Child support and mixed-status families an analysis using the Fragile Families and Child Wellbeing Study

Abstract

A large body of literature documents the importance of child support for children’s wellbeing, though little is known about the child support behaviors of mixed-status families, a large and rapidly growing population in the United States. In this paper, we use data from the Fragile Families and Child Wellbeing Study to investigate the impact of citizenship status on formal and informal child support transfers among a nationally representative sample of parents who have citizen children. Probit regression models and propensity score matching (PSM) estimators show that mixed-status families are significantly less likely to have child support orders and child support receipt compared to their citizen counterparts. We found that mothers’ knowledge of the child support system increases the probability of establishing paternity. However, cultural differences in knowledge of and perception about the U.S. child support system between mixed-status families and citizen families do not have an impact on the probability of getting a child support order, child support receipt, or in-kind child support. Rather, institutional factors such as collaborations between welfare agencies and child support enforcement agencies as well as state child support enforcement efforts have a significant impact on formal child support outcomes. The results are robust against different model specifications, measure constructions, and use of datasets. These findings have important policy implications for policy makers and researchers interested in reducing child poverty in complex family structures and underscore the need to revisit child support policies for mixed-status families.

I. Introduction

Children of immigrant parents are the fastest growing segment of the nation’s child population, accounting for 77 percent of the increase of children born in the United States between 1990 and 2007 (Fortuny and Chaudry 2009). As of 2009, the number of immigrant youth—defined as children under the age of 18 who are either foreign born or U.S. born to immigrant parents—was 17.3 million, or 23.2 percent of all children in the United States (Passel and Cohn 2011). The vast majority (93 percent) of children of immigrants are U.S. citizens, mostly by virtue of having been born in the United States (Fortuny and Chaudry 2009). As a result, the number of mixed-status families in which at least one parent is a noncitizen and at least one child is a citizen is surprisingly large. According to estimates by the Urban Institute, 32 percent of children of immigrants in 2007 lived in mixed-status families where the children were U.S. citizens and their parents were not.¹

With an estimated 5.6 million American citizen children with noncitizen parents,² the wellbeing of children in these families requires serious considerations by policy makers, service providers, and researchers interested in reducing poverty in complex family structures. American children living in mixed-status families are more likely to be impoverished. The Urban Institute estimates that children of immigrants are approximately 40 percent more likely to live in families that are poor, and nearly 70 percent more likely to live in low-income families with working parents (Chaudry and Fortuny 2010).

Prior literature has found that living in the United States can erode the family orientation of immigrants as reflected in increases in the rates of both divorce and births outside of marriage (Wu and Wolfe 2001, Tienda, Mitchell et al. 2006). Previous literature has also shown that a major factor in children’s impoverishment is the failure of non-custodial parents to provide child support (Nichols-Casebolt 1986, Freeman and Waldfogel 2001, Zedlewski, Giannarelli et al. 2010). It is thus prudent to examine citizenship status differentials in child support outcomes for children of low-income parents.

Nepomnyaschy & Donnelly (2014) is the first study that specifically examines the child support outcomes of immigrant families. The authors used the U.S. Current Population Survey – Child Support Supplement (CPS-CSS) data and compared the child support outcomes of foreign-born and native-born mothers. They found that foreign-born mothers are much less likely to have a formal child support agreement than native-born mothers, but they do not differ on the likelihood of receiving in-kind support, or on the amount of formal or informal child support received. In addition, Nepomnyashy & Donnelly identified nonresident fathers’ residence outside the United States as an important mechanism through which nativity affects the likelihood of having a child support order and receiving any in-kind support. While this article contributes important knowledge to our understanding of the economic circumstances of children of immigrants, there remain unanswered questions we attempt to address. For example, the CPS-CSS data they use are based on mothers’ self-reported child support outcomes, and do not include any information about the nonresident father. The authors used mothers’ characteristics as proxies for fathers’ characteristics; however, the extent to which assortative mating assumptions hold for immigrant groups is unknown. In addition, although the authors acknowledge the importance of child support enforcement and participation in social welfare programs such as TANF and SNAP on child support outcomes, their statistical models did not take these policy variables into consideration.

Our study improves upon previous child support studies in five important ways. First, we use the Fragile Families and Child Wellbeing Survey (FFCWS) as our main data source, which provides information on both mothers and fathers. The FFCWS provides a unique opportunity to measure the impact of fathers’ characteristics on child support outcomes. Second, we specifically focus on a subpopulation of the immigrant families, the fragile families where immigrant families are doubly disadvantaged by their citizenship status and their economic status. This is a particularly vulnerable population that deserves policy attention. Third, the study examines both individual-level and institutional-level determinants of child support outcomes. In particular, this study is one of only a few that quantify the impact of the anti-immigrant policy climate on noncitizens’ use of child support. We create proxies of such “chilling effects” by measuring the effect of states’ welfare generosity towards noncitizens and of public opinion on increasing federal spending to combat undocumented immigrants. Fourth, the study makes a methodological contribution by carefully considering the possibility of self-selection in the migration process and its bias associated with observed and unobserved heterogeneity among low-income immigrants. To this end, we employ a rich array of control variables to reduce omitted variables bias, including parents’ demographic and socio-economic characteristics, mother’s knowledge of and perceptions about the child support system, and state-level policies on child support and immigration. In addition, we use propensity score matching (PSM) to test and partially correct for selection bias. Fifth, we benchmark our analysis against the findings obtained using the Current Population Survey – Child Support Supplement (CPS-CSS) to test the sensitivity of our results against different model specifications and alternative datasets.

The findings from this study inform research on the links between parents’ citizenship status and child support outcomes in a sample of children born to “fragile families” (socially and economically disadvantaged families), an important policy target group. The results underscore the need to revisit child support policies for mixed-status families and the importance of designing effective programs and policies to meet the needs of this special population.

II. Background Information

In 2011, nearly 41 percent of all births in the United States in 2009 occurred outside marriage, up from 33.2 percent in 2000 and 18.4 percent in 1980 (Martin et al. 2011). The proportions in 2009 were even higher among ethnic minority populations: 53 percent for Hispanics and 73 percent for non-Hispanic blacks. At the same time, married couples face a high probability of marriage dissolution. It is estimated that between 40 and 60 percent of new marriages will eventually end in divorce (Williams, Sawyer et al. 2006). As a result, about half of all American children will spend some time in their life living apart from one of their parents by the age of 15 (Andersson 2002).

The rise in out-of-wedlock childbearing and the increase in single parenthood are major causes of high levels of child poverty (Bartfeld and Meyer 2001). The poverty rate among households headed by a single mother with one or more children under the age of 18 was 40.7 percent in 2010, while the comparable poverty rate for married couple households was only 8.8 percent (Seefeldt, Abner et al. 2012). Both poverty and absence of the father have been shown to have a significant negative effect on children’s future life chances, from lower educational achievement and behavioral problems to lower earnings in the labor market (McLanahan, Seltzer et al. 1994, Brooks-Gunn and Duncan 1997, Bronte-Tinkew, Moore et al. 2006).

Child support enforcement is one of the major tools that policy makers use to tackle the growing economic and social problems associated with nonmarital childbearing and poverty. The Child Support Enforcement (CSE) program, under Title IV-D of the Social Security Act (1975) tries to improve child well-being by establishing paternity, setting medical and financial support orders, and enforcing those orders through a federal, state, and local partnership. Improved child support enforcement is also a key component of the Personal Responsibility and Work Opportunity Reconciliation Act of 1996. In the law, Congress underscored the importance of family self-sufficiency and replaced an open-ended entitlement with Temporary Assistance for Needy Families (TANF), a time-limited benefit that has strict work requirements. The intent of these reforms was to increase the role of work and child support so that poor families do not need to rely on public cash assistance.

