Abstract
Changes in US social provisioning from 1996 to 2018 profoundly affected the well-being of single-mother families, with notable consequences for racial inequalities in poverty alleviation. Using an institutional approach to racial disparities alongside a sequence-independent decomposition analysis, we group policies and programs into five redistributive mechanisms—federal transfers, state transfers, federal income taxes, state income taxes, and federal payroll taxes—and estimate their effects on poverty reduction for White, Black, and Latino single-mother families. We find that although poverty declined, racial disparities in poverty rates persisted. State transfers had the greatest impact on poverty alleviation but declined substantially in effectiveness after 2012, while federal taxes, particularly for Latino families, became more effective. This study underscores enduring racial disparities in the reduction of single-mother poverty while highlighting the complex interplay of policy design, decentralization, and poverty outcomes.
In the years since Kathryn Edin and Laura Lein’s Making Ends Meet (1997) documented the lived reality of single mothers struggling to survive economically in the late 1980s and early 1990s, social provisioning for low-income families in the United States has been thoroughly reorganized. Most substantially, the landmark Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (PRWORA) ushered in a reconfigured safety net, both increasingly work-based, conditioning income assistance on workforce participation, and paternalistic, embedding new behavioral requirements and punitive consequences into social policy (Soss et al. 2011). PRWORA also facilitated greater safety-net decentralization with the shift from Aid to Families with Dependent Children (AFDC) to Temporary Assistance for Needy Families (TANF), replacing a federal entitlement with block grants that permit subnational governments’ greater responsibilities over financing, administration, and rulemaking of cash assistance (Bruch et al. 2018). Described by some as a shift to a “work-based safety net” (Heinrich and Scholtz 2009), income support has been largely replaced by the provision of in-kind benefits and services, namely food assistance programs and child health insurance (Bitler and Hoynes 2010; Bruch et al. 2026, this issue; Schmidt et al. 2025). Additionally, support for low-income families has increasingly been accomplished through the income tax system, particularly through federal and, to a lesser extent, state income tax credits specifically targeted at working families with children (Howard 1997; Hoynes 2019; Maag et al. 2023).
Female-headed families with children, the primary target population of welfare reform, have been the focus of considerable attention in studies of the effects of these policy reconfigurations. Scholars have examined the policy and program contours of how the single-parenthood “penalty” has changed over time (Brady et al. 2024; Brady et al. 2017) and how this penalty varies across the United States (Laird et al. 2018; Nicholson 2022). Given their emphasis on employment, previous research has also documented the labor-force participation, poverty, and income-security impacts of these policy shifts (Blank 2002; Bitler and Karoly 2015; Fox et al. 2015; Hoynes 2009; Ziliak 2015).
In this article, we build on the work of Edin and Lein (1997) by unpacking how and to what degree the United States’ social provisioning system has assisted single mothers over the nearly three decades since welfare reform. Our research disaggregates the impact of transfers from taxes to gauge how these changes have affected low-income single mothers. In so doing, we bring together the insights of safety-net and social-policy research with a growing body of fiscal sociology scholarship on the redistributive function of tax instruments (Martin and Prasad 2014; O’Brien 2017; Schechtl and O’Brien 2024). Drawing on an institutionalist perspective, we classify and package income sources into distinct redistributive mechanisms defined by their policy tool (transfer or tax) and intergovernmental design (federal or state). Bundling components of the social provisioning system in this way allows us to connect different institutional designs to their distributional consequences while foregrounding the effect of policy decentralization. This approach to aggregating programs according to their institutional design features serves as a complementary middle ground between approaches that examine poverty reduction from individual programs and income sources and those that aggregate all transfers and taxes to compare pre- and post-redistribution.
Background
Social provisioning in the United States is increasingly stratified, providing individuals and families with unequal levels and kinds of assistance based on specific categorical eligibility criteria, income and asset tests, and a host of behavioral conditions, including participation in the paid labor force (Cardona et al. 2022; Moffitt 2015). Shifts in the policy instruments and designs of social provisioning since the mid-1990s, including those instituted by PRWORA, reflect changes in perceptions of deservingness, responsibilities of citizenship, and economic interests in enforcing work (Clasen and Clagg 2007; Crabtree and Wehde 2023; Kreitzer et al. 2022; Piven and Cloward 1971; Soss et al. 2011).
In the wake of these substantial policy changes, scholarly research has explored how various groups have been differentially impacted (Moffitt 2015). Children, widely viewed as a particularly “deserving” group, have received special attention (National Academies of Sciences, Engineering, and Medicine 2019, 2023b). Motivated by concerns over who is left behind or no longer eligible for social support, scholars have explored whether these reforms have led to increases in extreme or deep poverty and have identified those most affected (Brady and Parolin 2020; Edin and Shaefer 2016). Similarly, using detailed household survey data, scholars have documented growing disparities among poor families, particularly between the employed and the not employed, as well as between families with and without children (Jackson and Fanelli 2023; Hoynes and Schanzenbach 2018; Moffitt 2015). Several of the policy changes mentioned earlier have been instrumental in fostering these inequities: the conditioning of cash assistance on employment; the 1993 expansion of the Earned Income Tax Credit (EITC), which was targeted at low-income families and designed to promote employment; the expansion of food assistance in the form of the Supplemental Nutrition Assistance Program (SNAP) and school-based meal programs; child health insurance (both Medicaid and the Children’s Health Insurance Program [CHIP]); and expansions of the Child Tax Credit (CTC). Together, this new package of supports reduces poverty to a much greater degree for families with children and for those with workers present (Brady and Parolin 2020; Bruch et al. 2023; National Academies of Sciences, Engineering, and Medicine 2023b; Parolin, Desmond, et al. 2023; Wimer et al. 2020).
