying the effects of maternal employment gaps on wellbeing after considering the effects of the strongest commonly identified predictors of maternal psychological wellbeing including race, educational level, marital status, number of children, income-to-needs ratio, attitudes towards maternal employment, and perceived availability of social support.
We further emphasize the dynamic nature of women’s employment. To our knowledge, prior research has not explored the potentially dynamic nature of gaps between a mother’s reported ideal and actual employment over time. Yet, one would expect variations in individual employment experiences over time, particularly during a critical transition period. Therefore, we employed multi-level growth curve models to assess both between-persons (i.e. inter-individual) and within-persons (i.e. intra-individual) change in a mother’s actual employment experiences, her employment preferences, her perception of available social support, and the gap between her actual and ideal employment over time. Given that no prior work distinguishing between-persons from within-persons change in maternal employment has been conducted, we suggest a preliminary hypothesis that the greater between-persons and within-persons variability between her ideal and actual employment experiences, the greater the threat to a mother’s wellbeing.
Finally, to understand the association between incongruence in actual and ideal work situations, we conduct preliminary analyses to distinguish between mothers who are over-employed (e.g. mothers who prefer to stay home, but work full time) and mothers who are under-employed (e.g. mothers who prefer to work full time, but stay home) to see which context may place mothers at greatest risk. Findings regarding this research question are mixed (Morris and Levine Coley, 2004). Based on role enhancement theories, we hypothesize that mothers who are under-employed will be at increased risk of depression and parenting stress. Based on role strain theories, we hypothesize that mothers who are over-employed will be at increased risk of depression and parenting stress.
Methods
Sample
The sample for this study came from the Study of Early Child Care, a comprehensive longitudinal study funded by The Eunice Kennedy Shriver National Institute of Child Health and Human Development, representing 10 geographic sites across the USA. The recruited 1364 healthy infants and their families were demographically similar to the families in the catchment areas (Early Child Care Research Network [ECCRN], 2001). The sample was not nationally representative, but participants represent a range of socioeconomic and sociocultural backgrounds. White, non-Hispanic children are somewhat over-represented, and children from ethnic minority groups are somewhat under-represented. Mean household income and maternal education are higher than the US average, but sample families were also more likely to receive public assistance than US families in general (ECCRN, 2001). The current analyses included 21 percent ethnic minority, 10 percent without a high school education, and 14 percent single mothers from the 1141 with complete data across the three-year span studied.
Measures
Control variables
At child ages 1, 6, 15, 24, and 36 months, interviewers visited families in their homes to collect data on demographic information such as the child’s ethnicity (dummy codes for African American and Other ethnicities are reported, with White, non-Hispanic as the reference group), maternal education (highest level of education completed, centered around high school graduation), number of children in the home, mother’s marital status (dummy codes for married, cohabiting, or single are reported, with ‘married’ as the reference group), and the family’s income-to-needs ratio (calculated by dividing the total family income by the poverty threshold based on family size). During these interviews, mothers also answered questions regarding ideal employment status, actual employment, perceived costs of maternal employment, social support, depressive symptoms, and parenting stress.
Employment variables
Actual maternal work/school hours per week were reported on a continuous scale at the 1, 6, 15, 24, and 36 month assessments. To report on ideal employment, mothers responded to the question ‘If you could have your ideal [employment] situation, what would it be right now?’ at the 1, 6, 15, 24, and 36 month assessments. Response options included: 1) work full time or combine work and school full time; 2) work part time; 3) go to school; 4) be at home full time or on full-time leave.
To create the employment match status variable, we converted maternal reports of continuous employment and school hours into mutually exclusive categories consistent with the four ideal employment categories (work full time/combine work and school full time, work part time, etc.). Mothers were identified as a ‘high match’ when they reported a perfect match between their ideal employment/school status, and their actual employment/school status. For example, a mother whose ideal circumstance was to work full time who reported working full time was given a high match status. In contrast, a mother who reported a complete mismatch between her ideal and actual circumstance was given a low match status. For example, a mother whose ideal circumstance was to work full time, but reported that she stayed home full time received a low match status; or a mother whose ideal circumstance was to stay home full time but reported that she worked full time received a low match status. All other mothers not in a clearly defined high match or low match status were categorized as ‘medium match.’ For example, a mother whose ideal circumstance was to work full time, but reported that she worked part time, or a mother whose ideal circumstance was to work part time, but reported that she stayed home full time, received a medium match status.