As a result, the relative importance of child support and earnings has increased since 1996. Cash assistance now constitutes a substantially smaller share of the income of poor custodial families. Child support represents, on average, 10 percent of poor custodial families’ income and 40 percent of income for poor custodial families who receive it (Sorensen 2010) . The child support program serves 17 million children, which represents nearly one in four children in the United States. Among publicly funded social services, only the Medicaid program serves more children (Sorensen 2010).

Despite the relative size and importance of child support enforcement, studies of the underlying demographic and socioeconomic characteristics of the individuals the program serves are very limited. In particular, there is no study that looks at mixed-status families and their child support outcomes. In this paper we use the FFCWS to study how mixed-status families differ from citizen families in their child support behaviors as well as to examine the determinants of child support outcomes for various groups with different legal status. We also use the CPS-CSS data to benchmark our main results against the child support outcomes of noncitizen custodial parents.

III. Literature Review

There is an extensive literature on child support because of its importance in modern American society and its use as a policy tool to combat child poverty. We focused on the different child support outcomes, the determinants of child support outcomes, and the role of legal status in child support to help guide our theoretical framework.

3.1. Child Support Outcomes

Lack of child support is often portrayed in the popular press as a compliance problem—that is, a problem stemming from failure to pay support obligations. The reality, however, is more complicated. In the case of nonmarital children, there are three key steps to obtain formal child support: a legal father must be identified, a support order must be issued, and support must be collected. Some parents may not wish to seek legal or government intervention, and the fathers elect to buy their children food, clothes, toys, medicines, school supplies, etc., which is often called “in-kind” child support. It is important to understand whether low-income families, especially mixed-status families, can navigate the child support system and what policies might help them to successfully obtain support.

To be eligible for child support, children born outside of marriage must first have the father legally identified through the process of establishing paternity. Aiming to stem the growth of the welfare rolls and recoup the cost of public benefits, Congress passed a series of provisions designed to promote the establishment of paternity, including in-hospital paternity programs, mandating genetic testing in disputed cases, and allowing the establishment of paternity up until a child’s 18th birthday. While rates of paternity establishment have increased to 97–99 percent in recent years,³ the rates ranged from 82 to 96 percent in 2003.⁴ The expansion of in-hospital paternity establishment has been found to be particularly effective in increasing paternity establishment rates (Sorensen and Hill 2004, Mincy, Garfinkel et al. 2005).

Having paternity established is a necessary but insufficient precursor to obtaining a support obligation for nonmarital children. Custodial parents without a child support order include those who have not established paternity, as well as those who have established paternity established but do not have orders. The most recent data from the Current Population Survey indicate that only about half (54.9 percent) of custodial mothers have child support orders, with lower rates among never-married mothers (44.2 percent) (Grall, 2001). The reasons most often cited for not establishing a formal legal agreement are the following: the other parent provided what he or she could for child support (34.4 percent); the custodial parent did not feel the need to go to court or get legal agreements (32.1 percent); or the other parent could not afford to pay (29.2 percent) (Grall 2001).

Enforcement of support orders has also been a target of federal and state policy. Emphasis has been placed on standardizing the collection process, such that support obligations are automatically withheld from parents’ paychecks. Other major policy changes intended to strengthen the enforcement system include provisions for intercepting tax refunds to collect overdue child support; placing liens on property; allowing criminal penalties for nonpayment of support; and, most recently, creating a centralized database to which employers report information about new hires, as a means of locating parents who have evaded the enforcement system. State-level child support enforcement efforts thus play an important role in paternity establishment, child support orders, and child support compliance rates (Garfinkel, Miller et al. 1998, Huang 2009). In addition, to recoup part of the costs of public benefits, Congress requires the recipients of public assistance to cooperate with efforts to establish paternity. Therefore, the probability of having a child support order is higher for mothers who received TANF or Food Stamps before the child’s birth (Mincy, Garfinkel et al. 2005).

When a child support order is set, the amount of the order is automatically withheld from the earnings of the noncustodial parent and sent by the employer to a centralized processing agency. Employers are also required to submit the Social Security numbers of new employees to the child support agency. In this way, the child support agency tracks parents’ employment changes and can begin a new withholding order. The administrative tracking systems do not work seamlessly, however. Researchers have noted that child support policies may create a disincentive among low-income parents to avoid the formal child support system and instead opt for underground employment and “under-the-table” or in-kind child support (Edin 1995, Johnson, Levine et al. 1999, Waller and Plotnick 2001, Roff 2008). This may be especially true for many noncitizens without proper documentation who live in the shadows to avoid deportation. Child support enforcement methods such as automatic wage withholding will not work well for a large proportion of such low-income mixed-status families.

Although there is a significant body of research on the patterns and correlates of formal child support, substantially less is known about informal support, which includes both in-kind contributions and cash contributions outside of the formal child support system. Informal support, especially the purchase of goods and services for the child, appears to be very common (Edin and Lein 1997, McLanahan, Garfinkel et al. 2001). Informal support is of particular interest because it reflects parents’ voluntary contributions to their children. Nepomnyaschy, Magnuson et al. (2012) found that low-income fathers’ provision of informal cash support (but not formal support) is associated with higher cognitive scores of children. A possible explanation is that fathers who voluntarily provide financial assistance on a regular basis are often more involved in their children’s lives and provide support in nonfinancial ways as well. Increasing our understanding of whether and how informal support is provided may provide important information on the economic wellbeing of vulnerable children.

3.2. Determinants of Child Support Outcomes

Previous literature has shown that many factors have an impact on a family’s child support outcomes, including the mother’s ability to navigate the child support system, the father’s ability and willingness to pay child support, and child support enforcement efforts by states.

A mother’s ability to navigate the child support system is important because the child support enforcement program is complex. Securing a child support order and obtaining child support payments may involve a substantial amount of time and perseverance on the part of custodial mothers. For this reason alone, the take-up rate of child support services by immigrant mothers may well be less than the rates for other similarly situated women who are citizens. This is likely to be true if immigrants use English as a second language, fear authorities, have limited knowledge of the child support enforcement program, or are from countries with different expectations about whether women should advocate for themselves (Rios-Salas 2014). Socio-demographic variables such as age, race, marital status, educational attainment, and employment status are often used as proxies for the mother’s ability to navigate the child support system, or as proxies for the father’s ability to provide child support. For instance, Beller and Graham (1996) and Hanson et colleagues (1996) documented a higher likelihood of support orders among better-educated mothers and among whites. Using pooled Current Population Survey data through 1992, Miller and Garfinkel (1999) found that orders are more common among older mothers and mothers with more children. In addition, Teachman (1990) found that mother’s higher income at the time of divorce increases the likelihood that she will be awarded child support.

A father’s ability and willingness to pay are also hypothesized to be strong indicators of child support outcomes. A substantial body of research has confirmed that fathers’ higher earnings or income (or frequently proxies for such) are associated with higher child support compliance (Sonenstein and Calhoun 1990, Beller and Graham 1996, Bartfeld and Meyer 2001). Enforcement tools, such as the centralized database for new hires, routine income withholding and interceptions of tax returns in the case of child support arrearages, have made the child support system more stringent and automated, especially for fathers in the formal labor market.

However, child support payments from low income fathers, especially those who are not in the formal labor market, depend not only on their income and ability to pay but also on their willingness to pay. Willingness to pay is often linked to the strength of fathers’ ties with the children and mother, the level of economic need of mothers and children, and the perceived fairness of the support obligation (Meyer and Bartfeld 1996, Lin 2000, Bartfeld and Meyer 2003). Graham and Beller (2002) modeled child support payments as a classic prisoner’s dilemma game and predicted a noncooperative (low-spending) equilibrium because of parents’ mistrust. A non-custodial father’s willingness to pay is expected to boost child support payments independent of income (Nepomnyaschy and Garfinkel, 2010).