As the primary target population of welfare reform, the impacts of these policy changes on single-mother families with children have received a great deal of attention in relation to their participation in paid labor and continued receipt of safety-net assistance (Blank 2007; Ziliak et al. 2000). Scholars have also examined the experiences of those who are “disconnected” (neither working nor receiving assistance) (Blank and Kovak 2009) and those in extreme or deep poverty (Edin and Shaefer 2016). Other research has shown that the design of tax-based anti-poverty programs that have grown in prominence since the mid-1990s (such as the EITC and CTC) tends to exclude or underserve the poorest single-mother households (Brehm and Malkova 2023; Hoynes and Patel 2018). Taken together, scholars have documented that single parenthood in the United States tends to be associated with a substantial “penalty” (Brady et al. 2017).
We know less about how the composition of support for single-mother families, and the distributional consequences of those supports, has varied by race and ethnicity in the post–welfare-reform period; however, there are several reasons to expect inequities. We know, for example, that many of the changes in policy instruments and program designs since the mid-1990s reflect the continued salience of racialized controlling images of target populations and the endemic nature of structural racism (Baker 2022a, 2022b; Collins 2000; Schneider and Ingram 1993; Soss et al. 2011; Williams 2021; Williams and Baker 2021). We further know that racist ideologies have historically been embedded in, and continue to shape, not only politics and policy design but also the administrative state that puts policies into practice (Alexander and Stivers 2020; Katznelson 2005; Williams and Duckett 2020).
Existing research on poverty rates among single parents and children documents the racial and ethnic disparities that result from these historical and contemporary racial dynamics and underscores the continued importance of attending to racial differences in anti-poverty and redistribution policies. Despite the widespread focus on children as an especially deserving target of social policies, child poverty in the United States has remained high, while the magnitude of poverty reduction across racial and ethnic groups has been markedly uneven. Black and Latino children have higher prevalences of poverty than White children, and while Black and Latino children have experienced sizeable declines in poverty since the mid-1990s, their poverty rates remain three times as high as those of White children (Baker 2022b). Although some scholars and policymakers maintain a concern with the role of family structure, and single motherhood in particular, in fostering racial gaps in child poverty, numerous recent studies have demonstrated the role of politics and policy (Baker et al. 2022; Brady et al. 2024; Parolin 2021), the enduring legacy of historically racialized institutions (Baker 2022a; Sáenz and Quintanilla-Muñoz 2025), as well as barriers to employment for immigrants and their families (Thiede et al. 2021) as major drivers of these disparities. Cross-state analyses have demonstrated, for example, how variation in states’ social policies, differentially magnifies the risks of poverty for Black and Latino children to a greater degree than White children (Laird et al. 2018).
Contributions
This article builds on these literatures by advancing an institutional approach to the analysis of racial disparities in single-mother poverty. We foreground the role that rules and regulations (policies and programs) play in shaping distributional outcomes (Brady 2009; O’Connor 2001) in ways that both reflect society’s interests, composition, and the relative power of groups, as well as how these outcomes feed back to reinforce these factors (Esping-Andersen 1990). This approach acknowledges the multiple ways that racialized and gendered assumptions about women, motherhood, labor, and care are reflected in the policy design and administrative practices of social policies (Fox 2012; Gordon 1994; Lieberman 1998; Quadagno 1994).
In applying this approach, we make two main conceptual contributions: first, we map institutional features and aspects of poverty governance (policy tools and intergovernmental arrangements) onto redistribution mechanisms. Second, we incorporate insights from fiscal policy scholarship, broadening our treatment of tax components including income taxes, both their credits and liabilities, as well as workers’ share of payroll taxes. Scholars examining racial inequities in social provision vary in the extent to which they incorporate the decentralized nature of transfers and taxes in the United States. While many scholars leverage spatial and temporal variation across states to identify the impact of specific policies (Bitler et al. 2017; Laird et al. 2018; Nicholson 2022; Nolan et al. 2016) and document racial disparities in the receipt and impact of social assistance, social insurance, and taxes (Gaines et al. 2021; McDaniel et al. 2017; O’Brien 2017; Parolin, Cross, et al. 2023; Skandalis et al. 2022), fewer explicitly articulate the ways in which decentralization may be tied directly to racial inequality (Bruch et al. 2019; Michener 2019; Soss et al. 2011). Building on this work, here we foreground the impact of institutional decentralization, characterized by the devolution of aspects of program administration, financing, and rulemaking to subnational governments (Bruch et al. 2018) as an important component of United States’ social provisioning in two primary ways: first, by disaggregating between transfers and taxes and, second, by distinguishing programs that afford state actors substantial discretion from those that do not (state and federal transfers and taxes, respectively). To do this, we assess the impact on family incomes of five redistributive-policy mechanisms—federal transfers, state transfers, federal income taxes, state income taxes, and federal payroll taxes. By separating income sources in this fashion, we foreground how policy decentralization functions vis-à-vis different policy instruments (Bruch et al. 2023).
Our analyses begin with an examination of poverty rates among single-mother families in 2018, our most recent study year. Throughout the article, we estimate poverty using the Supplemental Poverty Measure (SPM). We explore how poverty reduction attributable to our five mechanisms varies across White, Black, and Latino single-mother families, and we focus on how the receipt and size of distinct social programs and tax credits vary across these three groups. The second part of the article explores variation in poverty and poverty reduction for White, Black, and Latino single-mother families over time, beginning in 1996 when PRWORA was enacted into law. We trace how the magnitude and sources of poverty reduction for single-mother families have changed since the mid-1990s, with attention to how patterns of redistribution correspond to the design characteristics of our five mechanisms and how these patterns of redistribution vary by race. Our findings show how social policies, once designed and implemented, have exclusionary or inequitable impacts across racial groups and, in so doing, exacerbate and maintain racial inequality.