Perceived costs of maternal employment
The Attitude Toward Maternal Employment Questionnaire (Greenberger et al., 1988) assessed at the child’s one month interview was composed of six items including: ‘Children are less likely to form a warm and secure relationship with a mother who is working full time,’ ‘Children whose mothers work are more likely to be left alone and exposed to dangerous situations,’ and ‘Working mothers are more likely to have children with psychological problems than mothers who do not work outside the home.’ Mothers’ responses ranged from 1 = disagree very strongly to 6 = agree very strongly. Higher scores indicated a stronger belief that maternal employment is harmful to children. Attitudes toward maternal employment were only assessed once during the study. Cronbach’s alpha was .89. The total score is significantly associated with gender, education, women’s employment status, and sex role traditionalism establishing both concurrent and predictive criterion validity.
Social support was measured at 1, 6, 15, 24, and 36 months using the Relationships with Other People Scale (Marshall and Barnett, 1993), an 11-item measure on which mothers rate their perceived availability of support over the past month. Each item is scored from 1 = none of the time, to 6 = all of the time. Sample items include: ‘The people I care about make me feel they care about me,’ ‘The people I care about help me out,’ and ‘The people important to me accept me as I am.’ Higher scores represent greater perceived availability of support. Cronbach’s alpha was .91 at baseline. This measure has been significantly correlated with maternal mental and physical health symptoms (Marshall and Barnett, 1993), establishing predictive criterion validity.
Maternal depressive symptoms were measured at 1, 6, 15, 24, and 36 months using the Center for Epidemiological Studies Depression Scale (CES-D; Radloff, 1977). Respondents rate the frequency of 20 symptoms during the past week. Sample items include: ‘I was bothered by things that usually don’t bother me’; ‘I felt sad’; ‘I felt lonely’; ‘My sleep was restless.’ Response categories are ‘rarely or none of the time (less than 1 day),’ ‘some or a little of the time (1–2 days),’ ‘occasionally or a moderate amount of time (3–4 days),’ and ‘most or all of the time’ (5–7 days). Higher scores indicate greater depressive symptoms, with scores at 16 or above indicating need for clinical referral. Cronbach’s alpha was .86. The mean CES-D score for psychiatric inpatients has been reported to be higher than those for samples of nonclinical adults, establishing discriminative construct validity (see also Orme et al., 1986).
Parenting stress was measured at 15, 24, and 36 months using an adaptation of The Parent Role Quality Scale (Barnett and Marshall, 1991). Respondents were asked to indicate on a 4-point scale to what extent each of 20 items was a concern or a reward about raising their children. Sample concern items include, ‘How much of a concern is the financial strain?’ ‘How much of a concern is the unending responsibilities?’ ‘How much of a concern is feeling tied down because of the children?’ Sample reward items include, ‘How rewarding is seeing your child grow and change?’ ‘How rewarding is the meaning and purpose your child gives your life?’ ‘How rewarding is being needed by your child?’ Higher scores indicate greater parenting stresses than rewards. Cronbach’s alpha for the total score is .84. This measure has established good predictive validity as it was a significant predictor of emotional wellbeing for mothers, even after controlling for age, race, SES (socioeconomic status), and job experiences (Barnett and Marshall, 1991).
Analyses
We conducted multi-level growth curve models (Singer and Willett, 2003) in SPSS to independently examine patterns and predictors of maternal depression (Table 4) and maternal stress (Table 5) over time. Because time-varying variables are composed of between- and within-person variation, two separate variables were used in each model to account for each type of variation. To decompose between- and within-person effects for the categorical variables, these variables were kept in the raw scale (Bollen and Curran, 2006). Between-person variation was represented by a person’s score centered on the sample mean (e.g. ‘between-person assessment of income-to-needs ratio’ as seen in Level 2 equation below). Please note the mean of categorical time-varying variables represents the proportion of time points over which the mother was in a particular employment match classification (e.g. ‘between-person assessment of low match’ as seen in Level 2 equation below). Within-person variation was represented by the deviation from a mother’s average score on a given occasion, and thus denotes fluctuation over time relative to the mother’s usual level (e.g. ‘deviation from average income-to-needs ratio’ as seen in Level 1 equation below). We represented categorical within-person assessments the same way. For example, the ‘deviation from average medium employment match status’ (as seen in Level 1 equation below) denoted a mother’s fluctuation relative to her usual medium employment match classification.