Recognizing the discrepancies between non-custodial parents’ ability to pay and willingness to pay, child support policies have tried to assess the non-custodial parent’s ability to pay and attempt to enforce child support payments regardless of the willingness to pay (Sorensen and Hill 2004, Sinkewicz and Garfinkel 2009 ). State child support enforcement thus plays a key role in how much child support a father pays. More stringent child support enforcement policies are generally positively associated with formal child support payments (Freeman and Waldfogel 2001, Sorensen and Hill 2004). For example, Miller and Garfinkel (1999) found that never-married mothers in states with higher Office of Child Support Enforcement (OCSE) expenditures per female-headed family and more efficient collection systems have higher paternity establishment rates. Similarly, Argys et al. (2001) found that paternity is significantly more likely to be established in states with stronger genetic-testing requirements, and the likelihood of a child-support award is significantly increased in states with higher and more effective child-support enforcement expenditures. Freeman and Waldfogel (2001) constructed a child support enforcement index and found that never-married mothers are more likely to receive child support if they live in a state that has more stringent child support enforcement laws and whose OCSE programs spend more per case on enforcement.

In addition, a mother’s use of public benefits might increase her probability of getting awards, because welfare programs require mothers to sign over their rights to child support to the state (Beller and Graham 1996). Even if a mother did not want child support income from the father, state efforts to recover the costs of public benefits would remove her discretion. TANF and SNAP participants, for example, are required to cooperate with the state OCSE’s attempts to locate nonresident parents, establish paternity and orders, and collect support. Thus we hypothesize that mothers on TANF or SNAP would be more likely to receive formal child support.

Mixed-status families face greater barriers than citizen families to accessing means-tested welfare programs. For immigrants, the 1996 Personal Responsibility and Work Reconciliation Act (PRWORA) went well beyond conditioning access to cash benefits on work. In fact, the law made citizenship more central to the receipt of social benefits and granted state governments the power to determine immigrants’ eligibility for benefits. PRWORA barred TANF for at least five years for new legal permanent residents entering the United States on or after August 22, 1996, and increased requirements for the TANF agency to report to the Immigration and Naturalization Service persons whom the agency knows are not lawfully present in the United States (Broder and Director 2008). Consequently, participation in TANF has declined more sharply for immigrants than for non-immigrants (Borjas and Hilton 1996, Fix and Passel 1999). Therefore, we expect that immigrants are less likely to enter the child support system through TANF than are citizens. However, the mixed-status families that are already receiving TANF benefits are subject to the same program requirements to collaborate with child support agencies. Thus we expect that mixed-status families on TANF would be more likely to receive formal child support compared to citizen families.

However, states’ child support enforcement efforts and the collaboration between child support agencies and welfare agencies might have a negative effect on informal child support. Because formal child support paid to women currently or previously on TANF or SNAP is kept by the state to recoup welfare costs, low-income fathers and mothers may both prefer informal support to formal support (Roberts and Vinson 2004). In addition, regardless of mothers’ welfare status, fathers may prefer informal support because it may increase their ability to make decisions about child rearing and to monitor how their money is spent (Waller and Plotnick 2001). Empirical research using the FFCWS data shows that strong enforcement increases the amount of formal support but reduces the amount of informal support.

The Role of Legal Status in Child Support

Previous studies have shown that parents’ characteristics and parental relationships play important roles in child support outcomes. Willis (1999) argued that paternity establishment depends on the father’s preference parameters and the mother’s expenditure on the child. Therefore, the father’s willingness to contribute will be positively related to his income and negatively related to the mother’s income. In addition, a father is more likely to be altruistic towards his children if he has a close relationship with the mother, as evidenced by cohabitation, subsequent marriage, or multiple children together (Seltzer, Schaeffer et al. 1989, Castillo 2009).

For some mixed-status families, the choice to engage the child support enforcement system likely requires consideration of additional factors. Some custodial parents may balance the prospects of child support receipt with the possibility of losing their chances to become naturalized citizens. To establish legal paternity, parents need to either sign the Acknowledgement of Paternity form in the hospital in front of a notary public or other officers, or have the court do genetic testing to determine the biological father. In cases involving undocumented immigrants, coming into contact with officials might expose them to deportation risks and the risk of losing custody of their newborn child during the deportation process (Williams 2011). Furthermore, many legal immigrants fear that they might be perceived as a “public charge”⁵ if they utilize public services and that, in turn, might jeopardize their ability to adjust their status toward becoming citizens. Recent research suggests that public charge concerns, along with other “chilling effects” related to welfare reform and confusion about eligibility rules for benefits, have kept many legal immigrants from accessing benefits for which they are eligible (Fix and Passel 1999, Van Hook 2003, Kaushal 2005). However, there is little empirical evidence about how the “chilling effect” has worked to reduce immigrants’ willingness and ability to obtain governmental support. Watson (2010) and Vargas (2016) represent the first attempts to link immigrants’ program participation with the patterns of immigration enforcement in a particular state. We build upon their work and examine whether the anti-immigrant policy environment of a state, measured by a state’s welfare generosity towards immigrants and public opinion regarding illegal immigration influence a mixed-status family’s likelihood of getting child support.

In addition, the majority of immigrants in the United States come from countries in which child support issues receive less policy attention, and enforcement is relatively weak. Thus, cultural attitudes and perceptions about the financial responsibility of child rearing may prevent immigrant mothers from pursuing child support regardless of their citizenship status. However, Cuesta and Meyer (2012) found that custodial mothers in Colombia and the United States have similar issues when it comes to receiving financial support from their children’s father, despite the differences in economic development and in attention to child support policy in the two countries. It is therefore an empirical question whether culture and perceptions about the child support system would have an impact on the child support outcomes of mixed-status families in the United States

Although no current research focuses on the child support outcomes of mixed-status families, some studies have focused on racial and ethnic differences in child support outcomes. Beller and Graham (1986, 1996) found that single minority mothers are less likely to have a child support award or payment, even after controlling for characteristics that reflect the mother’s need for child support payments and the father’s ability to pay. Using the Fragile Families data, Mincy and Nepomnyaschy (2005) found no statistically significant difference between the award probabilities of black and white unmarried parents. However, among those with awards, blacks were far less likely to comply (make payments) than were whites. The authors also observed lower award probabilities among Hispanic fathers, but no difference in compliance rates between Hispanics and whites. In addition, Mincy and Nepomnyaschy (2005) included mother’s nativity as a control variable and found that U.S.-born mothers are significantly more likely than foreign-born mothers to have a child support order. We contribute to this literature by examining the mechanisms through which U.S.-born children of mixed-status families experience child support outcomes, and how legal status influences those outcomes.

IV. Data and Methods

4.1. Data

The main source of data for this analysis is the Fragile Families and Child Wellbeing Study (FFCWS), a longitudinal survey of 4,898 children (3,710 of whom were born to unwed parents) born in 20 large U.S. cities between 1998 and 2000. The survey interviewed the children’s biological mothers and fathers at the time of the child’s birth and approximately 1, 3, 5, and 9 years after the birth. The response rate for mothers for the baseline survey was 86 percent and for fathers the response rate was 78 percent.⁶

The FFCWS oversamples low-income unmarried parents and gathers data on both formal and informal child support agreements and transfer amounts in all follow-up surveys. While the FFCWS has detailed information on the type of child support arrangements, there are some limitations in using these data to answer our research questions. Mothers may underestimate the child support payments they receive, and fathers may overestimate their child support transfers. To what extent either parent misestimates the child support payment is debatable; however, we rely on the custodial mothers’ reports since these are less likely to have nonresponse bias relative to nonresident fathers’ reports (Nepomnyaschy and Garfinkel 2010).

