Abstract
We use data from the Fragile Families and Child Wellbeing Study (N = 12,140 person–waves) to identify characteristics associated with mothers’ having or lacking “personal safety net” support from family and friends. We focus on characteristics that are likely to increase the importance of having support available but may also interfere with the maintenance of supportive ties: poverty, poor physical and mental health, and challenging child rearing responsibilities. By capitalizing on distinctions among these types of personal disadvantages and among types of personal safety nets (financial, housing, child care, and emotional), we help to explain why personal disadvantages are associated with weaker support. Our paper contributes to the literature emphasizing the importance of reciprocity in support relationships and introduces the idea that families that are more difficult to help will have less support available.
Keywords: Exchange relationships, Fragile Families and Child Wellbeing, perceived support, personal safety nets, poverty, social support
The high rates of nonmarital childbearing and marital instability that characterize the U.S. family system, coupled with a relatively meager social welfare system, mean that family and friend networks often play an important role in supporting mothers and young children (Swartz, 2009). Having a “personal safety net” of basic supports is important for buffering against adverse events and has been shown to have a robust correlation with measures of well-being. In the romanticized version of support networks, network members are altruistic and willing to help in response to need. In reality, not all mothers have support available, and we lack a good understanding of which mothers lack support and why support is not available to them.
Previous research on support relationships has tended to fall into one of two categories. Some research focused on a particular community and used an ethnographic approach, illustrating the strong reciprocity norms that prevail in resource-scarce environments (Dominguez & Watkins, 2003; Menjivar, 2000; Nelson, 2000). In contrast, quantitative studies took a broader but less deep approach. These studies documented a correlation between personal disadvantages and lacking a safety net but did not explore the mechanisms linking the two (Harknett, 2006; Hogan, Eggebeen, & Clogg, 1993). Our paper aims to fill a gap at the intersection of these types of studies. Using national, longitudinal data from the Fragile Families and Child Wellbeing Study, we leverage distinctions among types of personal disadvantages and among types of support to test theories for why some mothers lack support.
Our paper pursues two related aims. First, we examine which mothers have and which mothers lack various types of support from friends or family. We pay particular attention to how mothers’ personal disadvantages—which we define as poverty, health problems, having a large number of children, or having an unhealthy child—are associated with personal safety nets. These personal disadvantages are likely to increase the need for support but may also decrease support availability. In defining support, we focus on perceived supports that are important for mothers caring for young children: anticipated financial, housing, child-care, and emotional support. Understanding who lacks a personal safety net to provide support in times of need is important for identifying at-risk populations and informing public safety net programs.
Second, our paper explores the reasons why some mothers lack support. Here, we take advantage of the conceptual distinctions among types of disadvantages and types of support. We use our pattern of results to make inferences about the underlying motivations and requirements for exchange relationships. We consider several theoretical pathways that may lead to a lack of support from personal networks: having a resource-poor network, having conditions that interfere with the formation and maintenance of social ties, and having characteristics that impose a greater burden on would-be support providers. Our research contributes to a literature on the limits of altruism and the importance of reciprocity in exchange relationships.
Theoretical Framework and Hypotheses
Having a personal safety net to draw on in times of need is correlated with well-being (Sarason, Sarason, & Pierce, 1990). A large body of research documented a robust relationship between personal safety nets (also referred to as perceived support) and better physical and mental health (Berkman, Glass, Brissette, & Seeman, 2000; House, Umberson, & Landis, 1988; Thoits, 1995). Prior research focused most often on the positive correlation between perceived emotional support and well-being, but recent research found that perceived financial and other instrumental supports were also positively related to material well-being and child well-being (Harknett, 2006; Henly, Danziger, & Offer, 2005; Ryan, Kalil, & Leininger, 2009).
Personal safety nets may be of particular importance for mothers whose personal disadvantages place them at greater risk of needing support. In addition to poverty, we also consider mothers to be at higher risk of needing assistance if they reported physical or mental health problems, had a large number of children to care for, or had a child in poor health.
In this paper, we focus for several reasons on perceived support rather than actual support received. First, measures of actual support received are conflated with need. Those who do not receive support are a mix of people who do not need any help and those who need help but have none available. Perceived support, in contrast, is unambiguously desirable. Second, prior research found that perceived support was more strongly correlated with psychological well-being than was received support (Thoits, 1995; Turner & Marino, 1994; Wethington & Kessler, 1986). Last, regular exchanges of support are relatively uncommon in the United States, so point-in-time measures of support received are not a good reflection of support availability (Swartz, 2009).
Why and in what ways should we expect mothers’ personal disadvantages to be related to perceived support? There are several possible pathways, which we display in Figure 1. The mechanisms in the figure correspond to the hypotheses described below.
Figure 1.
Conceptual Diagram
Hypothesis 1. Poverty will be associated most strongly with lacking a financial safety net and less strongly with other, non-monetary forms of support.
Personal disadvantages may be related to perceived support because of homophily of economic resources within social networks (Mechanism 1). Low-income mothers tend to be embedded in social networks that are also impoverished and, as a consequence, constrained in their ability to provide financial assistance (Lye, 1996; Swartz, 2009; McPherson, Smith-Lovin, & Cook, 2001). For this reason, even poor mothers with close kin relationships and friendships may nevertheless lack financial assistance because their friends and families lack financial resources. Consistent with this idea, there is some evidence that poverty interferes with some types of support more than others. For instance, researchers found that neighborhood poverty did not diminish support that required time but was negatively related to support that required material resources (Turney & Harknett, 2010). Although providing housing support requires that one have housing resources, families can and do “double up” (Luo, 2010), so we consider the poverty of network members less of an impediment to housing support than it is to financial support.
Hypothesis 2. Personal disadvantages will be negatively associated with all types of perceived support.
Mothers in poverty, poor health, and with greater child-rearing burdens are likely to face a dilemma: These mothers may need more frequent support because of their disadvantaged circumstances but be constrained in their ability to reciprocate support. Consequently, these mothers may be especially prone to falling behind in reciprocal support relationships and alienating would-be support networks.