The following equation illustrates the within-person portion of growth in depression, or Level 1 of our models, including significant predictors.
Level 1 Model: DEPRESSIONti = π0i + π1iTIMEti + π2i(TIMEti)(TIMEti) + π3i(Deviation from Average Actual Employment/School Hoursti) + π4i(Deviation from Average Ideal Employment/SchoolStatusti) + π5i(Deviation from Average Social Supportti) + π6i(Deviation from Average Medium Matchti) + π7i (Deviation from Average Low Matchti) + π8i (Deviation from Average Inc/Needti) + eti, where the predicted value for depression for each mother ‘I’ on a given occasion ‘t’ is a function of the mother’s average depression at the initial interview (intercept, π0i), plus the linear slope (π1i), plus the quadratic slope (π2i), plus the deviation from the mother’s own average actual employment/school hours (π3i), plus the deviation from her average ideal employment/school status (π4i), plus the deviation from her average perceived social support (π5i), plus the deviation from her average medium employment match status (π6i), plus the deviation from her average low employment match status (π7i), plus the deviation from her average income-to-needs ratio (π8i) and residual variation in depression (eti). Time was scaled in months since initial interview.
The following equation illustrates the between-person portion, or Level 2 of the models, including group mean centered predictors, time invariant predictors, and between-person random effects.
Level 2 Model: π0i = β00 + β01(Maternal Educationi) + β02(Cohabitingi) + β03(Singlei) + β04(Perceived Costs Employi) + β05(Children in home) + β06 (Social Support) + β07(Actual Employment/School) + β08(Ideal Employment/School) + β09(Low Matchi) + β10 (Inc/Needsi) + U0i; π1i = β10 + U1i; π2i = β20 + U2i, where the average depression score (intercept) is a function of the overall sample average depression score (β00) plus maternal education (β01), plus marital status (Cohabiting, β02; Single, β03), plus perceived costs of maternal employment (β04), plus the number of children in the home (β05), plus the between person assessment of social support (β06) plus the between-person assessment of actual employment/school (β07), plus the between-person assessment of ideal employment/school preference (β08), plus the between- person assessment of low employment match status (β09), plus the between-person assessment of the income-to-needs ratio (β10), plus a random deviation from the sample average (U0i). The linear change over time (π1i) in depression is a function of linear time, and random variations in change over time (U1i); and quadratic change in depression (π2i) is a function of quadratic time plus random variations in quadratic change over time (U2i).
Similar equations were used for maternal stress except that with only three time points available in the data set, a linear model with random variations in intercept and linear time fit the data better than a quadratic model.
Results
Attainment of work situation preferences
Table 1 presents the correlations and distributional properties of independent and dependent variables at baseline. Table 2 presents the work-situation preferences and employment match status of mothers at 1, 6, 15, 24, and 36 months after the child’s birth. Mothers’ preferences for employment are somewhat dynamic. Mothers who prefer full-time employment range from 10–12 percent over time; mothers who prefer part-time employment range from 47–49 percent; mothers who prefer to participate in some schooling represent 9–11 percent; and mothers who prefer to be at home range from 29–34 percent. Consistent with other nationally representative data, the largest preference for mothers of young children is part-time employment.
Correlations and distributional properties of independent and dependent variables at baseline (n = 1141)
p < .05; **p < .01; ***p < .001.
Costs mat emp. = Perceived costs of maternal employment.
Children in hm. = Number of children in the home
Notes: The reference group for ethnicity is ‘White non-Hispanic.’ The reference group for marital status is ‘married.’ The reference group for ideal employment circumstance is ‘ideal to stay home full time.’ The reference group for the employment match status variables is ‘high match.’ Baseline for all variables related to maternal depression is 1 month; baseline for time-varying variables (e.g. social support, work hours, ideal full, ideal part, ideal school, medium match, low match, income-to-needs, and depression) related to parenting stress is 15 months.
Percentage of sample by employment preference and employment match status at each time point
The gap between ideal and actual employment is more dynamic. While almost half of mothers preferred part-time employment, only 2–8 percent of mothers achieved their preference. Though a moderate number of mothers preferring no employment (30%) achieve their preference at baseline, by six months after birth only 17 percent of mothers achieve their preferred status; by 15 months this decreases again to 15 percent. Only 2–4 percent of mothers preferring to attend school actually achieve this. Finally only 1–5 percent of mothers preferring full-time employment achieved their preference over time. Findings in this sample expose multiple gaps between mothers’ preferred and actual employment. They also expose multiple sources of untapped talent in the marketplace for mothers in this sample.
