In this research note, we demonstrate that trends in the likelihood of child support agreements differ by marital history (i.e., never-married vs. ever-married) and by whether measures rely on the stock of families (i.e., all those in which children live apart from a parent) or the flow (i.e., those that include children who newly live apart from a parent) in a given year. While previous research has highlighted difference by marital history, the contrast between stock and flow is a new contribution. Trends are typically measured with reference to the stock of cases, even while the flow of cases, which more immediately reflects concurrent policy changes, is more relevant in many contexts. Interpretations of recent declines in child support agreements in the stock of cases—referenced as evidence for both mandating participation and the impracticality of requiring child support—may be better informed by considering the flow of cases. We find the flow of previously married mothers increasingly likely to have child support agreements while the likelihood is relatively consistent over time for never-married mothers. For both groups, using the flow measure, we find notable increases in agreements without payments due in the most recent period. These findings underscore the importance of differentiating stock and flow, and by marital history, in considering the proportion with agreements as an indicator of the effectiveness of current policy.
With one in four children living apart from one of their parents and at higher risk of poverty (Grall 2020), the child support program has a broad and important mission in the United States. Child support, in which a noncustodial parent (with whom the child does not live primarily, most often the father)1 provides financial support toward raising a child, is a crucial financial support for the millions of custodial-parent families who receive it (Grall 2020; Sorenson 2021). However, the most commonly used measure suggests that the proportion of custodial parents who have legal child support agreements has been declining over time.2 This pattern has sparked critiques of the child support program, from both those who see the decline as evidence that participation should be mandated for more families (e.g., Doar 2017) and those who see the decline as evidence of the poor fit between traditional child support orders and the economic and social situations of many families (e.g., Nepomnyaschy et al. 2022).
In this research note, we demonstrate that trends in the likelihood of child support agreements differ by marital history (i.e., never-married vs. ever-married) and by whether measures rely on the stock of families (i.e., all families demographically eligible or at risk for involvement with the child support system) or the flow (i.e., families newly eligible owing to a nonmarital birth or divorce) in a given year. We also consider differences for custodial mothers and fathers, where the data allow. While previous research has highlighted difference by marital history, the contrast between stock and flow is a new contribution. Trends are typically measured with reference to the stock of cases in child support research, even while the flow of cases, which more immediately reflects concurrent policy changes, is more relevant in many contexts.
The main source of population-level data on child support, the Child Support Supplement (CSS) of the Current Population Survey (CPS), shows that the proportion of custodial-parent families with agreements increased slightly from 1993 to 2003 (from 57% to 60%) but then declined, and it has been stable at about 50% since 2009 (Grall 2020). The growing share of custodial fathers, who are less likely to have agreements than custodial mothers, contributes to this decline, but agreements among custodial mothers have also seen a large decline, from a high of 64% in 2003 to 51% in 2017 (Grall 2020). Although there has been some (now dated) research on trends in child support (e.g., Case et al. 2003; Huang 2009), reasons for this decline have not been well understood.
In this research note we examine the trend in agreements in more detail. We look separately at never-married and ever-married parents, and we document a key limitation of the dominant current empirical approach, which examines the stock of all cases. Flow measures, which categorize cases by when they entered the system, more closely track responses to policy changes than do static stock measures, which can be muddled by changes in case composition over time and other issues (Abramowitz and Dillender 2017). Indeed, previous demographic work from a variety of policy realms, including welfare reform, the Affordable Care Act, the Earned Income Tax Credit, and others, relies on flow measures in outcomes rather than stock measures and shows notable differences in stock and flow outcomes (e.g., Abramowitz 2016; Bitler et al. 2004; Herbst 2011; Klerman and Haider 2004). Our purpose is not to make causal claims regarding the trends, but rather to clarify the distinction between trends in the stock and flow, which are useful to distinguish and understand.
To the best of our knowledge, there has been very little consideration of national trends in child support agreements using the flow of eligible cases, despite the interest in understanding the impact of policy changes that principally impact new cases. This research note emphasizes the importance of differences in stock and flow measures and how they may result in diverging understandings of population outcomes in child support. In addition, we document long-term trends for never-married and ever-married mothers (which are not published separately in the main child support report (Grall 2020)) and provide selected information for custodial fathers. We also provide more detail on types of child support agreements. While our focus is on legal agreements, and particularly legal agreements that require an amount due, we also present information on informal agreements.