DATA AND METHODS
We use the Current Population Survey’s Annual Social and Economic Supplement (CPS-ASEC), accessed via IPUMS (Flood et al. 2025) and analyze data from survey years 1995–2019. We construct three-year moving averages for our analyses of the composite years 1996 through 2018. Our selection of years allows us to examine changes in social provisioning for single-mother families over nearly a quarter century, from the passage of PRWORA, perhaps the most significant social welfare reform for single-parent families, until the onset of the COVID-19 pandemic and the dramatic changes to social welfare policy enacted in its immediate wake.
Our analyses focus on families headed by single mothers of working age. To define our sample, we made several decisions. First, we define families as all persons related to one another living at the same address, including coresidential partners and coresident children. This definition is consistent with the resource-sharing unit specified by the Supplemental Poverty Measure (SPM), which we use throughout our analyses.1 Second, to capture working-age families, we selected families where the head of the SPM resource unit was aged eighteen to sixty-four years old.2 Third, we define single mothers as those who report no spouse (coresidential or not), who do not live with a nonmarital partner, and who do not live with other adults. Excluding families with other adults allows us to focus on the impact of transfers and taxes specifically with respect to single mothers.3 We focus on female-headed single-parent families and define “mothers” as those who reside with one or more children under age eighteen.4 To examine racial-ethnic differences, we compare families with heads who identify as non-Latino White, non-Latino Black, and Latino.5
Measures
Using detailed income components provided by the CPS-ASEC, we construct a measure of pretax-pretransfer income (“market income”) and a measure of posttax-posttransfer income (“disposable income”). Our measure of market income is inclusive of wages and salaries, self-employment earnings, and farm earnings, as well as income from rent, dividends, and interest. We also include income from retirement, survivor, and disability pensions and annuities (other than from the OASDI program and the Veterans Administration), and income from friends and family, including alimony.6 Consistent with the SPM, we deduct medical out-of-pocket, work, and child support expenses from families’ market income.
We categorize transfers and taxes into five distinct redistributive mechanisms: federal transfers, state transfers, federal income taxes, state income taxes, and federal payroll taxes (Federal Insurance Contributions Act [FICA] taxes). We classify transfer programs as either federal or state according to whether programs have some degree of state discretion in their financing, administration, or rulemaking (Bruch et al. 2018).7 Two programs are categorized as federal transfers: OASDI and veterans benefits. State transfers include TANF, Unemployment Insurance (UI), General Assistance (GA), Workers’ Compensation (WC), Supplemental Security Income (SSI), child support, SNAP, school lunch subsidy, housing subsidy, energy assistance, and Special Supplemental Nutrition Program for Women, Infants, and Children (WIC).8 For TANF, SNAP, and SSI, we account for underreporting using a technique developed by Zachary Parolin (2019).9
Throughout our analyses, and following our earlier work (see Bruch et al. 2018; Bruch et al. 2023), we classify SNAP within the state transfers redistributive mechanism. Although SNAP is often considered a federal program, we classify it as a state transfer for several reasons. First, while states have relatively low levels of financing and policy rulemaking discretion, their moderate levels of discretion over the program’s implementation result in substantial variation in program participation across states (Bruch et al. 2018). SNAP benefits are federally funded, and their eligibility and benefit determinations are largely decided by federal rules; however, state and federal governments share financial responsibility for the program’s administration, which is conducted at the state, county, and local levels (Hoynes and Schanzenbach 2018). Moreover, the degree of state administrative discretion and flexibility has grown over time. The passage of PRWORA and other subsequent federal regulatory changes, such as the Farm Security and Rural Investment Act of 2002 and the Food, Conservation, and Energy Act of 2008, bolstered state discretion to implement strategies aimed at improving benefit access, determining exemption policies, and administering employment and training support programs. Recent work employing the administrative burdens framework has demonstrated that the wide variability in administrative rules affects program access and results in wide cross-state variation in recipiency (Herd and Moynihan 2018; Hertel-Fernandez 2024; Fox et al. 2023; Parolin, Cross, et al. 2023; Waxman and Joo 2019). For example, in 2022, according to the USDA, recipiency (the percentage of the population receiving SNAP benefits) varied across the states by a factor of five—from 5 to 25 percent (Jones 2024).
Federal and state income taxes, as well as payroll taxes (the employee-contribution portion of FICA taxes), are derived from estimates generated by the National Bureau of Economic Research (NBER) TAXSIM model for each family, which calculates total federal and state tax liabilities and credits (including the federal and state Earned Income Tax Credits and the Child Tax Credit) from survey data (Feenberg and Coutts 1993).10 We represent tax liabilities as negative income values, while tax credits are sources of income and therefore are represented as positive values.
Analytic Strategy
Our analyses estimate poverty reduction or impoverishment attributable to the five transfer and tax mechanisms. To measure poverty, we apply the US Census Bureau’s SPM thresholds using the market- and disposable-income definitions described earlier to estimate both the proportion of families that are market-income poor and the proportion of families in poverty after accounting for all transfer and tax mechanisms. We use SPM thresholds throughout our analyses, both because they incorporate a broader array of consumer expenditures and because these expenditures are geographically adjusted.11 All of our analyses use weights provided by the SPM to obtain nationally representative estimates.