Table 3 presents the complete demographic information on the four employment match history categories over time including majority high, majority medium, majority low, and no majority match. Mothers who were in the majority high category were more likely to be white, married, and have higher income-to-needs ratios than mothers in the other match categories.
Descriptive characteristics of independent and dependent variables by majority employment match status (n = 1141)
Notes: Mothers were classified in a majority match employment status group if they reported the specified match at least 3 of the 5 times points measured here (1, 6, 15, 24, and 36 months).
The following reflects the proportion of mothers in each employment match in the entire sample across all 5 time points: Majority high (20%), Majority medium (33%), Majority low (13%), and No majority status (34%).
Maternal depression
Results from baseline indicated that mothers reported an average depression score (intercept) of 10.76 after their transition to parenthood, suggesting that, on average, new mothers had ‘healthy’ levels of depressive symptoms. The intra-class correlation coefficient (ICC) of maternal depression was .50, indicating that half of the variance in the trajectory was owing to within-person change over time, while the other half was owing to between-person differences. A significant linear decrease of –.17 per month in depression emerged when linear effects of time (slope) were added to the model, followed by a slight increase of .004 per month when quadratic effects of time (quadratic slope) were added to the model. Random coefficients indicated a significant amount of variability around the intercept and the linear and quadratic slopes. In Model 1 of Table 4, a fixed quadratic slope was added to a random linear slope; Model 2 added a random quadratic slope; Model 3 added demographic predictors to the random intercept, linear, and quadratic slopes; and Model 4 added employment and income predictors. In both Model 3 and Model 4, we tested for interactions between the variables of interest and slope coefficients (linear and quadratic). No significant interactions were detected.
Unstandardized estimates (standard errors) of Level 1 and Level 2 predictors of maternal depression
p < .10; * p < .05; ** p < .01; *** p < .001.
Notes: AIC = Akaike Information Criteria; BIC = Bayesian Information Criteria.
Model 1 adds a fixed quadratic slope to a random linear slope; Model 2 adds a random quadratic slope; Model 3 adds basic demographic variables; Model 4 adds between-persons (BP) and within-persons (WP) estimates of employment and income variables.
Basic demographic predictors measured at baseline were first considered in the model (see Model 3, Table 4). All reported coefficients were unstandardized, with accompanying standard errors and p-values. Mothers with more than a high school education had significantly lower depression scores than mothers with less (b = –.41(.07), p < .001). Mothers who perceived employment to be detrimental to their child’s development had significantly higher levels of depression than those who perceived low costs of maternal employment (b =.13(.03), p < .001). Mothers who were single had significantly higher depression scores compared to mothers who were married (b = 1.35(.51), p < .01). A mother’s cohabiting status, her ethnicity, and the number of children in her home were not significantly associated with her depressive symptoms.
Social support was significantly associated with maternal depression. The significant between person-effect suggests that mothers who perceived more social support, on average, than other mothers in the sample, were likely to report lower depressive symptoms (b = −5.05(.25), p < .001). The significant within-person effect suggests that at times when a mother perceived more social support than her personal average she reported lower depressive symptoms (b = −2.88(.17), p < .001).
Of the employment and income predictors (see Model 4, Table 4), within-person effects for maternal work hours were negative and significant, suggesting that when a mother increased her work hours beyond her personal average hours of work/school per week, her depression was likely to decrease (b = –.03(.01), p < .001), lending some credibility to the role enhancement hypothesis. Between-person effects for maternal work hours were not significant, continuing to support the notion that deviations based on a mother’s personal experience matter more than the way her experience deviates from other mothers on average as suggested by Barnett (2004).
Mothers who preferred full-time work reported higher levels of depressive symptoms than mothers in the stay at home reference group (b = 1.73(.63), p < .01). No other reports of within-person or between-person ideal employment status were significant.
The between-person and within-person effects for income-to-needs ratio were both significant (see Model 4, Table 4). On average, mothers who reported lower income than others in the sample had significantly higher depression levels. On occasions where mothers had lower income-to-needs ratios relative to their own average, mothers were also more likely to report higher depression levels.