Our second hypothesis, that personal disadvantages will be associated with lacking all forms of support, follows from the idea that mothers’ personal disadvantages will act as an impediment to relationships with potential support providers, relationships that are typically contingent on reciprocity (Mechanism 2). Quantitative studies found that poverty was associated with less support (Fischer, 1982; Harknett, 2006; Henly et al., 2005; Hogan et al., 1993; Lye 1996). One possible explanation is provided by ethnographic research, which found that where resources were scarce, norms of reciprocity were strong (Dominguez & Watkins, 2003; Nelson, 2000; Menjivar, 2000). This research highlighted the stress and stigma that mothers experienced when they accepted financial or instrumental assistance and were unable to reciprocate and also the deterioration in personal relationships that resulted from unreciprocated support.
Poor health and depression were also linked to a lack of support in prior research (Dean & Ensel, 2006; House et al., 1988; Kawachi & Berkman, 2001; Thoits, 1995). Generally, this research focused on the lack of support as a cause of poor health or depression and has paid less attention to the possibility that depression or poor health may interfere with the maintenance of support relationships. Research on the stigma and social distance associated with mental illness suggests that depression may have a negative influence on personal safety nets (Link, Phelan, Bresnahan, Stueve, & Pescosolido, 1999).
Researchers found that child health problems had a destabilizing effect on marriage (Mauldon, 1992; Reichman, Corman, & Noonan, 2004). By the same token, maternal care for children in poor health may interfere with the maintenance of mothers’ relationships with other supportive adults and result in a lower perception of all types of support. The time and energy demands on mothers caring for many children are also likely to interfere with mothers’ relationships with supportive adults (Bird, 1997; Goldsteen & Ross, 1989).
Hypothesis 3. Personal disadvantages will be positively associated with all types of perceived support.
Alternatively, personal disadvantages may be positively associated with perceived support if social network members are responsive to need and motivated by altruism, or if norms of reciprocity are flexible (Mechanism 3). In theory, altruism is a key motivator in exchange relationships, particularly among kin (Lye, 1996). Reciprocity may be unbalanced, meaning that reciprocation does not need to come in the same form as the initial support (Ishii-Kuntz & Seccombe, 1989; Silverstein, Conroy, Wang, Giarrusso, & Bengtson, 2002). Reciprocation may also be delayed. Parents may help their adult children with an expectation of reciprocation when they become elderly (Silverstein et al.). Therefore, an inability to reciprocate in the short term, or in the same terms, need not decrease personal safety nets, and we may observe that needier mothers perceive more support. Some research finds evidence of altruism, in particular in transfers of actual support from parents to adult children. (Suitor, Sechrist, & Pillemer, 2007, and Fingerman, Miller, Birditt, & Zarit, 2009, are recent examples.)
Hypothesis 4. Having many children or a child in poor health will be negatively associated with perceived housing and child-care support. The relationship between these child characteristics and financial or emotional support will be relatively weaker.
Another possibility, which has received less attention to date, is that some personal disadvantages—such as having many children or a child in poor health—may affect perceived support because they raise the level of difficulty involved in providing supports such as child care or shared housing (Mechanism 4). We assume that caring for multiple children is more challenging than caring for one child. In prior research, the number of children in the home increased young mothers’ psychological distress and perceived burden from children (Bird 1997; Goldsteen & Ross 1989). Likewise, caring for a child in poor health is likely to be more challenging than caring for a healthy child. These child characteristics should not make it any more difficult for friends or family to provide financial or monetary support to mothers.
Hypothesis 5. The relationship between personal disadvantages and personal safety nets will be attenuated or eliminated in fixed-effects models.
A final possibility is that the correlation between personal disadvantages and personal safety nets is spurious and driven by unobserved characteristics (Mechanism 5). This would be the case if, for instance, mothers’ poor social skills or other unobserved characteristics lead both to personal disadvantages and to lacking a personal safety net. In this case, mothers’ unobserved characteristics act as confounders and the relationship between personal disadvantages and personal safety nets is weakened or eliminated when these characteristics are taken into account.
Other Correlates of Support
In prior research, younger parents received and perceived more support than older parents (Cooney & Uhlenberg, 1992). Parents who grew up with a single parent or who had children with multiple partners reported less perceived instrumental support (Swartz, 2009; White, 1992; Harknett & Knab, 2007). Immigrants reported lower levels of perceived support (Turney & Kao, 2009). Some researchers found that White mothers reported more financial support (Cooney & Uhlenberg; Eggebeen, 1992; Hogan et al., 1993), and others found that African American mothers reported more child-care support (Benin & Keith, 1995). Marriage and religious participation have long been considered markers of social integration in sociological research dating back to Durkheim (1951[1897]). Yet, recent research found that those who were married received and provided less instrumental and emotional support than those who were unmarried (Sarkisian & Gerstel, 2008). Prior research found a positive association between religious participation and many dimensions of social support (Bradley, 1995; Ellison & George, 1994).
METHOD
The data for this paper come from the longitudinal Fragile Families and Child Wellbeing Study. The study follows a birth cohort of children born in twenty U.S. cities with populations greater than 200,000 and is representative of nonmarital births in large U.S. cities with a comparison sample of marital births. Baseline information was collected around the time of a birth, and follow-up surveys were administered when the focal child was around 1, 3, and 5 years old. The birth that initiated inclusion in the study was a first birth for 38% of mothers and a higher order birth for 62% of mothers. Mothers’ response rates were 86% for the baseline survey. Among the baseline sample, 90%, 88%, and 87% responded to the Year 1, Year 3, and Year 5 surveys, respectively. The original Fragile Families sample consists of 4,898 focal children and their parents.
We restrict our analysis to the follow-up survey waves that took place 1, 3, and 5 years after the birth. We omit the baseline survey because perceptions of support may be biased by the recent birth experience. By focusing on the follow-up surveys, all mothers in our sample, including the first-time mothers, have had time to learn about the willingness of their friends and families to help them.
Our analyses are based on N = 12,140 person–waves of data, representing 83% of all possible person–waves (4,898 mothers surveyed at baseline × 3 survey waves = 14,694 person waves). Most of the excluded cases consist of mothers who did not respond to follow-up surveys. About 2% (n = 299) of cases are excluded because no children were living in the household, and our focus is on the support available to mothers. Another 2% (n = 285) of cases are excluded because of missing values on one of the measures of personal disadvantage. Less than 1% of cases are missing data on a control variable, and we use a single-imputation regression to predict these missing values. The results we present are consistent with results generated using listwise deletion.
The Fragile Families study oversampled unmarried parents by design. Following Winship and Radbill (1994), we present unweighted results that control for marriage, the main sample stratification variable. Separate analyses of married and unmarried samples yield a pattern of results similar to those we present.