To estimate poverty reduction attributable to each redistributive mechanism, we use a sequence-independent decomposition (Azevedo et al. 2012). This decomposition employs a Shapley value-based calculation to estimate the marginal poverty reduction, operationalized as changes in the poverty rate attributable to each mechanism irrespective of the order in which they are included among the decomposition’s components (Shorrocks 2013).12
To describe changes in poverty (either reductions or increases), we use three measures. First, we estimate the overall absolute poverty reduction as the raw difference between the market- and disposable-income poverty rates (that is, the percentage-point change). Second, we estimate the proportion of the overall absolute reduction attributable to each mechanism. Third, we estimate changes in poverty in relation to market-income poverty by calculating the proportion of market-income poverty reduced (or increased). Because income taxes include credits, they can increase or decrease families’ disposable income. The same is not true of workers’ payroll taxes, however, because they do not include credits and can only decrease families’ disposable income.13
To describe the poverty reduction or impoverishment attributable to each redistributive mechanism, we also calculate the rates of receipt and average values for each mechanism. The rates of receipt for transfer mechanisms reflect primarily self-reported information for the family,14 while the rates of receipt for tax mechanisms reflect estimates for the family based on tax simulations (as discussed earlier).15 For tax mechanisms, we disaggregate families into net-paying (the sum of liabilities exceeds those of credits), neither paying nor receiving, and net-receiving (the sum of credits exceeds those of liabilities). FICA taxes do not possess a credit component, so for this mechanism families are either paying or not paying. Average values for each tax mechanism as well as for each federal and state transfer program are calculated as the average among market-income-poor families receiving income from each tax mechanism or transfer program and are therefore always calculated among nonzero amounts.
RESULTS
In this section, we present the findings from our decomposition analyses in two parts. We focus first on 2018, our most recent study year, before exploring fluctuations in rates of poverty and patterns of poverty reduction over time from 1996 to 2018. Each part proceeds by first describing patterns of poverty and poverty reduction, illustrating variation among White, Black, and Latino single-mother families. We then disaggregate total poverty alleviation into magnitudes of reduction attributable to our five mechanisms, while exploring variation in magnitudes across groups. Our analyses of the state transfer mechanism unpack how variation in poverty reduction is traceable to differences in average values and rates of receipt for the mechanism’s component programs.
Poverty Reduction and Impoverishment Among Single-Mother Families, 2018
In 2018, market- and disposable-income poverty rates varied dramatically among White, Black, and Latino single-mother families (see table 1). More than half of Black and Latino families had income below the poverty line when only market-income sources were included (0.602 and 0.645, respectively). White families had a much lower market-income poverty rate of 0.447. After transfers, taxes, and work expenses were taken into account, poverty was reduced substantially for all three groups. However, comparing disposable-income poverty rates (accounting for all transfers and taxes), Latino families had the highest poverty rates (0.330), followed by Black families (0.285) and White families (0.219). Latino and Black families had the highest rates of disposable-income poverty even though they experienced greater levels of absolute poverty reduction compared with White families (0.316 and 0.317, respectively, compared with 0.228).
Poverty Reduction or Impoverishment by Policy Mechanism for White, Black, and Latino Single-Mother Families, 1996–2018
For all groups, state transfers and federal income taxes had the greatest poverty-reduction impacts, with more modest poverty reductions attributable to federal transfers and a small role attributable to state income taxes. Pulling in the opposite direction, payroll taxes (FICA) served to increase poverty for all groups. While this general pattern was observed across all three racial-ethnic groups, there were a few notable differences.
The poverty reduction attributable to state transfers was the largest for all three groups with 60–66 percent of market-income poverty being reduced. Almost all market-income-poor Latino, Black, and White single-mother families received one or more types of state transfers (98–99 percent), but they received markedly dissimilar total average values ($9,400 for White, $10,000 for Latino, and $11,100 for Black families) and experienced reductions in poverty in line with these differences (a 14-percentage-point reduction for White, a 20-percentage-point reduction for Latino, and a 21-percentage-point reduction for Black families).
Disaggregating state transfers into their component programs reveals three notable patterns (see table 2). First, the high rates of state transfer receipt are largely driven by the fact that most market-income-poor single-mother families receive assistance from two food assistance programs (SNAP and the National School Lunch Program [NSLP]) with rates of receipt ranging from 87 to 93 percent for SNAP and from 67 to 75 percent for NSLP. Second, for White families, child support is the next-most frequently received source of income (35 percent) and on average they receive $5,600 compared with Latino families (24 percent) who receive $4,900, followed by Black families who report the lowest rates of receipt (20 percent) and average amounts received ($3,700). Third, the receipt of income support transfer programs—TANF, SSI, UI, WC, and GA—is broadly similar across White, Latino, and Black single-mother families with markedly low rates of receipt but widely varying average levels of assistance. Even the two programs with the highest rates of receipt (SSI and TANF) are received by fewer than one in five market-income-poor families (14–15 percent for SSI and 9–15 percent for TANF). The average level of support provided by income-transfer programs is also remarkably low, especially relative to income levels needed to escape poverty—ranging from $3,300 to $4,600 for TANF and GA, from $4,300 to $4,600 for UI, and from $8,300 to $9,000 for SSI. The one exception is the Workers’ Compensation program; however, less than 1 percent of market-income-poor single-mother families receive this form of assistance.16
Recipiency Rates and Average Values for White, Black, and Latino Single-Mother Families, 1996–2018
Federal income taxes provided the second largest poverty reduction across all three groups, ranging from a 9-percentage-point decline in market-income poverty for White single-mother families to a 14-percentage-point decline for Latino families (table 1). A higher percentage of Black and Latino single-mother families received income (a credit) from federal income taxes (68 and 74 percent, respectively) compared with White families (64 percent, see table 3), but the average amount received was similar (ranging from $4,800 to $5,200, see table 1). A substantial portion of the poverty reduction attributable to federal income taxes came from EITC. Looking at the largest federal income-tax credit, the EITC, Latino families had the highest rate of receipt (74 percent) compared with 67 percent of Black and 64 percent of White single-mother families (table 2).