After controlling for key predictors of maternal depression, a mismatch between a mother’s ideal and actual employment status was a significant predictor of her depression (see Model 4, Table 4). A significant between-persons effect for low match status suggests that when compared with mothers in a high employment match, mothers with a low match status report significantly higher depression levels (b = 2.98(.74), p < .001). The significant within-persons effect for low match status suggests that on occasions where a mother in a low match status deviated from her usual employment match status, she reported higher levels of depression (b = .86(.29), p < .01). These findings supported our hypothesis, and further highlight the importance of understanding both between-persons variation and within-persons variation.
Prior research has been unable to test differences in wellbeing between mothers who stay at home but would prefer full-time employment, and those who work full time but would prefer to stay at home. We were able to explore such distinctions and found significant between-person (b = 6.52 (.97), p < .001) and within-person (b = .99(.34), p < .01) effects on depression for mothers who stayed home but preferred full time work or school options. In contrast, mothers who were employed/in school full time but preferred to stay at home did not have significantly different levels of depression from the high match reference group (between person: b = 1.31(1.08), p = .23; within-person: b = .39(.41), p = .35). In sum, mothers who stayed at home, but wanted to be engaged in some other full-time employment or schooling, persistently reported higher depression levels than the average. Mothers in other low match classifications did not. This poses the question: what can employers do to reach mothers of young children who are floundering at home, but might be flourishing in the workplace? While being in the workplace when one prefers to be at home with children may pose other personal concerns for mothers, these data do not provide evidence that over-employment is as risky to a mother’s mental health as under-employment may be.
Overall, predictors accounted for approximately 30 percent of the intercept variance, 7 percent of the linear slope variance, none of the quadratic slope variance, and 7 percent of the residual variance in depression based on pseudo R2 calculation (Singer and Willett, 2003).
Maternal stress
At baseline (15 months after child’s birth), mothers reported an average parenting stress score (intercept) of 34.23. No clinical markers for this stress score are available, but based on the full range of scores (20 to 63), this intercept suggests that mothers are feeling small to moderate levels of parenting stress. The ICC for stress was .60, indicating 60 percent between-person variability versus 40 percent within-person variability over time. Linear effects of time (slope) indicated an initial level of maternal stress at 34.16, followed by a non-significant increase of .007 per month. Random coefficients indicated a significant amount of variability around the intercept and the linear slope. In Model 1 of Table 5, a fixed linear slope was added to a random intercept; Model 2 added a random linear slope; Model 3 added demographic predictors to the random intercept and linear slope; and Model 4 added between-person and within-person estimates of employment and income predictors. In both Model 3 and Model 4, we tested for interactions between the variables of interest and the linear slope coefficient. No significant interactions were detected.
Unstandardized estimates (standard errors) of Level 1 and Level 2 predictors of maternal stress
p < .10, * p < .05; ** p < .01; *** p < .001.
Notes: AIC = Akaike Information Criteria; BIC = Bayesian Information Criteria. Model 1 adds a fixed linear slope to the empty model of severity; Model 2 adds a random linear slope; Model 3 adds basic demographic variables; Model 4 adds between-persons (BP) and within-persons (WP) estimates of employment and income variables.
Maternal education, perceptions of the costs of maternal employment, and the number of children in the home significantly predicted maternal stress (see Model 3, Table 5). Mothers with a high school education experienced less parenting stress than those without (b = .15(.07), p < .05). Mothers who perceived more costs of maternal employment for their child’s development (b = .12(.03), p < .001), and mothers with more children than other women in the sample were more likely to experience higher levels of maternal stress (b = .84(.14), p <.001).
Further, while the within-persons measure of a mother’s income-to-needs ratio was not significant, the between-persons measure of income-to-needs ratio was significant (see Model 4, Table 5). On average, the more money a mother reported having relative to her family’s needs, the less parenting stress she reported (b= –.12(.05), p < .05).
As expected, both between-persons and within-persons variations of mothers’ perceived social support significantly predicted her parenting stress. Mothers who received more social support than other mothers in the sample had decreases in their parenting stress (b = −3.86(.21), p < .001). Mothers who received more social support than their personal average also reported decreased parenting stress (b = −1.33(.16), p < .001).