Data and Methods
We use repeated cross sections of the CPS-CSS. The CSS is fielded in April biennially; households with a child who has a biological parent living outside of the household are eligible for participation. The CSS collects data on child support agreements (including whether they are legal or informal), child support payments and receipt (both awarded and actual amounts) in the previous year, and other relevant child support details. We use data from 1994 through 2018, covering all available years of comparable data.3 Because custodial mothers represent the vast majority of custodial parents (80%), and because they are more economically vulnerable than custodial fathers (Grall 2020), we focus primarily on them, though we also include some measures for fathers. Our sample starts with custodial parents who are demographically eligible for child support, that is, have at least one child who lives apart from their other parent. From that group, we then limit to those with children younger than 18. We show trends separately for never-married and ever-married custodial parents.4
The survey asked custodial parents whether they have ever had a child support agreement of any kind in place, and, if they respond affirmatively, they are asked whether the agreement was informal or legal, or whether a legal agreement is pending.5 The custodial parent is also asked whether they were due any child support in the last year. Some custodial parents report that they have (or had) a legal agreement but that no payments were due in the last calendar year. We are primarily interested in legal agreements involving child support, and specifically in legal agreements that require an amount due (a child support order), since only legal agreements can be enforced.
We examine both stock and flow measures to understand how these different approaches may influence the understanding of trends. Our stock measure uses the calendar year prior to the year that the participant was included in the CSS (the period covered by whether payment was due). For the flow, we aim to use the year that the custodial parent was first eligible to enter the child support system. For parents with an agreement, their flow year is based on their self-reported first year of the agreement. For parents without an agreement, we assign a year intended to proxy the year they would become eligible for child support. For ever-married custodial parents, we use parent's self-reported year of most recent divorce when available.6 For ever-married parents without a divorce year, and for never-married parents, we use the year of the oldest child's birth.7 In all instances, we only include reports up to 12 years following the point of entry.
We group flow years into six evenly divided periods (1994–1997, 1998–2001, 2002–2005, 2006–2009, 2010–2013, 2014–2017). This allows us to avoid any issues of sparse data. We document the proportion of cases with a legal agreement within each stock year and within each flow year bin.
We focus our discussion on custodial mothers, who account for at least 80% of custodial parents over the entire period (the share of custodial fathers grew from 16.0% to 19.8% of the stock of all custodial-parent cases). While custodial mothers remain dominant, the composition by marital status has changed substantially. With increases in nonmarital fertility and declines in marital fertility, the proportion of custodial mothers who have not been married has been increasing, from less than 30% of the stock of cases in 1993 to more than 40% by 2013. Never-married mothers may be less likely to have legal agreements, since divorce requires a legal act and provisions for child support are often part of the divorce process. As a result, in the analyses that follow, we focus on never-married and ever-married custodial mothers separately, and further examine differences in stock and flow measures.
Looking at the stock of cases (Figure 1 and online appendix Table A1), we see that for both never-married and ever-married mothers, the highest proportion with a legal agreement is in 2003.8 However, the patterns over the full period differ. For never-married mothers (panel a), an increase in the 1990s is followed by a decline, leaving the proportion with a legal agreement virtually the same in 2017 (36.4%) as it was in 1993 (36.2%), though the point estimate for the proportion with a legal agreement with payment due in 2017 is slightly higher than in 1993 (growing from 30.7% to 32.5%; the difference is nonsignificant). The proportions of never-married mothers with pending agreements and informal agreements are relatively small and show no consistent trends.
For ever-married custodial mothers (Figure 1, panel b), modest but nonsignificant gains in the 1990s were followed by substantial declines. After the peak in 2003 (65% with legal agreements and 60% with an amount due), the proportions decline steadily, though not substantially, after 2011. For ever-married mothers, in 2017 just more than half reported a legal agreement, and just less than half reported that payments were due. The proportions with pending legal agreements and informal agreements are both small.
In summary, considering the stock of cases, we see the understanding that has prevailed since early in this century: the proportion of child support cases with legal agreements, and the proportion due child support, declined for both never-married and ever-married mothers, though for never-married mothers this decline followed a period of comparable growth.