Rates of Recipiency, Payment, and Neither Payment of Recipiency for Tax Mechanisms for White, Black, and Latino Single-Mother Families, 1996–2018
While federal transfers account for a much lower proportion of market-income poverty reduction among single-mother families, there were considerable differences in the overall reduction attributable to this mechanism. For White single-mother families, market-income poverty was reduced by 4 percentage points, compared with only 1–2 percentage points for Latino and Black families, respectively (table 1). These differences reflected the higher rate of receipt of federal transfers by White families (16 percent) compared with Black (10 percent) and Latino families (6 percent) and the higher average value received ($17,900 compared with $12,100 and $10,700, respectively).
Disaggregating federal transfers into OASDI and Veterans’ Benefits showed that White single-mother families had higher rates of receipt and average amounts for both (see table 2).17 For example, 16 percent of White families received income from OASDI compared with 6 percent for Latino families and 9 percent of Black families, with an average amount received of $17,400 compared with $10,900 and $11,400, respectively. Veterans’ Benefits were received by very few market-income-poor single-mother families (less than 1 percent for all three groups); however, the average amount received followed the same pattern as OASDI, with White families receiving a greater average benefit ($23,500) compared with Black and Latino families ($22,400 and $5,400, respectively).18
State income taxes reduced poverty very little for single-mother families—generally less than 1-percentage-point reduction in market-income poverty. Like federal income taxes, Latino families had the highest rate of receipt of income from this redistributive mechanism (40 percent) compared with White and Black families (32 and 23 percent, respectively; see table 3). Our measure of state income taxes includes all state-level EITCs. However, state-level EITCs tend to be modest, and the credit is refundable in just half of the US states.19
Finally, we found that payroll taxes (FICA), increased poverty between 3–4 percentage points for all three groups of single-mother families (table 1). Latino single-mother families had the highest rate of FICA payment (74 percent), followed by Black and White families (68 and 64 percent, respectively; see table 3). It is important to clarify that the poverty-increasing effects reported here referred to the contemporary period; this did not take into account the fact that, in the future, these single mothers may receive poverty-reducing cash transfers financed by their FICA contributions.
Single-Mother Poverty Reduction and Impoverishment Since PRWORA, 1996–2018
Rates of market- and disposable-income poverty have declined for Latino, Black, and White single-mother families from 1996 to 2018 (see figure 1 and table 1). However, the declines were much more precipitous for Latino and Black families. Declines in market-income poverty for Latino families were the sharpest, dropping 12 percentage points (from 76 percent in 1996 to 64 percent in 2018), with disposable-income poverty declining 21 percentage points (from 54 percent in 1996 to 33 percent in 2018). Black families experienced similar declines in poverty, with market-income poverty decreasing 10 percentage points (from 70 percent to 60 percent between 1996 and 2018) and disposable-income poverty declining 18 percentage points (from 46 to 28 percent). White families, on the other hand, while having much lower rates of market- and disposable-income poverty throughout the period, experienced the smallest declines in poverty (from 47 percent market-income poverty in 1996 to 45 percent in 2018 and from 29 percent disposable-income poverty in 1996 to 22 percent in 2018).
Racial and ethnic disparities in poverty reduction widened substantially over the period between 1996 and 2018. White, Black, and Latino single-mother families experienced the lowest levels of absolute poverty reduction in 1996 (18, 23, and 21 percentage points, respectively) and in 2002 (17, 19, and 18 percentage points, respectively). By 2012, Black and Latino single-mother families experienced the greatest, and most similar absolute poverty reductions (37 and 35 percentage points, respectively). These reductions fell slightly by 2018 (to 32 percentage points). White single-mother families experienced the smallest absolute poverty reductions throughout the period (from a low of 17 percentage points in 2002 to 28 percentage points in 2012). However, when overall poverty reduction was measured as the percent of market-income poverty that was reduced, White single-mother families experienced the greatest poverty reduction—ranging from 38 to 57 percent of market-income poverty reduced between 1996 and 2012. From 1996 to 2018, Latino and Black single-mother families experienced similar percentages of market-income poverty reduction (28 and 34 percent, respectively, in 1996; and 48 and 56 percent in 2012; and 49 and 53 percent, respectively, in 2018).
The relative contributions to poverty reduction and impoverishment attributable to each of the five transfer and tax mechanisms remained largely consistent from 1996 to 2018. State transfers provided the greatest reduction, followed by federal income transfers and taxes, across the entire period, while on the impoverishment side, federal payroll taxes increased poverty (see figure 2). However, the magnitude of the contributions attributable to these mechanisms in terms of absolute poverty reduction (or increase) and the proportion of the total poverty reduction shifted over time.