Dependent variables
Measures of perceived instrumental support are based on the mothers’ responses to three survey questions asked at each follow-up wave: “If you needed help during the next year, could you count on someone to … Loan you $200? Provide you with a place to live? Help you with emergency child care?” We construct a perceived instrumental support scale using principal components factor analysis. The alpha reliability for the scale is.72.
We also analyze the three components of the perceived instrumental support scale individually, along with a separate measure of emotional support. We do not include our measure of emotional support in the scale because emotional support was only measured in the last follow-up survey (Wave 4). Our emotional support measure is based on the question: “Is there any special person you know that you feel very close with—someone you share confidences and feelings with; someone you can depend on?” The financial, housing, child care, and emotional support variables are coded 0 = no or don’t know and 1 = yes to the questions above. Don’t know responses were given by no more than 1% of mothers for any of these measures.
Descriptive statistics for our dependent and independent variables appear in Table 1. As shown, most mothers anticipated having each type of support available, but 16% lacked perceived financial or housing support, 12% lacked perceived child-care support, and 8% lacked perceived emotional support.
Table 1.
Mother and Child Characteristics: Descriptive Statistics (n = 12,140, from 4,618 Women)
| Means or %s | SD | |
|---|---|---|
| Perceived support scale | 0.011 | 0.794 |
| Anticipates availability of $200 loan (%) | 84.1 | -- |
| Anticipates availability of place to live (%) | 84.1 | -- |
| Anticipates availability of childcare (%) | 87.8 | -- |
| Has someone to confide in (emotional support)a (%) | 91.5 | -- |
| Below poverty line (%) | 39.1 | -- |
| Near poor (100–200% poverty) (%) | 26.9 | -- |
| Middle income (200–500% of poverty) (%) | 26.8 | -- |
| High income (>500% of poverty) (%) | 7.3 | -- |
| Poor health scale (1=excellent to 5=poor) | 2.3 | 1.0 |
| Depressed (based on CIDI criteria, self-reports) (%) | 17.2 | -- |
| Has one child (%) | 30.6 | -- |
| Has two children (%) | 35.6 | -- |
| Has three children (%) | 19.7 | -- |
| Has four or more children (%) | 14.2 | -- |
| Focal child poor health scale (1=excellent to 5=poor) | 1.5 | 0.8 |
| Has less than high school education (%) | 27.8 | -- |
| Has high school diploma (%) | 24.6 | -- |
| Has some post-secondary education (%) | 34.7 | -- |
| Has college degree (%) | 12.9 | -- |
| Employed (%) | 57.2 | -- |
| Married (%) | 34.3 | -- |
| Cohabiting (%) | 30.5 | -- |
| Never attends religious servicesb (%) | 14.4 | -- |
| Occasionally attends religious servicesb (%) | 63.4 | -- |
| Attends religious services weeklyb (%) | 22.2 | -- |
| Age (years) | 28.3 | 6.2 |
| Whiteb (%) | 22.1 | -- |
| Blackb (%) | 47.7 | -- |
| Hispanicb (%) | 26.5 | -- |
| Other raceb (%) | 3.7 | -- |
| Immigrantb (%) | 15.5 | -- |
| Has child with different partner (%) | 40.0 | -- |
| Lived with both parents at age 15b (%) | 43.4 | -- |
Note: Unless otherwise noted, characteristics were measured in Waves 2–4.
Measured in Wave 4.
Measured in Wave 1.
Independent variables
Our independent variables of primary interest are measures of three types of personal disadvantage: poverty, physical or mental health problems, and child rearing burden. All of these variables were measured at each survey wave.
Our poverty measure is based on the ratio of household income to the poverty threshold, which varies by year and family size. In the year 2000, for example, the income-to-poverty ratio for a family of four would be the family income divided by $17,050. We separate income-to-poverty ratios into four categories: poor (below the poverty line), near poor (less than twice the poverty line), middle income (between two and five times the poverty line), and high income (more than five times the poverty line).
To capture health status, we include a continuous measure of mothers’ self-rated health, ranging from 1 = excellent to 5 = poor. Separate analyses found that the relationship between the health scale and dependent variables follows a linear pattern, supporting a continuous variable specification. We also use a measure of whether a mother meets the Composite International Diagnostic Interview Short Form (Version 1.0, November 1998) criteria for depression based on her reports of three or more depressive symptoms or her current use of anti-depressant medication (Kessler, Andrews, Mroczek, Ustun, & Wittchen, 1998). Separate analyses found that the relationship between depressive symptoms and dependent variables was not linear but was instead better represented by the dichotomous categories.
The number of children is based on a household roster that asked the age and relationship of each household member to the mother. We count household members identified as biological or adopted children, stepchildren, or foster children who are between the ages of 0 and 17 in our measure of number of children. Separate analyses restricted only to biological children yield similar results. In our analysis, the number of children in the household is measured with dummy variables indicating that two, three, or four or more children reside in the household. One child is the omitted category.
Our measure of child health is based on mothers’ reports of the focal child’s overall health, ranging from 1 = excellent to 5 = poor. The focal child is the child whose birth initiated the family’s inclusion in the Fragile Families study. The health status of other children in the household, those that were not the focal child, was not collected in the surveys.
Control variables
Our analysis controls for mothers’ basic demographic characteristics, such as age, race, and ethnicity. We also control for mothers’ education level (less than high school, high school only, some college, or college degree), employment status (employed or not employed), relationship status (married, cohabiting, or neither), religious attendance (weekly, occasional, or rarely or never), and whether the mother was raised by both biological parents. Control variables were measured at each survey wave with the exception of religious attendance and time-invariant characteristics such as race, ethnicity, and whether mothers grew up with two parents, which were measured in the baseline survey.
Analytic approach
Our first set of models uses pooled, person–wave data with random effects. In the pooled data, mothers contribute one observation for each wave in which they were interviewed. The random effects adjust for the nonindependence of observations for mothers who contribute more than one observation. These random-effects models take advantage of all of the variation in the data: variation across individuals and over time within individuals. These models tell us, for instance, whether mothers who are poor also tend to lack various types of personal safety nets. These models leave open the possibility that unmeasured characteristics of mothers are driving the relationship between personal disadvantages and support deficits.