We focus on child support outcomes at 3 years, because child support receipt for unmarried mothers stabilizes by the time children are 3 years old (Bartfeld and Meyer 2001). The FFCWB data include 4,898 births at baseline. However, 667 respondents dropped out by the 3-year follow up. Mothers who report being married to the father at the 3-year follow-up are excluded from our analysis sample. This includes 1239 cases in total: 1,014 women were married at the baseline and remained married at the 3-year follow-up, while 225 women who were not married at the baseline got married before the 3-year follow-up. These individuals were not asked any child support questions and are thus not included in our analysis. We dropped an additional 99 cases in which the fathers were unknown, deceased, or had primary custody of the focal child. Our final analysis sample consists of all 2,893 mothers for whom child support information was available. We call this sample the child support eligible sample. However, 240 mothers in the child support eligible sample were living all or most of the time with the focal child’s father when the child was 3 years old. While many of these children would have paternity acknowledged, it would be relatively unusual for a child support agency to pursue child support, and very unusual for the mother to pursue child support, from a resident partner. In fact, these 240 mothers were not asked questions about a child support order, child support receipt, or in-kind child support, although 86 percent of them had paternity established. Therefore, the mothers who were not married but were living together with the fathers are included in the analysis of paternity, but excluded from the analysis of the other three outcomes. We call the smaller sample the child support plausible group. Table 1 details the steps of the sample selection criteria we used to generate the analysis sample.

Table 1.

Process of Sample Selection using the FFCWS Data

Full Sample	4,898
Not interviewed at 3- year follow-up	667	Dropped out of survey. Those dropped out between waves are on average more disadvantaged (less education, less household income, and less likely to be married to baby’s father at baseline survey, etc.)
Father deceased, unknown or has primary custody of child	99	Not asked paternity/child support questions
Mother and father married at the 3-year follow-up	1,239	Not asked paternity/child support questions. 1,014 mothers were married at the time of the child’s birth and stayed married at the 3-year follow-up; 225 mothers were not married at the child’s birth but got married by the 3-year follow up
Final Sample	2,893	Child support eligible sample
Sub-sample	2,343	Child support plausible sample (excluded those whose parents were living all or most of the time together at the 3-year follow up)

Since the FFCWS oversamples low-income fragile families in major U.S. cities, we conducted sensitivity analysis using the 2001 Current Population Survey – Child Support Supplement (CPS-CSS) to address concerns regarding the representativeness of the FFCWS data of the child support eligible population. The CPS-CSS is a large, nationally representative household survey conducted in April of every other year, and data are matched to the Annual Demographic Supplement representing data from March of that year. The survey identifies households with absent parents and provides data on child support arrangements, visitation rights of an absent parent, amount and frequency of actual versus awarded child support, and health insurance coverage. However, the survey does not provide information on the characteristics of the noncustodial parents and thus makes the study of mixed-status families impossible. We therefore examine instead the impact of custodial parents’ citizenship on child support outcomes using the CPS-CSS data as a benchmark for our study of mixed-status families using the FFCSW. The variable PRSELIG designates whether a parent was eligible to be asked the questions on the CSS; that is, they are a parent with a biological, adopted, or stepchild under age 21 living in the household and that child has a biological or adopted parent living outside of the household. There are 4,934 child support eligible respondents in the 2001 survey.

4.2. Measurements

a. Child Support Outcomes

Despite the wealth of child support information collected by the FFCWS, there are some limitations associated with these survey data. First, like most population-based samples, the reliability of self-reported data is a concern. Mothers may report only if they actually received child support, not whether it was paid, because in most states mothers who are receiving TANF or SNAP do not receive anything due to the recovery policy. Moreover, some respondents might have limited knowledge of the child support system and be confused by legal terms. For example, 67 mothers in the FFCWB study reported that they had a child support order, but that paternity was never established. This is not possible within the child support enforcement program.

We tried three different coding methods to test the robustness of our models to the treatment of these cases: (1) treating everyone who reported having a child support order as having paternity established; (2) treating everyone who reported having no paternity established as having no child support order; and (3) deleting those 67 observations from our sample as unreliable data. We estimated the models using all three coding schemes and found that the results were not sensitive to how these observations were coded. We report only the results using the first coding method, in which we treat everyone who reported having a child support order as having paternity established. The results using the other two coding methods are available upon request.

A second limitation is that many observations on formal or informal child support payments are missing: Parents typically say whether payments have been made, but the payment amounts are often not reported. Therefore, it is not feasible to use the actual amount of child support received as a dependent variable. Instead, we use binary variables for outcome variables, indicating: (1) whether paternity is established, (2) whether a child support order is established, (3) whether any child support payment is received, and (4) whether any in-kind support is provided.

The CPS-CSS asks similar questions about child support orders, child support payments, and in-kind child support. However, there are no paternity establishment questions. In the sensitivity analysis, which used the CPS-CSS data, we therefore focused our analysis on the other three child support outcomes.

b. Parents’ Citizenship Status

Since all focal children in the FFCWS are born in the United States and are thereby automatically U.S. citizens, we define mixed-status families as families with at least one parent who is foreign-born and not a U.S. citizen. We imputed citizenship status for respondents in Oakland and Austin because the citizenship question was not asked in these two FFCWS cities. We made the assumption that all U.S.-born respondents are U.S. citizens. For the non–U.S.-born respondents, we used data in the remaining 18 cities to build a logistic model to predict whether the non U.S.-born respondents are naturalized citizens based on their socio-demographic characteristics, including age, education, income, marital status, country of origin, current state of residence, year of entry into the United States, and language used to complete the interview. The pseudo R-squared of the logistic model for naturalized citizen mothers is 0.33, indicating a good model fit. We then used this model to predict the citizenship status of the non U.S.-born respondents in the two cities. There are 86 non U.S.-born respondents (or 27 percent of all non-U.S.-born respondents in the analysis sample) whose citizenship status is imputed. Complete model specification and prediction results are available upon request.

After the imputation, there were 319 mixed-status families in the FFCWS sample (11%), 1,938 families where both parents are U.S. citizens (67%), and 636 families for which citizenship information is missing (22%). Fifty-seven percent of all noncitizen mothers and 56 percent of all noncitizen fathers are from Mexico, while 7 percent of all noncitizen mothers and 9 percent of all noncitizen fathers are from Central America/the Caribbean.

Due to the small sample size of the noncitizen parents in the FFCWS, we were not able to separately test the impact of the citizenship of mothers and fathers on child support outcomes. We therefore defined our key independent variable as mixed-status families, which include families with only noncitizen mothers, only noncitizen fathers, and both noncitizen mothers and fathers, in comparison to citizen families where both parents are U.S. citizens.

To test the robustness of our analysis, we also ran the models with the FFCWB data using only mother’s citizenship as the indicator for the family’s immigration status. In addition, we ran the models with noncitizen mothers using data from the CPS-CSS. It is not possible to link the data for custodial and noncustodial parents in CPS. In these models, we use only covariates that are the same or similar across the two surveys, including mother’s citizenship; mothers on TANF, noncitizen mothers on TANF, mothers on SNAP, and noncitizen mothers on SNAP; mother’s age, race, education level and employment status; household income; child support enforcement index; state generosity toward noncitizens, and anti-immigrant sentiment based on American National Election Studies (ANES) data.

c. TANF Participation and SNAP Receipt

Because of welfare agencies’ emphasis on recouping welfare expenses through child support, we expect that custodial parents’ participation in welfare programs to be a key factor for their child support outcomes. Both the FFCWS and the CPS-CSS ask whether the respondent received any TANF or SNAP benefits within the past 12 months. We constructed a variable for TANF participation and a variable for SNAP receipt based on these questions, and created an interaction term between these variables and immigration status to specifically test whether immigrants participating in these welfare programs have different child support outcomes.

d. Child Support Enforcement Index

Based on the literature, we characterize the strength of the child support system in the different states using the CSE index constructed by Huang, Garfinkel, and Waldfogel (2004). This index is the normalized sum of a legislative index and three CPS-based measures of CSE. The legislative index includes laws covering genetic testing, paternity establishment, numerical guidelines, presumptive guidelines, wage withholdings under delinquency, immediate wage withholdings for new cases, universal wage withholdings, and state income tax fund interception (Huang et al. 2004). The CPS measures that compose the CSE index are the percentage of mothers with child support received, the ratio of child support collections to child support guidelines, and the average support payment divided by the maximum AFDC/TANF benefit. The CSE index in 1999 ranges from 64 to 189 with a mean of 108; a higher number indicates a stronger child support system.

e. State Welfare Generosity toward Noncitizens

The 1996 Personal Responsibility and Work Reconciliation Act (PRWORA) denied most types of federally funded means-tested assistance to noncitizens who arrived after 1996, and limited the eligibility of many noncitizens already living in the United States. However, PRWORA allowed states to use their own funds to provide public assistance to immigrants and to determine their own eligibility criteria. States have taken different approaches to social welfare programs for immigrants and their children. Some states have extended benefits to legal resident noncitizens and others allow access to legal immigrants only after a period of U.S. residency, but none routinely give benefits to unauthorized immigrants.