For single-mother families, state transfers accounted for the greatest magnitude in poverty reduction for Latino, Black, and White families from 1996 to 2018 (see table 1). In 1996, state transfers reduced market-income poverty by 12–18 percentage points for these groups, accounting for 69–77 percent of the total poverty reduction. The impact of state transfers declined in 2002, but then increased for all groups during the Great Recession, reaching a high point in 2012 before declining dramatically through 2018, though remaining higher than in 1996 (see figure 2, panel A). However, the rate of decline since 2012 varied across the three groups. For Black families, poverty reduction attributable to state transfers fell by 6 percentage points from 2012 to 2018; for Latino families, the decline was 4 points (from 27 and 24 points to 21 and 20 points, respectively), while the decline for White families was 5 points (from 19 to 14 points). The result was a divergence since 2012 in the degree of absolute poverty reduction from state transfers across groups, between Black and Latino single mothers, on the one hand, and White single-mother families, on the other.
Disaggregating state transfers by examining the rates of receipt and average amounts of specific programs provides greater insight into these shifts in poverty reduction over time (see table 2). The traditional cash-assistance transfer program for single-mother families, TANF (and pre-1997, AFDC) showed the most dramatic decline in rates of receipt and average amounts received (plummeting from 54–68 percent receiving in 1996 to 9 to 15 percent by 2018, with average benefits ranging from $4,800 to $7,700 in 1996 declining to between $3,200 and $4,100 in 2018). Across the entire period, the other income support programs (UI, SSI, GA, and WC) were received by relatively few single-mother families. The SSI program provided an increased amount of support from 1996 to 2018 through higher rates of receipt and average amounts received, but this increase was concentrated among White single-mother families (increasing from 11 to 15 percent receiving assistance through this program and increasing from an average benefit of $6,800 to $8,300).
The high-water mark for poverty reduction attributable to state transfers in 2012 was due to both an increase in the rates of receipt and average benefit amounts for two programs—SNAP and UI (see table 2). In 2012, SNAP was received by 86 to 95 percent of market-income-poor single-mother families, with an average benefit ranging from $4,400 to $5,200; UI, on the other hand, while received by a much lower percentage of these families (8–13 percent), provided higher average benefits ($6,800–$7,300).
Child support and housing subsidies were the two programs that had the starkest racial and ethnic disparities in rates of receipt and average benefits from 1996 to 2018 (see table 2). Across the entire period, White market-income-poor single-mother families received child support at a much higher rate (ranging from 35 to 40 percent) compared with 20 to 25 percent for Black families and 19 to 26 percent for Latino families, and the average amount received for White families was also higher (ranging from $5,300 to $6,500) compared with $3,500 to $4,200 for Black families and $3,800 to $5,200 for Latino families. Black single-mother families received housing subsidies at a higher rate (ranging from 41 to 46 percent) compared with 25 to 31 percent of Latino families and 20 to 22 percent of White families.
Federal taxes were the second-most substantial mechanism for reducing single-mother family poverty, again, due in large part to the EITC (see table 1). The poverty reduction attributable to federal taxes steadily increased throughout this period (see figure 2, panel B). In 1996, absolute poverty reduction attributable to federal taxes ranged between 6 and 7 percentage points, and by 2018 this increased to between 9 to 14 percentage points. Latino families experienced the most dramatic poverty reductions attributable to federal taxes between 1996 and 2018 (from 6 to 14 percentage points, accounting for 26 and 43 percent of the total poverty reduction for those years). For White families, the estimated rates of receiving the largest federal tax credit (the EITC) declined from 1996 to 2018 (from 67 to 64 percent). Rates of receipt increased for Black and Latino families (from 55 to 67 percent and from 50 to 74 percent, respectively). However, the average amount received increased during this period for all three groups (from $2,700 to $3,600 for White and Black families, and from $2,600 to $3,800 for Latino families). Though the average credit was far lower, the CTC also grew in value and rates of recipiency for all three groups over this period.
While far less substantial than state transfers or federal taxes, the contribution of federal transfers to poverty reduction for market-income-poor single-mother families was relatively consistent from 1996 to 2018 (see figure 2, panel A). White single-mother families experienced the largest poverty reductions attributable to federal transfers from 1996 to 2018 (3 to 4 percentage points across the entire period, comprising between 10 to 16 percent of their total poverty reduction). In comparison, the poverty reduction attributable to federal transfers for Black and Latino families ranged from 1 to 2 percentage points and between 4 and 10 percent of their total poverty reduction. The higher poverty reduction for White families can be attributed to their higher rates of receipt of OASDI (ranging from 14 to 16 percent between 1996 and 2018) compared with 9 to 10 percent for Black families and 6 to 10 percent for Latino families, and to the higher average amounts received (ranging from $14,100 to $17,400 between 1996 and 2018 compared with $9,100 to $11,400 for Black families and $10,000–$13,600 for Latino families).
The magnitude of poverty reduction attributable to state income taxes increased between 1996 and 2018; however, it still provided the lowest poverty reduction of all mechanisms (see figure 2, panel B). This increase in poverty reduction attributable to state income taxes was most substantial for Latino families, increasing from 0.0 to 0.5 percentage points, while for White and Black families the rate of poverty reduction hovered around 0. This small poverty-reduction impact was due to both a high percentage of poor families neither paying nor receiving state income taxes, which ranged from 57 to 86 percent for a given year and group (see table 3), and the small average values received, which ranged from less than $100 to just over $600 (see table 1).
While federal and state income taxes (and federal and state transfers) reduce poverty among market-income-poor single-mother families, federal payroll taxes served to further impoverish these families from 1996 to 2018 (see figure 2, panel B). In the case of FICA, the impacts were relatively similar over time (ranging from a 2- to 3-percentage-point increase in poverty for White and Black families, and a 3- to 4-percentage-point increase for Latino families) (see table 1). However, the percent of families paying these payroll taxes increased for Black and Latino families (from 55 and 50 percent in 1996 to 68 and 74 percent in 2018, respectively), while it decreased for White families (from 68 to 64 percent) (see table 3).