The second set of models includes person fixed effects. These model estimates are based solely on over-time variation within individuals. The fixed effects hold constant all time-invariant characteristics of mothers. Therefore, the relationships in the fixed-effects models are more likely to imply a causal relationship between personal disadvantages and supports. These models allow us to identify whether a change in the predictors of interest—poverty, health, and child rearing burdens—is associated with a change in support for individual women over time. These models also include controls for observed changes in marital and employment status, accounting for the influence of a divorce or job loss. Analyses were conducted in Stata SE 10.0 using the xtreg and xtlogit commands.
We compare the random-effects models to fixed-effects models using a Hausman test. The Hausman test results suggest that we can reject the null hypothesis that the coefficients from the random- and fixed-effects models are statistically indistinguishable. We present both random-and fixed-effects models because both contribute to our analytic goals. The random-effects models show the extent to which personal disadvantages and deficits in personal safety nets go hand in hand and address our aim of describing who lacks a personal safety net. The fixed-effects models come closer to estimating causal relationships between personal disadvantages and personal safety nets.
RESULTS
Table 2 shows the bivariate relationships between personal disadvantages and personal safety nets. In the absence of statistical controls, the gap between the highest income mothers and poor mothers is large: 97% of high-income mothers and 74% of poor mothers report having a financial safety net. The gaps between high-income and poor mothers in housing and child-care support are similar but slightly smaller in magnitude. High-income mothers are more likely than poor mothers to have emotional support, but the gap is smaller than for the instrumental supports. Poorer physical and mental health and challenging child-rearing responsibilities are also associated with weaker personal safety nets.
Table 2.
Bivariate Relationships Between Key Independent Variables and Perceived Support (n = 12,140, from 4,618 Women)
| Mother anticipates availability of: | ||||
|---|---|---|---|---|
| $200 loan (%) |
Place to live (%) |
Child care (%) |
Emotional supporta (%) |
|
| Mother's income | ||||
| High income (>500% of poverty) | 97.4 | 94.6 | 96.0 | 97.1 |
| Middle income (200–500% of poverty) | 93.1 | 92.5 | 94.6 | 95.8 |
| Near Poor (100–200% poverty) | 85.1 | 84.6 | 88.2 | 91.5 |
| Below poverty line | 74.3 | 75.9 | 81.2 | 87.3 |
| Mother's physical and mental health | ||||
| In good, very good, or excellent health | 85.8 | 85.8 | 89.1 | 91.8 |
| In fair or poor health | 69.5 | 70.7 | 76.9 | 87.7 |
| Not depressed | 85.7 | 86.0 | 89.2 | 91.6 |
| Depressed | 73.9 | 72.9 | 79.2 | 89.4 |
| Mother's childrearing responsibilities | ||||
| Has one child | 85.8 | 88.2 | 91.0 | 92.6 |
| Has two children | 85.5 | 85.9 | 89.5 | 92.1 |
| Has three children | 82.6 | 81.4 | 85.2 | 90.0 |
| Has four or more children | 75.6 | 72.2 | 77.9 | 89.4 |
| Focal child in good, very good, or excellent health | 84.1 | 84.4 | 87.8 | 91.4 |
| Focal child in fair or poor health | 69.9 | 63.2 | 75.3 | 81.0 |
Note: Unless otherwise noted, characteristics were measured in Waves 2–4.
Measured in Wave 4.
Table 3 presents our multivariate results. Our findings are consistent with Hypothesis 1, that poor mothers lack a financial safety net because their friends and family members are also likely be poor. This hypothesis predicts that poverty will be associated most strongly with lacking a financial safety net and less strongly with non-monetary support. We do find that the support gap between the highest and lowest income groups is largest for financial support, but the gap is also significant and only slightly smaller in magnitude for housing and child-care support. The gap is somewhat smaller for emotional support, consistent with the prediction of Hypothesis 1.
Table 3.
Perceived Support Regressed on Personal Disadvantages and Control Variables
| Perceived support scalea (OLS) |
Mother anticipates availability of: | ||||
|---|---|---|---|---|---|
| $200 Loana (logit) |
Place to livea (logit) |
Child carea (logit) |
Emotional supportb (logit) |
||
| Below poverty line | −0.208 *** | −1.856 *** | −1.223 *** | −1.324 *** | −0.909 * |
| (0.035) | (0.317) | (0.251) | (0.274) | (0.426) | |
| Near poor (100–200% poverty) | −0.119 *** | −1.429 *** | −0.836 *** | −1.03 *** | −0.677 |
| (0.034) | (0.314) | (0.247) | (0.269) | (0.419) | |
| 200–500% poverty line | −0.037 | −0.839 ** | −0.239 | −0.371 | −0.201 |
| (0.031) | (0.308) | (0.239) | (0.259) | (0.411) | |
| Above 500% of poverty line (ref.) | |||||
| Poor health scale | −0.054 *** | −0.261 *** | −0.197 *** | −0.269 *** | 0.060 |
| (0.007) | (0.039) | (0.038) | (0.040) | (0.061) | |
| Depressed | −0.145 *** | −0.525 *** | −0.757 *** | −0.523 *** | −0.181 |
| (0.018) | (0.097) | (0.093) | (0.098) | (0.158) | |
| One child (ref.) | |||||
| Two children | 0.000 | 0.160 | −0.108 | −0.062 | −0.080 |
| (0.018) | (0.110) | (0.108) | (0.116) | (0.180) | |
| Three children | −0.046 * | 0.131 | −0.310 * | −0.312 * | −0.148 |
| (0.023) | (0.133) | (0.129) | (0.136) | (0.195) | |
| Four or more children | −0.162 *** | −0.131 | −0.775 *** | −0.748 *** | −0.140 |
| (0.028) | (0.149) | (0.143) | (0.149) | (0.211) | |
| Focal child poor health scale | −0.036 *** | −0.082 + | −0.242 *** | −0.094 + | −0.191 ** |
| (0.009) | (0.049) | (0.047) | (0.049) | (0.073) | |
| Less than high school (ref.) | |||||
| High school diploma | 0.093 *** | 0.393 ** | 0.289 * | 0.298 * | 0.259 |
| (0.025) | (0.125) | (0.120) | (0.125) | (0.161) | |
| Post-secondary education | 0.107 *** | 0.507 *** | 0.296 * | 0.329 ** | 0.428 ** |
| (0.023) | (0.120) | (0.115) | (0.120) | (0.152) | |
| College degree | 0.110 ** | 1.097 *** | 0.653 ** | −0.083 | 0.284 |
| (0.037) | (0.238) | (0.212) | (0.212) | (0.284) | |
| Employed | 0.043 ** | 0.310 *** | 0.175 * | 0.172 * | 0.067 |