Employment variables for parenting stress provided more mixed results. Mothers who increased work hours over time (relative to their personal average) experienced decreases in their parenting stress (b = –.015(.008), p < .05), supporting the role enhancement hypothesis of maternal employment. A view of mother’s preferences for employment cast a slightly different result. Mothers who preferred full-time employment (b = 1.68(.52), p < .01), part-time employment (b = 1.49(.51), p <.01), or school attendance (b = 1.74(.66), p < .01) all reported significantly higher parenting stress levels than mothers who reported a preference to stay home full time, providing support for the role strain hypothesis.
Inconsistent with our hypotheses, after controlling for actual employment, ideal employment, and social support, no significant effect of low employment match on maternal stress was found (see Model 4, Table 5). However, mothers who moved into a medium match status experienced higher levels of parenting stress than those who more persistently stayed in a high match status (b =.49(.25), p < .05). It would appear from these mixed findings that considering a mother’s employment hours and her preferences for employment separately may be just as important to understanding a mother’s parenting stress as studying the gap between her ideal and actual work, particularly in the context of the strong buffering effects of social support on maternal stress.
Overall, predictors accounted for approximately 32 percent of the intercept variance, 50 percent of the linear slope variance, and 4 percent of the residual variance in depression based on the pseudo R2 calculation (Singer and Willett, 2003).
Discussion
In the current study, we explored the likelihood of mothers attaining their work situation preferences during the first three years of their child’s life. We further explored whether the gap between ideal and actual work situations was related to maternal depression and maternal stress over time. The analysis confirms previous studies indicating that many mothers in our national sample are not in their preferred work situations (Jacob, 2008; Reynolds, 2003), but goes beyond these studies by exploring the first three years of a mother’s transition to parenthood. The time surrounding the birth of a baby is one of the most critical times of fluctuation in a mother’s employment circumstances (Huston and Holmes, 2004), but prior research has not explored mothers’ preferences over such a large period of time.
Further, our study is able to connect a gap between a mother’s ideal and actual employment status with her depression. Prior work focused on the importance of a mother’s preference in her wellbeing outcomes but did not distinguish ‘between-persons’ changes for mothers from ‘within-persons’ changes. Our study now confirms that a gap between a mother’s ideal and actual employment status is significantly related to variations in her personal gap, as well as the comparison of her gap to other mothers. In other words, mothers’ employment preferences and actual employment circumstances are dynamic, and either variations in their own experiences or deviations from group averages both predict maternal wellbeing. The significant effect of match status was salient even while controlling for the strongest commonly identified predictors of maternal psychological wellbeing. Thus, attaining desired work situations seems to be an important consideration in mothers’ emotional wellbeing. The inability to reconcile work and family preferences with realities over an extended period of time may result in chronic strain and increased vulnerability to stresses in other life domains (Nomaguchi et al., 2005). Opposing theories bring about questions regarding the detriments of over-employment versus under-employment for mothers. We emphasize that mothers who stay at home, but would prefer to be working full time are at the greatest risk for higher depressive symptoms.
Data from this sample highlight the tremendous difficulty mothers may face in attaining their preferred work situations, particularly when children are young. The most common challenge for mothers in this sample seems to be working more than they would like during these early years, though we also identified groups of women (up to 6%) who are not working as much as they would like. While our analyses of low match classifications suggest that over-employment does not have the same deleterious effects on maternal depression as under-employment might, the trend toward significance in the between-persons effect for medium match status (b = 1.13 (.66), p =.09), along with the significant within-persons effect of a medium match status on parenting stress (b = .49(.25), p < .05) suggest that increasing or decreasing a mother with young children’s work hours may also place her at risk for higher depressive symptoms and should not go unnoticed. Employers who are sensitive to a new mother’s work preferences and to any possible transitions away from her employment preference may help buffer the negative effects by helping her ease into her new work status.
We return to the finding that mothers who perceived maternal employment to be detrimental to their child’s development had significantly higher levels of depression and parenting stress. Findings from research on workplace flexibility suggest that the structure of work demands may diminish negative effects of the conflict between work and family demands. Employees with access to flexible workplace options show a greater commitment to their organization and less intentions of turnover (Grover and Crooker, 1995). Beneficial effects of workplace flexibility appear to be greatest when flexibility in both place and time are combined (Hill et al., 2010), with the most obvious benefits for women with children under the age of five years. In an interesting twist, not missing dinner because of work enabled women to work longer hours without reporting the same level of conflict (Jacob et al., 2008). By providing flexible work options that reflect mothers’ work preferences and concerns about their child’s development, employers may be able to minimize concerns, and decrease mothers’ risk for depression and parenting stress.