DISCUSSION AND CONCLUSION
Returning to the motivation of this paper, what do our findings tell us about how single mothers have been assisted since the publication of Making Ends Meet? Our analysis documents three key takeaways. First, even after accounting for a multitude of income sources that low-income mothers pull together to make ends meet, too many of them remain in poverty (between 22 and 33 percent of single-mother families by 2018). Second, overall poverty reduction has increased from 1996 to 2018, particularly for Latino and Black single mothers, and this has resulted in a reduction in racial-ethnic disparities in disposable-income poverty. Yet, as other scholarship has documented, despite this progress, substantial racial-ethnic poverty gaps remain (Baker 2022b). By our estimates, nearly one-third of Black single-mother families (29 percent) live in poverty, compared to just over one-fifth of White single-mother families (22 percent). Third, we find that throughout this period, state transfers (including SNAP) are the mechanism of poverty reduction that plays the most significant role in decreasing single-mother family poverty, particularly among Black and Latino families. Federal taxes also, to a substantial degree, lessen single-mother poverty, with Latino families increasingly benefiting from the EITC and CTC over time.
This income-focused analysis should be understood within the broader context of how single-mother families are making ends meet and how these strategies have changed since welfare reform, while acknowledging its limitations of our analysis. Unlike other research, which is based on detailed financial information, such as that presented in Making Ends Meet or in other innovative articles in this double issue (see Pilkauskas and Bruey, 2026), our income-based analysis relies on the income sources detailed in the CPS-ASEC. Thus, our study omits some potential income sources (see Hill et al. 2026, this issue, regarding state paid leave) and suffers from some known weaknesses in capturing others (including, for example, benefit underreporting in the means-tested programs). More importantly, income-based analyses such as ours, regardless of the data collection instrument, cannot fully capture several important shifts in how low-income single-mother families are surviving.
One of the most significant shifts in social policy over the past three decades has been the move away from direct income support toward the provision of in-kind benefits and services, such as childcare subsidies and the provision of child health insurance (see Ananat et al. 2026, this issue; Bruch et al. 2026, this issue; Kwon et al. 2026, this issue). While we can observe the decline of cash assistance in terms both of rates of receipt and diminished benefit amounts, we cannot directly capture the expense-reducing impacts of many social supports and services (see Gonalons-Pons et al. 2026, this issue). Similarly, we only partially capture related changes, such as the increased focus on child support enforcement and collections. Although we observe rising child support payments to poor single-mother families, we lack data on the complex income and social dynamics between custodial and noncustodial parents that more robustly capture how parents support their children (Dwyer Emory et al. 2026).
Another important shift that we cannot capture is the changing terms of receiving assistance. Policy changes stemming from increased conditionality, stricter eligibility criteria, expanded sanctions for noncompliance (Soss et al. 2011), and an increase in administrative rules and procedures that impose burdens on applicants (Herd et al. 2023) have made the receipt of assistance more challenging. Though we may be able to observe these changes indirectly, to the extent that making assistance more challenging or costly to receive shapes recipiency rates, these paternalistic and punitive policy shifts have been vividly documented in qualitative work that details their experiential dimension (see Hughes et al. 2026). Finally, reflecting the residual and continually contracting nature of United States’ social provisions is the shift toward more liberal access to and increasing reliance on credit to smooth consumption (Prasad 2012). As social policy designs increasingly fail to align with the reality of new economic and social risks facing low-income families (Hacker 2006; Meyers et al. 2011), borrowing is likely to become a more common strategy for coping with income volatility, financial instability, and precarity (Berger, Brown, et al. 2024; Dodini et al. 2024; Laprise and Wiedemann 2025; Rhodes et al. 2026).
It is also important to acknowledge how our sample selection shapes our findings. First, our focus on single-mother families without coresident adults limits our ability to examine the broader range of strategies that economically vulnerable single mothers use, particularly when faced with a limited safety net. These strategies include receiving informal economic support from friends and family (Harknett 2006; Kalil and Ryan 2010), choosing to live in multigenerational household arrangements, and by doubling up (see Harvey 2025). Black and Latino families, particularly those with low incomes, are more likely than White families to live in such arrangements (Cohen and Casper 2002; Richard et al. 2022). While our analyses compare only families with similar compositions, future research should expand our work to include single-mother families living in complex households and compare those analyses with the findings we present here. Second, given the space limitations of this study, we were unable to examine how the immigration status of single mothers’ conditions their access to our redistributive mechanisms. As other scholars have shown, the risks of poverty are higher for recent Latino immigrants (Theide et al. 2021), and so future research should explore how immigration status shapes poverty reduction from the forms of social provisioning we have analyzed here.
Despite these limitations, our findings highlight the crucial role of race and ethnicity, and policy design, in shaping how single-mother families have made ends meet since welfare reform three decades ago. The institutional features of social provisioning in the United States are not race-ethnicity neutral; they reflect and reinforce hierarchies rooted in racialized and gendered assumptions about work, family, and deservingness. As our results and prior scholarship suggest, decentralization plays a significant role in reproducing racial inequality. Given the continued importance of state transfer programs in reducing poverty among single-mother families, future policy decisions must recognize the challenges of decentralized social provision. Efforts to support low-income single mothers should account for the complexity and variation in federal and state arrangements, particularly how state discretion in rulemaking, financing, and administration shapes distributional outcomes.