When we consider trends as measured by the flow of child support cases, however, the picture is substantially different (Figure 2). For never-married mothers (panel a), the proportion with legal agreements declines initially and then remains relatively stable, between 41% and 44%, from the early 2000s on. There is somewhat more fluctuation in the proportion with payments due, which declines to 35% in the final period. Moreover, the proportion with informal agreements rises from approximately 7% in the early cohorts to 14% in the most recent cohort. For never-married mothers, then, the simplest outcome measure (the proportion with legal agreements) shows some modest differences: stock measures show a decline since the mid-2000s, and the flow measures suggest an initial decline followed by relative stability.
For ever-married mothers, the stock and flow comparisons are quite different. Whereas the stock of cases suggests a slight increase in agreements for ever-married mothers through the mid-2000s followed by a decline, the flow of cases indicates a decline in the likelihood of legal agreements among cases entering the system in the first years of our analysis followed by a substantial increase in legal agreements for cases entering in 2010 and beyond (Figure 2, panel b). Similar to the case for never-married mothers, among ever-married mothers there is some increase in the proportion with legal agreements that do not require payments. These increases in legal agreements, combined with an increase in informal agreements in the same period, result in a lower level of recent custodial mothers with no agreement; beginning in 2010, fewer than 20% of all ever-married cases have no agreement in place.9
We also examined the trend in child support orders for custodial fathers (not shown). Small sample sizes make it difficult to discern patterns among never-married custodial fathers. But for the stock of ever-married custodial fathers, we find a lower probability of an order, relative to mothers, with no consistent time trend. In contrast, considering the flow of ever-married fathers, about 30% had legal orders from the mid-1990s to the mid-2000s, after which there was a dramatic increase in orders to 64.0% in 2010–2013 (54.4% with an order requiring payments) and 74.1% in 2014–2018 (52.9% with an order requiring payments). Thus, among ever-married custodial fathers, we also find that considering the flow of cases, rather than the stock of cases, yields different results.
The prevailing narrative that child support orders and agreements are becoming less likely is the result of looking only at the stock of cases. Looking at the flow of cases reveals that legal agreements for ever-married mothers (and fathers) are actually becoming more likely. Among never-married mothers, rather than the decline since the early 2000s shown in the stock of cases, we find a stable proportion with formal orders since the 2000s in the flow of cases of never-married mothers. Our detailed differentiation of agreement types further suggests increases in informal agreements and in legal agreements in which nothing is due, resulting in a substantial decline in the flow of never-married mothers, ever-married mothers, and ever-married fathers with no agreement.
In recent years, the apparent decline in child support orders, as shown by the standard stock measure, has been interpreted as a sign that the system is not working well—though critics have not agreed on the nature of the problem. On the one hand, some have argued that the formal child support system is punitive and is not well matched to the situations of many families, especially those in which low-income never-married couples are mandated to establish paternity and a child support order as a result of receiving means-tested assistance, including Medicaid coverage for the child's birth (Edin et al. 2019; Nepomnyaschy et al. 2022).10 One corollary has been a focus on voluntary agreements. On the other hand, the apparent decline in orders in the stock of nonmarital cases since the early 2000s has been cited as evidence that the child support system should increase enforcement and, particularly, mandate paternity and child support order establishment for children in families receiving means-tested benefits. Fewer families face such mandates since the dramatic decline in receipt of cash assistance following the passage of the Temporary Assistance for Needy Families in 1996 and associated policy changes. However, considering the flow of orders among never-married mothers, formal orders are fairly stable, and there has been a substantial increase in informal orders—not apparent when considering the stock of cases. While only an estimated 34.8% of never-married mothers have an order with payments due, 59.0% of the most recent cohort of never-married mothers have some agreement when we examine the flow of cases.
While our results point to the potential importance of considering alternative measures of the prevalence of child support orders, these findings should be seen in the context of several limitations. Our measures of the flow of child support orders depend on the dating of potential eligibility for child support, and the date must be estimated for those without orders (however, alternative dating rules lead to the same conclusions). Sample sizes of those becoming eligible for support in any year are relatively small, precluding annual estimates of the flow, as well as measures of trends for never-married custodial fathers. We do not have information on both parents, so these analyses do not consider the extent to which parents are supporting (or could support) children economically. Finally, sample size constraints limit us to presenting only the national picture, even though states differ in their policies and may differ in their trends.