It is also important to place these findings in the larger socio-economic context. As stated earlier, though this is a large data set, these data are not nationally representative. Though quite comparable to nationally representative data, these data slightly over-represent white mothers and mothers with a higher income-to-needs ratio. Based on our demographic results, non-white mothers and mothers with lower income-to-needs ratios are more likely to be in a low match or medium match employment status. Thus, it is possible that a nationally representative data set would reveal even more discrepancies between employment preferences and actual employment experiences. Nationally representative data sets rarely contain data on maternal employment preferences, but this would be an important next step in data collection efforts.
Further, changing rates of maternal employment and increasing proportions of women as major family breadwinners in the current economic climate provide evidence of the influence of economic need in employment discrepancies for mothers (Hartmann, 2008). Thus, although micro-level explanations for preferences (Hakim, 2000) suggest that women may derive less central meaning from their employment (Caven, 2006) and/or may desire to care for their own families as well as develop their careers (Hays, 1996; Still, 2006), economic realities override the influence of preference. A combination of factors including single-parent families, job instability, and declining real wages for men likely influenced the decision to become employed earlier than many of these mothers preferred (Mischel et al., 1999).
Implications and future directions
The findings suggest a need to focus on understanding mothers’ work situation preferences, enabling them to better attain preferences, or mitigating the effects of maternal employment gap on wellbeing. Exploring options for adjusting work expectations to better reflect reality, or obtaining workplace flexibility options may help address mothers’ feelings of depression and stress. The benefits of addressing maternal stress and depression related to work situations are further illustrated in the implications of maternal wellbeing on children as well as work outcomes including job performance, commitment, engagement, retention, and morale. Policy makers and family life educators would benefit from more extensive evaluations of mothers’ work preferences, the meaning ascribed to those preferences, and information regarding why they are or are not attained. Such research is needed to inform employers and policy makers about the need for options such as schedule flexibility, telecommuting, and part-time employment. Expanding the availability of these employment options could contribute to enabling more mothers to attain their preferred work situations with inherent implications for enhanced maternal wellbeing and benefits to families.
Limitations
There are some limitations to consider in analyses drawing on the SECC data. Only three time points were available to evaluate the maternal stress trajectory, limiting the potential to accurately capture the full nature of change in that trajectory. The difference in the number of time points that could be included for maternal depression and maternal stress also prevented a multivariate multi-level statistical comparison as to whether the gap between ideal and actual work situations more strongly predicted one or the other measure of maternal wellbeing.
The study is limited in data on the work situations of these mothers such as questions on the use of work-flexibility programs, coworker support, and other important factors that would predict work situation preferences. Further, no qualitative data existed to provide additional explanation for the findings. Finally, some concerns about shared method variance may be raised because measures of predictor and criterion variables were not always available from different sources. Podsakoff et al. (2003) suggest that one possible remedy is to ‘proximally or methodologically separate the measures by having respondents complete the measurement of the predictor variable under conditions . . . different from the ones under which they complete the measurement of the criterion variable’ (pp. 887–888). This remedy exists in the SECC data set as mothers responded at separate occasions using different media (phone call interviews versus pencil and paper responses in the home) for the criterion and predictor variables. The use of different stories to introduce each measure and psychologically distance one measure from another further minimized common method variance.
Taking these limitations into account, this data set provides access to unique insights on multiple dimensions of maternal wellbeing for a relatively large and diverse sample over an important transitional period. Previous explorations of the potential discrepancy between actual and preferred work situations for mothers have been limited to cross-sectional analyses, with substantially fewer variables, and a very limited view of associated outcomes (Jacob, 2008).
Footnotes
Acknowledgements
We are grateful to the Eunice Kennedy Shriver National Institute of Child Health and Human Development Early Child Care Research Network for designing and carrying out the data collection for this project. The NICHD Study of Early Child Care is a study directed by a Steering Committee and supported by NICHD through a cooperative agreement that calls for scientific collaboration between the grantees and the NICHD staff. The content of this project is solely the responsibility of the named authors and does not represent the official views of the Eunice Kennedy Shriver National Institute of Child Health and Human Development, the National Institute of Health, or individual members of the Network. We are also grateful to Lesa Hoffman, Joseph Olsen, and Jeremy Yorgason for helpful comments on earlier drafts.
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.