We use the Urban Institute’s index of state generosity toward noncitizens. This index measures the extent to which particular states offered their state-provided safety nets to the immigrant population after 1996. Such regulations may include the provision of Medicaid and TANF to immigrants before or after August 22, 1996, within or after the 5-year ban period, and the creation and expansion of food, General Assistance, and other health insurance programs. For a detailed explanation of how the scale was constructed, see Zimmerman and Tublin (1999). This index of state generosity is a four-point ranked ordinal variable, with 4 representing the most generous state and 1 representing the least generous state. We recoded this ordinal variable into a binary variable and defined a “generous state” as a state where state-funded assistance was either “most available” or “somewhat available,” and a “not generous state” as a state where state-funded assistance was either “less available” or “least available.”

f. Anti-Immigrant Environment

There is no readily available measurement for the “chilling effect” of the anti-immigrant environment. Both Watson (2010) and Vargas (2016) used federal immigration enforcement efforts as a proxy for the anti-immigrant environment. However, since immigration laws are enforced at the federal level by the Special Agent in Charge (SAC) districts, and the SAC districts areas do not align with states borders, using the number of immigrants apprehended or deported in each SAC district as a proxy for the anti-immigrant environment in each state or city within the specific SAC district is problematic.

Another potential anti-immigrant measure is the level of anti-immigrant legislation at the state level. The Migration Policy Institute (MPI) used the StateNet database in LexisNexis and Westlaw to locate all state legislation related to immigrants that has been passed, been rejected, or has expired across the 50 states since 2007.⁷ The National Conference of State Legislatures (NCSL) created a database to track legislation related to immigrants on topics such as budgets, education, public benefits, law enforcement, and employment since 2008.⁸ However, since there was not much state-level immigration legislation around 2000, the relevant time frame for our study, the creation of an anti-immigrant measure based on anti-immigrant laws was not feasible.⁹

Instead, we used the ANES data to directly measure public opinions about illegal immigration.¹⁰ The ANES is the key source for political opinion and participation information in the United States. It is a national probability sample conducted every two years (presidential and mid-term elections). The sample population is restricted to adult citizens, and the sample size usually ranges from 1,200 to 2,200. There is a wide range of questions on contemporary policy issues, the current election of that year, and party identification and participation, as well as standard demographic and socio-economic status variables. Specifically, the ANES asks each respondent whether he or she would like to see an increase or decrease in federal spending on border security to combat illegal immigration. The answers to this question in the years 2000, 2002 and 2004 are collapsed and aggregated to the state level to generate the fraction of state residents who would like to see an increase in federal enforcement against illegal immigration. We fully recognize that only some of the mixed-status families in the FFCWB data have one or more parents that are undocumented. However, given the salience of illegal immigration, public opinion about federal enforcement against illegal immigration is a good proxy for the anti-immigrant environment at the state level.

g. Parents’ Characteristics

Finally, we include a set of control variables commonly used in the literature, including indicators for the mother’s need for child support income, the father’s ability to provide child support, and the parents’ relationship (which might impact the parents’ attitudes towards child support). In addition, we included the mother’s knowledge of and perceptions about the child support system to eliminate bias that could arise from systematic differences between immigrants and native U.S. citizens in regard to their child support behaviors. These questions are only available in the FFCWS, and not available in the CPS-CSS, or any other nationally representative datasets. Specifically, answers to the question “If BF (boyfriend) doesn’t want to marry the mother, could he be required to pay child support?” are used as a proxy for mothers’ knowledge of the child support system, and answers to the question “How important is it for BF (boyfriend) to provide regular financial support to children?” are used as a proxy for mothers’ perceptions about child support.

Table 2 provides summary statistics for all variables used in our models.

Table 2.

Descriptive Statistics

Variable	N	Mean	SD	Min	Max
Outcome Variables
Paternity Establishment (N= 2863)	2,863	.834	.372	0	1
Child Support Order (N=2630)	2,630	.320	.467	0	1
Child Support Receipt (N= 2491)	2,491	.170	.367	0	1
In-kind Child Support (N= 2135)	2,135	.690	.463	0	1
Independent Variables
Immigration related Variables
Mixed Status Families	2,257	0.141	0.348	0	1
Noncitizen Mothers	2,814	0.071	0.258	0	1
Mixed-Status Families on TANF	2,148	0.033	0.178	0	1
Noncitizen Mothers on TANF	2,643	0.012	0.109	0	1
Mixed-Status Families on SNAP	2,147	0.061	0.239	0	1
Noncitizen Mothers on SNAP	2,643	0.027	0.162	0	1
Mother’s Knowledge of and Need for CS
Mother has good knowledge of the CS system	2,861	0.500	0.500	0	1
Mother thinks financial support from father important	2,884	0.828	0.377	0	1
Mother received TANF last year	2,719	0.429	0.495	0	1
# of economic hardships mother experienced	2,708	1.328	1.714	0	12
Mother had income from earnings last year	2,879	0.681	0.466	0	1
Mother has children with other men	2,567	0.413	0.492	0	1
Mother is older than 30 years old	2,893	0.152	0.359	0	1
Mother has more than high school education	2,893	0.541	0.498	0	1
Mother is Hispanic	2,870	0.262	0.440	0	1
Mother is Black	2,870	0.570	0.495	0	1
Mother is of other race (White is the omitted category)	2,870	0.025	0.156	0	1
Father’s Ability/Willingness to Provide CS
Father ever in jail	2,860	0.051	0.219	0	1
Father worked when child was 1 year old	2,556	0.690	0.463	0	1
Father has children with other women	2,619	0.443	0.497	0	1
Father and mother in steady relationship	2,893	0.686	0.464	0	1
State-level Policy Environment
CS enforcement index	2,893	107.971	35.822	64	189
State generosity toward noncitizens	2,893	0.642	0.480	0	1
ANES anti-immigrant sentiment	2,893	0.624	0.067	0.542	0.755

Source: Fragile Families and Child Wellbeing Survey, the Baseline, Age 1 and Age 3 Data.

4.3. Empirical Strategies

The analysis examines the effect of parents’ legal status on their child support outcomes. We ran probit models with a set of binary indicators as our dependent variables, including paternity establishment, child support order establishment, child support receipt, and in-kind child support. We report the marginal effects calculated from probit coefficients and z-statistics. The marginal effects presented should be interpreted as the change in the probability of the dependent variable associated with a discrete change in a binary independent variable and an incremental change in a continuous independent variable, holding all other variables at their means.