| (0.014) | (0.081) | (0.079) | (0.083) | (0.125) | |
| Married | 0.091 *** | 0.563 *** | 0.342 ** | 0.379 ** | 0.464 ** |
| (0.020) | (0.119) | (0.114) | (0.121) | (0.172) | |
| Cohabiting | 0.074 *** | 0.434 *** | 0.194 * | 0.259 ** | −0.005 |
| (0.016) | (0.092) | (0.090) | (0.095) | (0.140) | |
| Attend religious services rarely/never (ref.) | |||||
| Occasional | 0.124 *** | 0.575 *** | 0.386 ** | 0.545 *** | 0.470 ** |
| (0.027) | (0.132) | (0.127) | (0.128) | (0.155) | |
| Weekly | 0.115 *** | 0.480 ** | 0.396 ** | 0.539 *** | 0.186 |
| (0.032) | (0.159) | (0.152) | (0.156) | (0.185) | |
| Age | −0.005 ** | −0.012 | −0.041 *** | −0.021 * | 0.001 |
| (0.002) | (0.009) | (0.008) | (0.009) | (0.011) | |
| Race (reference=White) | |||||
| Black | −0.118 *** | −0.721 *** | −0.699 *** | −0.568 *** | −0.240 |
| (0.026) | (0.151) | (0.142) | (0.146) | (0.201) | |
| Hispanic | −0.045 | −0.458 ** | −0.277 + | −0.313 + | −0.239 |
| (0.030) | (0.171) | (0.162) | (0.167) | (0.224) | |
| Other race | −0.151 ** | −1.105 *** | −1.008 *** | −0.654 * | −0.345 |
| (0.054) | (0.284) | (0.260) | (0.277) | (0.355) | |
| Immigrant | −0.153 *** | −0.642 *** | −1.112 *** | −0.348 * | −0.670 *** |
| (0.030) | (0.155) | (0.145) | (0.154) | (0.181) | |
| Has child with a different partner | −0.093 *** | −0.390 *** | −0.374 *** | −0.422 *** | −0.166 |
| (0.020) | (0.106) | (0.101) | (0.106) | (0.137) | |
| Lived with both parents at age 15 | 0.073 *** | 0.397 *** | 0.374 *** | 0.376 *** | 0.070 |
| (0.020) | (0.108) | (0.102) | (0.105) | (0.132) | |
| N | 12140 | 12140 | 12140 | 12140 | 3916 |
Notes: OLS or logit coefficients and (standard errors) are shown.
Results based on person-wave data with random effects.
Emotional support was only measured at Wave 4, so this model is cross-sectional.
p < .001;
p < 0.01;
p < 0.05;
p < 0.10.
As predicted by Hypothesis 2, we find that personal disadvantages are associated with lacking all types of personal safety nets. Poverty is associated with lacking support. This relationship is apparent even after controlling for many of the other disadvantages that often accompany poverty. Compared with high-income mothers, poor mothers average 0.208 points lower on the perceived instrumental support scale, which has a range of about 3 points. Mothers in the near-poor income group also report significantly less support than mothers in the high-income group. This pattern holds across support types.
Also consistent with Hypothesis 2, poor physical and mental health are associated with lacking a safety net. The gap in support between mothers in the best and worst health is about 0.216 on the 3-point perceived instrumental support scale. Mothers who report symptoms of depression average 0.145 points lower on the perceived instrumental support scale. Mothers with poor health or depression are significantly less likely to report having financial, housing, or child-care support available to them, but there is no significant relationship to emotional support. This weaker result for emotional support is not an artifact of emotional support only being available in one survey wave: Analyses restricted to Wave 4 data confirm the stronger relationship for instrumental (financial, housing, child care) compared with emotional support.
We also find support for the Hypothesis 2 prediction that mothers’ challenging child rearing responsibilities—having many children to care for or a child in poor health—are negatively associated with the perceived instrumental support scale. The average difference in the perceived instrumental support scale between mothers with one child and mothers with four or more children is 0.162 on the 3-point scale. The gap in support between mothers with focal children in the best and worst health is about 0.144 on the perceived instrumental support scale. This pattern holds across support types, though not all coefficients are statistically significant.
We find no support for Hypothesis 3, that personal disadvantages increase safety net support by triggering an altruistic response to need. We find no instances in which a personal disadvantage is positively associated with personal safety net support.
As predicted by Hypothesis 4, child characteristics that increase caregiver burden are associated with lacking the types of support that require direct contact with children—providing a place to live and child care. The support for Hypothesis 4 is relatively stronger for number of children than for child health. Consistent with Hypothesis 4, having many children is associated with lacking housing and child-care support but is not associated with financial or emotional support. Also as predicted, a child’s poor health is associated with lacking housing and child-care support, though the child-care coefficient is weak (p < .10). Although not predicted by Hypothesis 4, a child’s poor health is associated with lacking emotional support.
The relationships between control variables and perceived support are consistent with expectations and prior research. Low education and unemployment are associated with weaker personal safety nets. Although education and employment are not characteristics we focus on, these correlations are consistent with the general pattern that personal disadvantages are associated with lacking a personal safety net. Mothers’ embeddedness in marriage, cohabitation, and religious institutions are positively and significantly associated with personal safety nets. Non-White and immigrant mothers are less likely to have a personal safety net. Multipartnered fertility is negatively related to support, and growing up with two parents is positively related to support.
Table 4 presents results from our fixed-effects analyses, which control for all characteristics of mothers that are time-invariant. As predicted by Hypothesis 5, the fixed-effects specification reduces the magnitude of the relationships between personal disadvantages and personal safety nets. In most cases, however, a statistically significant relationship persists. Emotional support could not be included in our fixed-effects analyses, because this type of support was only measured in Wave 4.
Table 4.