Notwithstanding these limitations, our findings reinforce the importance of considering never-married and ever-married mothers separately, and also of considering custodial fathers. Our primary contribution is demonstrating differences in stock and flow measures, a difference that is especially important for our understanding of a variety of policy-relevant outcomes. As we examine policy effectiveness, we must understand how the use of stock measures may muddle changes in population demographics and caseload composition with changes in the outcomes of interest.
The authors wish to acknowledge the anonymous reviewers and participants in the 2022 fall conference of the Association for Public Policy Analysis and Management, particularly Robert Doar and Maretta McDonald, for helpful feedback on early versions of the manuscript. The authors also thank James Spartz for his editorial and graphics assistance.
A difficult issue is determining appropriate terms that recognize the various types of parenting that occur postseparation. “Custodial” and “noncustodial” parents can refer to the parent with custody, but some have different arrangements for physical custody (with whom the child lives) and legal custody (who makes major decisions), and some parents share custody (either physical or legal). One resolution is to use the labels “custodial” and “noncustodial” depending on whether one owes or is to receive child support, but this leaves questions about the best term for parents who do not have agreements for child support. Some research uses “resident” and “nonresident” parents, but this is also difficult when children live half-time (or substantial portions of time) with each parent. In recognition of these difficulties, in this research note we use the imperfect terms “custodial” and “noncustodial,” following the U.S. Census Bureau’s terminology (Grall 2020).
One source of data on child support is the administrative records of the child support program (the Title IV-D program), summarized in annual reports to Congress (see Office of Child Support Enforcement 2022). However, these data include only those served by the program, and we are interested in all families who may be eligible for child support. The primary source of data on this population is from the U.S. Census Bureau, and that is what we use here.
For consistency, we use CPS-CSS data available via IPUMS (Flood et al. 2022). Some previous studies have used CPS-CSS data made available via NBER, which include a number of additional variables, including flags for imputed measures, but these data are not available after 2014.
The CPS-CSS data include those with children younger than 21; however, in some states, child support obligations end at age 18. A relatively small proportion of the parents in our ever-married sample have nonmarital coresident children. Owing to data limitations in identifying these parents in publicly available data from recent years, we include these parents as ever-married parents for consistency across years. Our estimates for ever-married parents are robust to excluding currently married parents from our analyses as well as controlling for marital status.
Note that individuals may have a legal agreement for one child and an informal agreement for another; this occurs primarily when there have been multiple partners. This complication is ignored in these data.
In the publicly available IPUMS, this variable is present only prior to 2012.
We use several checks to test assumptions about assigning flow years (i.e., the year a case becomes potentially eligible for child support). If we ignore information on the year of the order and instead assign potential eligibility years on the basis of divorce year (when available) or age of oldest, we find that nearly three quarters of cases would be dated within one year. Additionally, we use Stata’s hot deck imputation procedure to impute flow year for cases without an agreement by considering number of children, age of children, education level, and marital status. We find similar patterns of results when testing these alternate assumptions. Finally, because age of youngest child could be a better indicator for the end of the union (though not the end of an earlier union), we estimate results using age of youngest child rather than age of oldest child. Results are generally robust to this specification, though we find a slight decrease in legal orders for never-married custodial mothers in recent years (still not statistically significant), and increases for ever-married mothers are smaller in magnitude for the two most recent periods.
For never-married mothers, the proportion of cases with legal agreements peaks at around 46%, though the proportion with a legal agreement with payment due peaks two years earlier, at 41%.
The increase in informal agreements and legal agreements with no payments due may reflect, in part, an increase in shared physical custody (measured by mothers reporting the child spent the equivalent of at least three months in the past year with the other parent). Shared physical custody is associated with an increase of 3–4% in the probability of an informal agreement or a legal agreement with no payment due. Almost one third of ever-married mothers report shared physical custody (regardless of agreement) for flow years 2014–2018, an increase from one fifth of mothers in flow years 1993–1997. Mothers with shared physical custody account for one quarter of those with an agreement but no payment due, and 37% of those with informal agreements for flow years 2014–2018.
Noncustodial parents may also be ordered to pay medical support—either cash medical support or health insurance.