Since we are using observational data and it is not possible to observe what the child support outcomes would have been if the mixed-status families were citizen families, we have to determine whether what we detect using regression models are the true impact of the immigration status. Following the counterfactual framework (Rubin 1977, Rosenbaum and Rubin 1983), we used propensity score matching (PSM) to select the best control matches (citizen families) for each individual in the treatment group (mixed-status families) based on all other preexisting observed characteristics. When the relevant differences between any two units are captured in the observable (pretreatment) covariates, which occurs when outcomes are independent of assignment to treatment conditioned on pretreatment covariates, matching methods can yield an unbiased estimate of the treatment impact.¹¹ In this study, we performed PSM as a robustness test, but there are some important caveats. Propensity score matching is done only for characteristics that are observable and measured. It assumes no selection bias based on unobserved characteristics. For example, immigrants in the United States might possess an unobserved or unmeasured characteristic that is fundamentally different from those of native U.S. citizens, such as the lack of self-advocacy. If this characteristic also impacts child support behaviors, causal claims based on matching results become questionable. However, Cook, Shadish, and Wong (2008) found that matching individuals on location helps to eliminate biases based on location that are otherwise unobservable and that create nonequivalencies between the treatment and comparison groups. Accordingly, we matched mixed status families with U.S. citizen families using the mother’s city of residence in addition to standard demographic and social-economic variables, as well as the mother’s knowledge of and perceptions about the child support system. We used the nearest neighbor matching without replacement to create a matched sample of mixed-status families and U.S. citizen families, and then ran our probit models using the full set of control variables listed in Table 2 to estimate the average treatment effect (ATE).

Since the FFCWS oversamples low-income fragile families in major U.S. cities, one might be concerned about the representativeness of the FFCWS data of the child support eligible population. We conducted sensitivity analysis using the 2001 CPS-CSS because the respondents were facing the same policy environments in terms of child support policy, immigration policy and welfare policy as the respondents of the FFCWS. We limited the sample to mothers aged 15–65 who had a child under the age of 21 living in the household while the biological father lived outside of the household. In the CPS-CSS, custodial mothers, not nonresident fathers, report the child support information. It is not possible to study the child support behaviors of mixed-status families using these data since the CPS-CSS does not allow the linking of custodial mothers’ information with that of noncustodial fathers. In addition, CPS-CSS offers a far less extensive list of child support questions. Paternity establishment is not consistently defined in CSS, and there are no questions about mothers’ knowledge of and perceptions about the child support system. To make the analysis based on the FFCWS comparable with the analysis based on the CPS-CSS, we used noncitizen mothers as the measurement for immigration status and limited the covariates to only those that are also available in CPS-CSS. The outcome variables are child support order, child support receipt, and in-kind child support.

V. Results and Discussion

5.1. Descriptive Statistics

Table 3 shows the differences in child support outcomes between mixed-status families and citizen families. Among all families with citizenship information, 62 percent of citizen families are eligible for child support, compared to only 41 percent of mixed-status families. This is primarily due to the higher marriage rate in mixed-status families than in citizen families (37 percent of mothers in mixed-status families vs. 25 percent of those in citizen families were married to the focal child’s father when the child was born). However, among the child support eligible families, the rate of having paternity established is high for both groups (87 percent for citizen families and 82 percent for mixed-status families). This is consistent with findings in previous studies which show that there has been considerable success in increasing the rate of paternity establishment, even in the very low-income population and in fragile families where relationships are volatile.

Table 3.

Sample Size for Child Support Outcomes by Family’s Citizenship Status

		Citizen Families	Mixed Status Families	Total
Full Sample	N	3,122	781	3,903
CS Eligible	N % of full sample	1,938 62%	319 41%	2,257 58%
CS Plausible	N % of full sample	1,733 57%	269 34%	2,042 52%
Paternity Established	N % of CS eligible	1,690 87%	263 82%	1,953 87%
Child Support Order	N % of CS plausible % of paternity established	592 33% 35%	34 13% 13%	626 31% 26%
Child Support Receipt	N % of CS plausible % of having CS order	319 18% 54%	20 7% 59%	339 17% 54%
In-kind Support	N % of CS plausible	1,077 61%	97 36%	1,174 57%

Source: Fragile Families and Child Wellbeing Survey, the Baseline, Age 1 and Age 3 Data.

A striking difference between mixed-status families and citizen families lies in the rate of having a formal child support order. Among those who are plausible to get child support, about 33 percent of citizen families have a formal child support order, while the rate for mixed-status families is only 13 percent. However, among those who have a formal child support order, the rate of receiving child support payment is slightly higher for mixed-status families than for citizen families (59 percent vs. 54 percent). These results indicate that the primary barrier for receiving child support payment in mixed-status families is the low rate of having a formal child support order. In addition, noncustodial parents in mixed-status families are much less likely than those in citizen families to provide in-kind support for the child (36 percent vs. 61 percent).

5.2. Probit Regression Results

Table 4 reports the probit regression results of the determinants of paternity establishment and child support orders. As expected, no statistically significant impact of legal status was found on paternity establishment. However, the results show that legal status has a statistically significant negative impact on the probability of getting a child support order, child support payments, and in-kind child support. A mixed-status family is 19.9 percent less likely than a family with both citizen parents to have a formal child support order. This impact is statistically significant at the 0.001 level. Similarly, a mixed-status family is 13.3 percent less likely than a family where both parents are citizens to have received child support payments in the past year. In addition, mixed-status families are also 18.6 percent less likely to receive in-kind child support. These results provide evidence for our hypothesis that U.S. citizen children in mixed-status families are faring much worse than those in citizen families in terms of child support outcomes.

Table 4.

Probit Regressions using FFCWS Data: Full Model, Marginal Effects

	Paternity	Child Support Order	Child Support Receipt	In-kind Support
Immigration-related Variables
Mixed-status families	−0.033	−0.199***	−0.133***	−0.186*
	(0.035)	(0.041)	(0.032)	(0.075)
TANF	−0.003	0.101**	0.028	0.034
	(0.018)	(0.033)	(0.028)	(0.028)
Mixed-status families on TANF	−0.033	0.361*	0.314	−0.091
	(0.065)	(0.156)	(0.172)	(0.135)
SNAP	−0.006	0.031	0.024	−0.048
	(0.019)	(0.032)	(0.027)	(0.029)
Mixed-status families on SNAP	0.016	−0.096	−0.030	0.094*
	(0.040)	(0.110)	(0.096)	(0.046)
Mother’s Knowledge of and Need for CS
Mother has good knowledge of the child support system	0.038**	−0.009	0.003	−0.040
	(0.013)	(0.024)	(0.020)	(0.022)
Mother thinks financial support from father important	0.013	0.018	0.031	0.008
	(0.018)	(0.032)	(0.026)	(0.029)
Number of economic hardships mother Experienced	−0.005	−0.007	−0.004	0.001
	(0.004)	(0.007)	(0.006)	(0.006)
Mother had income from earnings last year	−0.024	−0.029	−0.029	0.057*
	(0.014)	(0.028)	(0.024)	(0.027)
Mother has children with other men	−0.013	−0.023	−0.022	−0.008
	(0.014)	(0.026)	(0.022)	(0.024)
Mother is older than 30 years old	0.017	−0.015	0.005	0.026
	(0.017)	(0.035)	(0.031)	(0.030)
Mother has more than high school education	0.037*	0.027	0.019	−0.005
	(0.015)	(0.026)	(0.022)	(0.023)
Mother is Hispanic	0.004	0.014	−0.045	0.026
	(0.024)	(0.043)	(0.032)	(0.036)
Mother is Black	−0.020	0.026	−0.028	0.060
	(0.021)	(0.036)	(0.029)	(0.035)
Mother is of other race	−0.026	0.065	0.040	0.137***
	(0.054)	(0.092)	(0.078)	(0.019)
Father’s Ability/Willingness to Provide CS
Father ever in jail	−0.075	0.053	−0.065	−0.236***
	(0.045)	(0.061)	(0.042)	(0.071)
Father worked when child was 1yrs old	0.047**	−0.003	0.087***	0.138***
	(0.017)	(0.027)	(0.020)	(0.027)
Father has children with other women	−0.037*	0.022	0.012	−0.059*
	(0.015)	(0.025)	(0.021)	(0.023)
Father and mother in steady relationship	0.031	−0.164***	−0.114***	0.138***
	(0.018)	(0.032)	(0.029)	(0.030)
State-level Policy Environment
Child support enforcement index	0.000	0.002***	0.001***	−0.000
	(0.000)	(0.000)	(0.000)	(0.000)
State generosity toward noncitizens	0.004	−0.018	−0.004	0.059*
	(0.015)	(0.027)	(0.023)	(0.026)
ANES anti-immigrant sentiment	0.030	0.158	0.006	0.016
	(0.115)	(0.208)	(0.180)	(0.189)
N	1,724	1,528	1,454	1,120
Chi-squared	65.656	155.391	92.947	128.182
P	0.000	0.000	0.000	0.000

Source: Fragile Families and Child Wellbeing Survey, the Baseline, Age 1 and Age 3 Data.