Perceived Support Regressed on Personal Disadvantages, Controls, and Person Fixed Effects
| Perceived support scale (OLS) |
Mother anticipates availability of: | |||
|---|---|---|---|---|
| $200 loan (logit) |
Place to live (logit) |
Child care (logit) |
||
| Below poverty line | −0.101 * | −0.832 + | −0.531 | −1.167 ** |
| (0.044) | (0.500) | (0.374) | (0.430) | |
| Near poor (100–200% poverty) | −0.087 * | −0.821 + | −0.393 | −1.22 ** |
| (0.043) | (0.498) | (0.368) | (0.423) | |
| 200–500% poverty line | −0.051 | −0.666 | −0.190 | −0.877 * |
| (0.039) | (0.481) | (0.355) | (0.405) | |
| Above 500% of poverty line (ref.) | ||||
| Poor health scale | −0.027 ** | −0.081 | −0.054 | −0.176 *** |
| (0.008) | (0.050) | (0.049) | (0.053) | |
| Depressed | −0.075 *** | −0.208 + | −0.378 *** | −0.164 |
| (0.021) | (0.123) | (0.115) | (0.125) | |
| One child (ref.) | ||||
| Two children | −0.013 | 0.122 | −0.140 | −0.075 |
| (0.024) | (0.166) | (0.159) | (0.174) | |
| Three children | −0.061 + | 0.104 | −0.394 + | −0.331 |
| (0.034) | (0.214) | (0.208) | (0.230) | |
| Four or more children | −0.085 + | 0.197 | −0.535 * | −0.599 * |
| (0.045) | (0.267) | (0.265) | (0.287) | |
| Focal child poor health scale | −0.007 | 0.089 | −0.140 * | 0.023 |
| (0.011) | (0.063) | (0.062) | (0.065) | |
| Less than high school (ref.) | ||||
| High school diploma | −0.030 | −0.589 | −0.149 | 0.058 |
| (0.065) | (0.386) | (0.365) | (0.421) | |
| Post-secondary education | −0.010 | −0.140 | −0.016 | −0.010 |
| (0.056) | (0.288) | (0.295) | (0.328) | |
| College degree | −0.112 | −0.040 | −0.677 | −1.867 + |
| (0.099) | (0.749) | (0.702) | (1.084) | |
| Employed | −0.002 | 0.103 | 0.012 | −0.144 |
| (0.016) | (0.102) | (0.101) | (0.109) | |
| Married | 0.027 | 0.148 | 0.000 | 0.163 |
| (0.028) | (0.182) | (0.177) | (0.189) | |
| Cohabiting | 0.063 ** | 0.353 ** | 0.139 | 0.308 * |
| (0.020) | (0.117) | (0.117) | (0.124) | |
| Age | 0.007 + | 0.048 + | 0.000 | 0.032 |
| (0.004) | (0.025) | (0.024) | (0.026) | |
| Has child with a different partner | −0.040 | 0.007 | −0.186 | −0.25 |
| (0.038) | (0.224) | (0.212) | (0.247) | |
| N | 12140 | 2595 a | 2730 a | 2349 a |
Notes: OLS or logit coefficients and (standard errors) are shown.
p < .001;
p < 0.01;
p < 0.05;
p < 0.10.
In logistic regression models with fixed effects, sample sizes reflect mothers with variation across survey waves, because mothers without variation are automatically excluded.
Table 4 shows that the negative relationship between poverty and the perceived instrumental support scale persists but becomes smaller in magnitude after controlling for all of mothers’ stable characteristics. Short-term increases or decreases in mothers’ incomes are associated with respective increases or decreases in perceived instrumental support. Similarly, the negative relationship between mothers’ poor physical and mental health and the perceived instrumental support scale also persists in our fixed-effects analyses. A deterioration (or improvement) in physical or mental health between survey waves is associated with a decline (or gain) in perceived support.
The fixed-effects analysis shows that an increase in the number of children a mother is caring for is associated with a decline in perceived instrumental support. Although the coefficients for three or four children are only marginally significant, jointly they are statistically significant, suggesting that a growth in the number of children in the home is associated with declines in perceived instrumental support. Consistent with Hypothesis 3, an increase in the number of children is significantly associated with a decrease in housing and child-care support, the types of support which require direct contact with children.
Changes in the health of the focal child between survey waves are not associated with the perceived instrumental support scale. Deterioration in focal-child health is associated with a significant decline in one type of perceived support: emergency housing. This is consistent with Hypothesis 3, which predicted that characteristics associated with caregiver burden will tend to diminish perceived support.
DISCUSSION
Help from friends and families can be crucial for averting and weathering crises, and simply having confidence that help is available if needed offers peace of mind. Unfortunately, not all mothers anticipate that friends or family will be able to help them in times of need. Among our sample of urban mothers who are raising young children, 16% lack perceived financial or housing support, 12% lack perceived child-care support, and 8% lack perceived emotional support. Using a large sample that represents U.S. families in urban areas, we find that personal disadvantages go hand-in-hand with lacking a personal safety net that can be called on for financial, housing, child-care, or emotional support. We observe this pattern across a wide range of characteristics: Low income, poor health, depression, large family size, and having a child in poor health are all associated with lacking a personal safety net. Our work extends previous research, which has documented associations between low income and less receipt of support (Hogan et al., 1993), by analyzing a wider range of personal disadvantages and shifting the focus to the availability of personal safety nets.
Our research builds on community studies that stressed the importance of reciprocity in the context of scarce resources (Dominguez & Watkins, 2003; Menjivar, 2000; Nelson, 2000). These prior studies focused on small samples of mothers in a poor Boston neighborhood, Salvadoran immigrants in a San Francisco neighborhood, and mothers in a rural community, respectively, and were rich in their ethnographic detail. Our research adds a comparative dimension to this body of research. We demonstrate that poor mothers and mothers with other disadvantages are more likely to lack a personal safety net compared with their more advantaged counterparts. We show that the relationship between personal disadvantages and lacking a personal safety net spans all types of disadvantage and all types of support. The pervasiveness and consistency of the relationships are consistent with our hypothesis, derived from previous ethnographic research, that mothers with personal disadvantages have difficulty maintaining supportive social ties because of challenges in adhering to norms of reciprocity.
Our evidence suggests two theoretical influences that amplify the relationship between personal disadvantages and lacking a personal safety net. First, social networks tend to be homophilous in terms of economic resources, and poor mothers are especially likely to lack financial support. Second, some families are more difficult to support than others, and families with many children or a child in poor health are especially likely to lack housing and child-care support. In contrast, and as we predicted, these families are no more or less likely to have financial support available. The process of providing a loan to a family with many children or a sick child should be no more onerous than providing a loan to a family with one healthy child. Therefore, the pattern of findings across these support types lends credence to our theory that the onus of providing support may be an underappreciated influence in support relationships.