Marginal effects represent the discrete change from the base level holding all other variables at the means.

The Delta-method standard error for the marginal effects is shown in parentheses.

^*p<0.05,

^**p<0.01,

^***p<0.001

Our results show that welfare agencies play a big role in connecting mothers with the child support system by establishing a formal child support order. Families that had received TANF since the focal child’s birth were 10.1 percent more likely to have a formal child support order than families not on TANF. Receiving welfare benefits has an even greater impact on the child support outcomes of mixed-status families. TANF participation since the focal child’s birth increases mixed-status families’ probability of having a child support order by almost 36 percent, and this impact is statistically significant at the 0.05 level. However, TANF participation does not have a statistically significant impact on paternity establishment, child support receipt, or in-kind child support. In addition, although there are similar cooperation requirements in other programs such as SNAP, receiving SNAP benefits does not have a statistically significant impact on any of the child support outcomes examined.

The mother’s knowledge of the child support system has a positive and significant impact on the probability of having paternity established. It appears to have no effect on the probability of having a child support order, child support receipt, or in-kind support. The father’s ability to pay as measured by his employment status increases the likelihood of paternity establishment, child support receipt, and in-kind child support, as expected. The father’s willingness to pay also matters for child support outcome. Fathers who have children with other women are less likely to establish paternity and provide in-kind support. On the other hand, mothers who are in a steady relationship with the father are less likely to be in the formal child support system, but more likely to receive in-kind support from the father.

Institutional factors such as state-level child support enforcement efforts significantly improve formal child support outcomes such as child support order and child support receipt. A one unit increase in the Child Support Enforcement Index (ranged from 64 to 189 with a mean of 108), for example, is associated with a 0.2 percent increase in the likelihood of a family getting a formal child support order, and a 0.1 percent increase in the likelihood of a family receiving any child support. These impacts are statistically significant at the 0.001 level.

State generosity towards noncitizens seems to have a positive and significant impact on the likelihood of receiving in-kind child support, but does not have a statistically significant impact on formal child support outcomes. The anti-immigrant sentiment constructed from the ANES data also does not have a statistically significant impact on child support outcomes. These results indicate that the “chilling effect” that was found in immigrants’ utilization of welfare or Medicaid programs is not evident in the child support system. The caveat however is that mixed-status families are less likely to use TANF, and TANF has a large positive effect on child support outcomes.

5.3. Propensity Score Matching Results

Appendix A presents the covariate balance between the treatment group and the control group before and after matching. As the table shows, the differences between the mixed status families and the citizen families before matching are statistically significant on almost all of the matching variables, indicating that there are systematic differences between the treatment and control groups and therefore a quasi-experimental method such as PSM is warranted. All of the group differences become insignificant after matching, indicating a high quality match. The mean bias was reduced from 36.4 to 5.0, and the LR chi2 was reduced from 383.56 to 8.09, indicating the statistically significant diminishing of the group difference after matching. We also checked to make sure the common support assumption is satisfied.

As can be seen from column 4 in Table 5, the matching estimates are consistent with the probit regression results, although the magnitudes of the significant coefficients are slightly different. The matching results show that mixed-status families are 20.7 percent less likely than citizen families to have a child support order, 14.4 percent less likely to receive child support payments, and 17.1% less likely to receive in-kind child support. The impact of legal status is not significant for paternity establishment.

Table 5.

Summary Results Table

Outcome	Family type	Probit model w/all covariates (Preferred)	Matching estimator w/all covariates	Probit Model w/all covariates	Probit model w/only covariates available in CPS
		FFCWS	FFCWS	FFCWS	FFCWS	CPS
Paternity	Mixed-status family	−0.033	−0.009
	Mother noncitizen			−0.035	−0.061
CS Order	Mixed-status family	−0.199***	−0.207***
	Mother noncitizen			−0.219***	−0.240***	−0.189***
CS Receipt	Mixed-status family	−0.133***	−0.144***
	Mother noncitizen			−0.137**	−0.134***	−0.138***
In-kind Support	Mixed-status family	−0.186*	−0.171*
	Mother Noncitizen			−0.101	−0.084	−0.063

Source: Fragile Families and Child Wellbeing Survey, the Baseline, Age 1 and Age 3 Data.

Notes: Marginal effects represent the discrete change from the base level holding all other variables at the means.

Nearest neighbor matching method is used. Population average treatment effects are reported. The number of matched mixed-status families is 157 for paternity establishment, 136 for CS order, 131 for CS receipt, and 72 for in-kind support.

^*p<0.05,

^**p<0.01,

^***p<0.001

5.4. Sensitivity analysis using CPS-CSS data

To address the concern that the FFCWF oversamples low-income fragile families and as such is not representative of the child support eligible population, we used the CPS-CSS data from 2001 as an additional validity check. Column 5, 6, and 7 in Table 5 presents the marginal effects derived from probit regressions using noncitizen mothers as the measurement of immigration status and using the full list of control variables available in the FFCWS, only the control variables in the FFCWS that are comparable to the CPS-CSS variables, and the CPS-CSS data respectively. The sign, magnitude and significance levels are largely consistent across all the model specifications, indicating that the negative relationship we found between mixed-status families and child support outcomes is robust and not sensitive to the construction of measures, the model specifications, and the datasets used. Table 6 reports the full regression results of using noncitizen mothers as the indicator of immigration status and only the covariates that are comparable in the FFCWS and the CPS-CSS.

Table 6.

Comparison of Regression Results Using FFCWS Mothers Only Sample and CPS data, using only variables available in CPS: Marginal Effects

	Child Support Order		Child Support Receipt		In-kind Support
	FFCWS	CPS	FFCWS	CPS	FFCWS	CPS
Mother noncitizen	−0.240***	−0.189***	−0.134***	−0.138***	−0.084	−0.063
	(0.041)	(0.037)	(0.029)	(0.036)	(0.070)	(0.039)
TANF	0.103***	−0.016	0.017	−0.039	−0.025	−0.077*
	(0.025)	(0.033)	(0.021)	(0.032)	(0.027)	(0.033)
Noncitizen mothers on TANF	0.102	−0.089	0.136	0.035	−0.243	0.102
	(0.150)	(0.135)	(0.164)	(0.135)	(0.162)	(0.064)
SNAP	0.015	0.056*	−0.011	0.018	−0.052	−0.001
	(0.027)	(0.024)	(0.022)	(0.025)	(0.029)	(0.021)
Noncitizen mothers on SNAP	0.084	0.195*	0.052	0.240**	0.059	−0.086
	(0.138)	(0.077)	(0.128)	(0.088)	(0.110)	(0.100)
Mother employed	0.014	0.011	0.007	−0.005	0.032	0.017
	(0.022)	(0.018)	(0.018)	(0.018)	(0.024)	(0.016)
Household income	−0.002**	0.001	−0.000	0.001	0.001*	0.001*
	(0.001)	(0.001)	(0.000)	(0.001)	(0.001)	(0.001)
Mother over 30 years old	−0.018	0.076***	0.012	0.049**	0.036	0.075***
	(0.028)	(0.018)	(0.024)	(0.018)	(0.030)	(0.017)
Mother has more than high school education	0.046*	0.109***	0.043*	0.110***	0.007	0.049*
	(0.021)	(0.022)	(0.017)	(0.021)	(0.023)	(0.021)
Mother is Hispanic	−0.013	−0.089***	−0.053*	−0.082***	0.016	−0.024
	(0.036)	(0.025)	(0.026)	(0.024)	(0.039)	(0.023)
Mother is Black	−0.011	−0.140***	−0.057*	−0.150***	0.075*	−0.056**
	(0.030)	(0.020)	(0.024)	(0.019)	(0.034)	(0.019)
Mother is of other race	0.050	−0.122***	0.078	−0.143***	0.181***	−0.089*
	(0.074)	(0.035)	(0.065)	(0.032)	(0.050)	(0.037)
Child support enforcement index	0.002***	0.001*	0.001**	0.000	−0.000	0.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
State generosity toward noncitizens	−0.026	−0.027	−0.017	−0.004	0.056*	0.008
	(0.023)	(0.016)	(0.019)	(0.015)	(0.025)	(0.014)
ANES anti-immigrant sentiment	0.255	−0.030	0.005	−0.004	0.156	−0.013
	(0.170)	(0.025)	(0.142)	(0.025)	(0.182)	(0.022)
N	2389	4798	2263	4798	1944	3595
Chi-squared	151.085	224.668	55.746	193.559	62.843	119.574
P	0.000	0.000	0.000	0.000	0.000	0.000

Source: Fragile Families and Child Wellbeing Survey, the Baseline, Age 1 and Age 3 Data.

The Current Population Survey – Child Support Supplement, 2001.

Notes: Marginal effects represent the discrete change from the base level holding all other variables at the means.

The Delta-method standard error for the marginal effects is shown in parentheses.

^*p<0.05,

^**p<0.01,

^***p<0.001

VI. Conclusions

Very little empirical evidence exists on immigrants’ participation in the child support system in the United States This is regrettable because children of immigrants represent the fastest growing share of children in the United States. With 5.6 million U.S. citizen children in mixed-status families, and with half of the children of immigrants living in poverty, the wellbeing of these children is ripe for policy attention. This research provides an insight into the child support behaviors of mixed-status families living in fragile families. Our data analysis suggests that although paternity establishment rates are similar between mixed-status families and citizen families, mixed-status families are less likely to have a formal child support order, less likely to have received child support payments, and less likely to receive in-kind support. Since mixed-status families have a greater risk of being poor and are less likely to be eligible for federally funded welfare programs such as TANF, SNAP, EITC, or Medicaid, the child support system represents one of the few possibilities that these low-income custodial parents have for obtaining financial support to raise their children. However, our research indicates that children living in mixed-status families are falling through the cracks in a child support program that does not reflect their realities. According to the Fragile Families data, one out of seven nonmarital children is born to mixed-status families in the 20 cities surveyed between 1998 and 2000. However, only 11 percent of the child support eligible mixed-status families are getting child support orders, compared to 31 percent for citizen families. The Migration Policy Institute estimates that there are 14.9 million second-generation children—those who were born in the United States to at least one foreign-born parent in 2011.¹² The number of U.S. citizen children in mixed-status families who are eligible for child support but do not have a child support order is strikingly high and should be of great concern for policy makers, service providers, and researchers interested in reducing poverty. We expect this number to continue rising rapidly given the fertility trend of immigrants in the United States and the decline in marriage rates. How to better design the child support program to take this vulnerable population into consideration is a topic that needs further research and more policy attention.

While the nation’s attention is on reforming current immigration policy, it is important to note that American children of mixed-status families are not receiving the government services for which they are eligible. Child support agencies need to create target programs to reach this underserved population. Child support professionals, community-based organizations, policy makers, advocates, and researchers should engage in assisting mixed-status families to obtain this important financial resource.

Our analysis also shows that the consistent policy emphasis on using the child support system to promote family self-sufficiency and child wellbeing has paid off in many respects. The paternity establishment rate is very high, even in fragile families. State-level child support enforcement efforts have a significant positive impact on the likelihood of child support order and child support receipt. Moreover, mothers who are already receiving welfare benefits are more likely to have a child support order, indicating that the incentive for states to recoup part of the welfare expenses through child support payments has been successful. Nevertheless, the proportion of families with a formal child support order is low, especially in mixed-status families.

As discussed throughout the paper, there are some important limitations of our study, especially with the use of the self-reported FFCWS data. First, we did not have detailed immigration status information, so we could not investigate whether undocumented immigrant parents behave differently than legal immigrant parents. Our hypothesis about the impact of an anti-immigrant policy environment thus needs to be tested using better data. Second, we did not have sufficient data to model the amount of child support received and arrears accrued. The dummy variable child support payment does not capture the differences in the amount of child support received. In addition, the FFCSW data only provide information on children born in the United States, while some children in mixed-status families have been born elsewhere. This paper is not able to capture the complicated family structure of mixed-status families in terms of citizenship and child support outcomes. Further research is required to investigate whether and how the child support system works to promote the well-being of vulnerable children, including those in mixed-status families.

Appendix A. Check for Covariates Balance before and after PSM

	U(Unmatched)	Mean		% Bias Reduction		t-test
	M (Matched)	Treated	Control	%bias	|bias|	t	p>|t|
TANF	U	0.20588	0.39558	−42.2		−5.68	0
	M	0.21304	0.24783	−7.7	81.7	−0.88	0.377
SNAP	U	0.42017	0.59371	−35.2		−5.06	0
	M	0.43478	0.44348	−1.8	95	−0.19	0.851
Mother’s city of residence	U	8.0336	8.6165	−11.8		−1.65	0.098
	M	8.113	8.5087	−8	32.1	−0.88	0.382
Mother has good knowledge of the CS system	U	0.33613	0.53681	−41.3		−5.8	0
	M	0.34348	0.41304	−14.3	65.3	−1.54	0.125
Mother thinks financial support from father important	U	0.7395	0.84337	−25.7		−3.97	0
	M	0.75652	0.78261	−6.5	74.9	−0.66	0.508
# of economic hardships mother experienced	U	0.98319	1.3072	−19.6		−2.76	0.006
	M	0.9913	1.1348	−8.7	55.7	−0.98	0.329
Mother had income from earning last year	U	0.55882	0.73159	−36.7		−5.49	0
	M	0.57391	0.61304	−8.3	77.4	−0.85	0.394
Mother has children with other men	U	0.34454	0.40629	−12.8		−1.81	0.071
	M	0.33913	0.34783	−1.8	85.9	−0.2	0.845
Mother is older than 30 years old	U	0.21849	0.13855	21		3.22	0.001
	M	0.20435	0.2087	−1.1	94.6	−0.11	0.909
Mother has more than high school education	U	0.37815	0.58969	−43.3		−6.17	0
	M	0.3913	0.43913	−9.8	77.4	−1.04	0.299
Mother is Hispanic	U	0.76471	0.21687	130.8		18.96	0
	M	0.75652	0.75652	0	100	0	1
Mother is Black	U	0.14706	0.58568	−102.1		−13.2	0
	M	0.15217	0.15652	−1	99	−0.13	0.898
Mother is of other race	U	0.04622	0.02075	14.2		2.37	0.018
	M	0.04783	0.03913	4.8	65.9	0.46	0.648
Father ever in jail	U	0.01261	0.03882	−16.6		−2.04	0.042
	M	0.01304	0.0087	2.8	83.4	0.45	0.654
Father worked when child was 1 year old	U	0.83613	0.7095	30.5		4.09	0
	M	0.83043	0.8087	5.2	82.8	0.61	0.545
Father has children with other women	U	0.30252	0.40361	−21.2		−2.98	0.003
	M	0.3087	0.3087	0	100	0	1
Father and mother in steady relationship	U	0.85294	0.80321	13.2		1.82	0.069
	M	0.85652	0.86957	−3.5	73.8	−0.41	0.685

Source: Fragile Families and Child Wellbeing Survey, the Baseline, Age 1 and Age 3 Data.