In theory, we might expect social networks to be responsive to the needs of mothers as they arise (Lye, 1996), and some research found just such a relationship when analyzing actual support received (Fingerman et al., 2009; Suitor et al. 2007). Other research suggested that family members were flexible in terms of the timing and type of reciprocation (Silverstein et al., 2002). Nevertheless, our pattern of results suggests that other motives and limiting conditions dominate in shaping whether mothers report having personal safety nets. We find no instances in which personal disadvantages are associated with stronger safety nets, and we typically find just the opposite. Although friends and family may in some cases provide some extra support in times of need, mothers’ personal disadvantages weaken rather than strengthen mothers’ perceptions of personal safety net support.
We found that mothers’ personal disadvantages are more strongly related to instrumental supports than to emotional support, a finding that we did not predict and that to our knowledge this has not been a topic of previous research. This variation between instrumental and emotional support types provides an illustration of the complex, multidimensional nature of perceived support (Sarason et al., 1990). We propose that emotional support may be more easily provided and more easily reciprocated than instrumental support. Another possibility is that norms of reciprocity are stronger and more rigid when it comes to tangible instrumental supports: We speculate that there may be less scorekeeping when it comes to the provision of emotional support; perhaps an expression of gratitude may be considered reciprocation enough.
Our analysis of the Fragile Families data included measures of mothers’ personal disadvantages and perceived supports at three points in time, allowing for fixed-effects analyses. The results from our fixed-effects analyses suggested that after taking into account each mother’s stable characteristics—even those characteristics that cannot be observed directly—a relationship persists between personal disadvantages and lacking support from personal safety nets. The relationships are somewhat weaker in the fixed-effects specification, which rely on variation over time within mothers, but are largely consistent with estimates that include variation across mothers. The attenuation of estimated effects with the addition of fixed-effects suggests that unobserved characteristics play a role in the relationship between personal disadvantages and personal safety nets. Alternatively or in addition, the longer term effects of personal disadvantages may not be fully captured in the fixed-effects specification, which relies on short-term changes across survey waves. It may be tempting to view the fixed-effects specification as superior, but it is plausible that the long-term effects of poverty and other disadvantages may be different, and larger, than the short-term effects. Although we cannot be certain about the true magnitudes of the estimated relationships, we consider our random-effects models to be upper-bound and our fixed-effects models to be lower-bound estimates. Importantly, in almost all cases, both specifications yield a consistent pattern of results and support the same set of inferences.
The study of support relationships is by nature complex. Our data allowed us to make comparisons across several dimensions of support, but other important dimensions were not available. For instance, we lack information on the source of support, and expectations of reciprocity are different for friends versus family members (Lye, 1996; Silverstein et al., 2002). We also have only partial data for two variables: Perceived emotional support was only collected at one wave, so we were not able to include it in the fixed-effects analysis, and child health was only collected for the focal child in each family, so we are not able to capture the full effect of poor child health. Other studies have more comprehensive measures of various types of perceived support, but lack representative samples or longitudinal data. In contrast, our measures of support were limited, but our sample of urban mothers is representative of a large and policy-relevant population.
Our overall finding that personal disadvantages are associated with lacking a personal safety net is consistent across types of disadvantages, types of support, and model specifications. One remaining question, however, is the causal ordering in the relationship between personal disadvantages and personal safety nets. Theory and prior research suggest that bidirectional, self-reinforcing relationships are likely. In particular, personal disadvantages are likely to interfere with support relationships, and personal safety-net deficits, in turn, are likely to reinforce and entrench personal disadvantages.
In principle, public safety-net programs are intended to fill the gap when a mother’s own resources and the resources she can get from friends and family are not enough to stave off serious hardship. Our paper documents a process of cumulative disadvantage: Mothers who appear to have the greatest need for support seem to be the ones least able to maintain supportive social ties. What can public safety-net programs do for these vulnerable mothers? An accessible and comprehensive public safety net can obviously provide a range of direct benefits to these mothers. In theory, public safety-net programs could also provide an important indirect benefit if, by providing an external source of cash and in-kind assistance, these programs helped mothers to regain their good standing in exchange relationships and rekindle supportive social ties. In reality, public policy in the United States has been moving farther away rather than closer to this ideal in recent years. In an era in which welfare receipt is often contingent on work requirements, the very same mothers who lack personal safety nets are likely to have difficulty meeting the welfare systems’ quid pro quo.
Contributor Information
Kristen S. Harknett, Department of Sociology, 271 McNeil, 3718 Locust Walk, University of Pennsylvania, Philadelphia, PA 19104 (harknett@sas.upenn.edu)
Caroline Sten Hartnett, Departments of Sociology and Demography, 239 McNeil, 3718 Locust Walk, University of Pennsylvania, Philadelphia, PA 19104 (csten@pop.upenn.edu).
REFERENCES
- Benin M, Keith VM. The social support of employed African-American and Anglo mothers. Journal of Family Issues. 1995;16:275–297. [Google Scholar]
- Berkman LF, Glass T, Brissette I, Seeman TE. From social integration to health: Durkheim in the new millennium. Social Science & Medicine. 2000;51:843–857. doi: 10.1016/s0277-9536(00)00065-4. [DOI] [PubMed] [Google Scholar]
- Bird CE. Gender differences in the social and economic burdens of parenting and psychological distress. Journal of Marriage and the Family. 1997;59:809–823. [Google Scholar]
- Bradley E. Religious involvement and social resources: Evidence from the data set“Americans’ Changing Lives”. Journal for the Scientific Study of Religion. 1995;34:259–267. [Google Scholar]
- Cooney TM, Uhlenberg P. Support from parents over the life course: The adult child’s perspective. Social Forces. 1992;71:63–84. [Google Scholar]
- Dean A, Ensel WM. Modelling social support, life events, competence, and depression in the context of age and sex. Journal of Community Psychology. 2006;10:392–408. [Google Scholar]
- Dominguez S, Watkins C. Creating networks for survival and mobility: Social capital among African-American and Latin-American low-income mothers. Social Problems. 2003;50:111–135. [Google Scholar]
- Durkheim E. In: Suicide: A study in sociology. Simpson G, Spaulding JA, editors; Simpson G, translator. New York: Free Press; [Google Scholar]
- Eggebeen DJ. Family structure and intergenerational exchanges. Research on Aging. 1992;14:427–447. [Google Scholar]
- Ellison CG, George LK. Religious involvement, social ties, and social support in a southeastern community. Journal for the Scientific Study of Religion. 1994;33:46–61. [Google Scholar]
- Fingerman K, Miller L, Birditt K, Zarit S. Giving to the good and the needy: Parental support of grown children. Journal of Marriage and the Family. 2009;71:1220–1233. doi: 10.1111/j.1741-3737.2009.00665.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Fischer CS. To dwell among friends: Personal networks in town and city. University of Chicago Press; 1982. [Google Scholar]
- Goldsteen K, Ross CE. The perceived burden of children. Journal of Family Issues. 1989;10:504–526. doi: 10.1177/019251389010004005. [DOI] [PubMed] [Google Scholar]
- Harknett K. The relationship between private safety nets and economic outcomes among single mothers. Journal of Marriage and Family. 2006;68:172–191. [Google Scholar]
- Harknett K, Knab JT. More kin, less support: Multipartnered fertility and kin support among new mothers. Journal of Marriage and Family. 2007;69:237–253. [Google Scholar]
- Henly JR, Danziger SK, Offer S. The contribution of social support to the material well-being of low-income families. Journal of Marriage and Family. 2005;67:122–140. [Google Scholar]
- Hogan DP, Eggebeen DJ, Clogg C. The structure of intergenerational exchanges in American families. American Journal of Sociology. 1993;98:1428–1458. [Google Scholar]
- House JS, Umberson D, Landis KR. Structures and processes of social support. Annual Review of Sociology. 1988;14:293–318. [Google Scholar]
- Ishii-Kuntz M, Seccombe K. The impact of children upon social support networks throughout the life course. Journal of Marriage and the Family. 1989;51:777–790. [Google Scholar]
- Kawachi I, Berkman LF. Social ties and mental health. Journal of Urban Health. 2001;78:458–467. doi: 10.1093/jurban/78.3.458. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kessler RC, Andrews G, Mroczek D, Ustun B, Wittchen H. The World Health Organization Composite International Diagnostic Interview Short-Form (CIDI–SF) International Journal of Methods in Psychiatric Research. 1998;7:171–185. [Google Scholar]
- Link BG, Phelan JC, Bresnahan M, Stueve A, Pescosolido BA. Public conceptions of mental illness: Labels, causes, dangerousness, and social distance. American Journal of Public Health. 1999;89:1328–1333. doi: 10.2105/ajph.89.9.1328. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Luo M. “Doubling up” in recession-strained quarters. New York Times. 2010 December 29;:A1. [Google Scholar]
- Lye DN. Adult child–parent relationships. Annual Review of Sociology. 1996;22:79–102. [Google Scholar]
- Mauldon J. Children’s risks of experiencing divorce and remarriage: Do disabled children destabilize marriages? Population Studies. 1992;46:349–362. [Google Scholar]
- McPherson M, Smith-Lovin L, Cook JM. Birds of a feather: Homophily in social networks. Annual Review of Sociology. 2001;27:415–444. [Google Scholar]
- Menjivar C. Fragmented ties: Salvadoran immigrant networks in America. Berkeley: University of California Press; 2000. [Google Scholar]
- Nelson MK. Single mothers and social support: The commitment to, retreat from, reciprocity. Qualitative Sociology. 2000;23:291–317. [Google Scholar]
- Reichman N, Corman H, Noonan K. Child health and parents’ relationship status. Demography. 2004;41:569–584. doi: 10.1353/dem.2004.0026. [DOI] [PubMed] [Google Scholar]
- Ryan RM, Kalil A, Leininger L. Low-income mothers’ private safety nets and children’s socioemotional well-being. Journal of Marriage and Family. 2009;71:278–297. [Google Scholar]
- Sarason BR, Sarason IG, Pierce GR. Traditional views on social support and their impact on assessment. In: Sarason BR, Sarason IG, Pierce GR, editors. Social support: An interactional view. New York: John Wiley and Sons; 1990. [Google Scholar]
- Sarkisian N, Gerstel N. Till marriage do us part: Adult children’s relationships with their parents. Journal of Marriage and Family. 2008;70:360–376. [Google Scholar]
- Silverstein M, Conroy SJ, Wang HT, Giarrusso R, Bengtson VL. Reciprocity in parent–child relations over the adult life course. Journals of Gerontology Series B: Psychological Sciences and Social Sciences. 2002;57:S3–S13. doi: 10.1093/geronb/57.1.s3. [DOI] [PubMed] [Google Scholar]
- Suitor JJ, Sechrist J, Pillemer K. Within-family differences in mothers’ support to adult children in Black and White families. Research on Aging. 2007;29:410–435. [Google Scholar]
- Swartz TT. Intergenerational family relations in adulthood: Patterns, variations, and implications in the contemporary United States. Annual Review of Sociology. 2009;35:191–212. [Google Scholar]
- Thoits PA. Stress, coping, and social support processes: Where are we? What next? Journal of Health and Social Behavior. 1995;35:53–79. [Extra issue] [PubMed] [Google Scholar]
- Turner RJ, Marino F. Social support and social structure: A descriptive epidemiology. Journal of Health and Social Behavior. 1994;35:193–212. [PubMed] [Google Scholar]
- Turney K, Kao G. Assessing the private safety net: Social support among minority immigrant parents. Sociological Quarterly. 2009;50:666–692. [Google Scholar]
- Turney K, Harknett K. Neighborhood socioeconomic disadvantage and perceptions of social support among new parents. Journal of Family Issues. 2010;3:499–524. doi: 10.1177/0192513X09347992. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wethington E, Kessler RC. Perceived support, received support, and adjustment to stressful life events. Journal of Health and Social Behavior. 1986;27:78–89. [PubMed] [Google Scholar]
- White L. The effect of parental divorce and memories of family problems on relationships between adult children and their parents. Journal of Family Issues. 1992;13:234–250. doi: 10.1093/geronb/50b.1.s24. [DOI] [PubMed] [Google Scholar]
- Winship C, Radbill S. Sampling weights and regression analysis. Sociological Methods & Research. 1994;23:230–257. [Google Scholar]