APPENDIX
ASEC Income Components and SPM Expenditures
FOOTNOTES
↵1. The SPM resource-sharing unit expands the Census’s definition of a family to include coresidential partners, foster children under age twenty-two, related children older than age seventeen, as well as any coresident children aged fifteen and younger (Provencher 2011; Wimer et al. 2021). Research comparing the OPM and SPM over time find the SPM’s estimated poverty rates to be lower, due in part to the inclusion of coresidential partners within the resource-sharing unit (Fox et al. 2015). Moreover, recent research suggests that families and SPM units are the units at which individuals are sharing economic resources, especially within low-income households, and even when families are doubled up within the same household and sharing some costs (Berger, Cancian, et al. 2024; Harvey 2025). We refer to SPM resource-sharing units as families throughout.
↵2. While the US Census Bureau provides no uniform definition of this category, the SPM unit head corresponds to the first person in the roster of household members interviewed by the Current Population Survey interviewer. The ordering of each household roster is enumerated by a household line number. In cases where there is only one SPM unit within a household, the SPM unit head is equivalent to the householder or household head and has the lowest line number for that household. Typically, the householder is assigned the lowest line number, followed by the householder’s spouse or partner, the householder’s children, and other members of the family. In cases where there is more than one SPM unit in a household, additional SPM heads are identifiable as the individuals within that unit with the lowest line number (J. Creamer, personal communication, April 9, 2025).
We exclude units headed by retirement-eligible individuals (that is, adults ages sixty-five and older) from our analyses, given the substantial differences in eligibility for transfer programs and tax policies for working-age versus retirement-age individuals and families.
↵3. We examine only SPM units in which the head’s marital status is single, divorced, or widowed, and where the head has no coresidential same- or opposite-sex partner. Our analyses include only families headed by an unmarried and unpartnered female with children under age eighteen. This means we exclude all families that contain one or more coresidential adults who are not partnered with the head, including adult children. Our analyses therefore exclude multigenerational families that contain single mothers, and families in which single mothers reside with non-partner adults, including roommates or other relatives. We include only single-mother families without other adults because coresidential non-partnered adults, particularly individuals who are of retirement age, may possess tax and transfer resources that are otherwise unavailable to single mothers. Parents or grandparents, for example, may be eligible to receive retirement benefits, which have substantial anti-poverty impacts for retirement-age individuals and can also benefit children (Berger et al. 2022).
Moreover, we find that single mothers living without a coresidential non-partner adult, such as a parent or roommate, are more common than single mothers sharing a family with a non-partner adult. Using a pooled sample of working-age families from the 2017–2019 ASEC, we find that among working-age single-mother families, more than three-fourths (76 percent) contain no coresidential non-partner adults. In supplemental analyses, we separately estimate poverty rates and poverty reduction and impoverishment for single-mother families that include non-partner coresidential adults (see online appendix table A.5 at https://www.rsfjournal.org/content/12/1/67/tab-supplemental; all online appendix material can be found here). We find that these families have lower levels of market- and disposable-income poverty and, in general, lower magnitudes of poverty reduction. State transfers remain the most substantial mechanism in terms of absolute poverty reduction for all groups; however, federal transfers play a larger role in poverty reduction for families with coresidential adults. Particularly among White and Black families, federal transfers play a larger role in absolute poverty reduction than federal tax credits. This is likely because these families include older adults receiving Old-Age, Survivors, and Disability Insurance (OASDI) retirement income. Most adults of retirement age are eligible to receive the retirement benefits component of OASDI, known colloquially as “Social Security.” As other research has shown, Social Security substantially reduces poverty for qualified non-working age individuals (Meyer and Wu 2018).
↵4. We limit our sample to female single parents because women comprise the vast majority of single parents in the US and differ in many important ways from male single parents, particularly with regard to their poverty rates (Kramer et al. 2016; Wimer et al. 2021). Additionally, our focus is informed by the fact that, among families with children, single mothers have historically been the focal population of social provisioning.
↵5. We classify single-mother families based on the self-identified race of the head. Our focus is informed by the fact that non-Latino White, non-Latino Black, and Latino individuals are the largest racial and ethnic groups in the United States. We use the term Latino rather than Hispanic throughout. While we recognize that there are alternative gender-neutral terms, including Latinx and Latine, we choose not to use these terms because their pronunciation is less comfortable for many Spanish speakers, because they are most commonly used among English-speaking academic and economically advantaged audiences, and because Latino corresponds to the wording used by the Census Bureau. Although we do not include this in our empirical work, we acknowledge the importance of investigations of racial inequalities in poverty both between non-Latino White and Black populations and among Asian American, Native Hawaiian, Pacific Islander, and Native American groups, as well as comparisons within and between these populations, as other scholars have begun to do (Baker et al. 2022).
↵7. State discretion over financing, administration, or rulemaking varies considerably across different programs. For an extended discussion of the discretion level classifications for each program, see Bruch et al. (2018) and Bruch et al. (2023).
↵8. There have been changes over time in how programs are captured as separate income components in the CPS-ASEC. Beginning in 2014, alimony income was eliminated as a separate income component and merged with the “other income not otherwise classified” component.
↵9. Prior research finds that underreporting of SSI, SNAP, and TANF in the unadjusted CPS-ASEC data is substantial (Meyer et al. 2009; Parolin 2019; Stevens et al. 2018). Following Parolin (2019), we correct for underreporting using the Urban Institute’s Transfer Income Model (TRIM3), which imputes benefit values for the programs SNAP, SSI, and TANF and aggregates them into applicable families’ total state-transfer income.
↵10. We use TAXSIM version 35, the most recent version at the time of this writing, for our estimations. For more information on the TAXSIM program, see the official TAXSIM website, available through the National Bureau of Economic Research (2022). Although the Census provides estimates of tax credits, liabilities, and contributions using its microsimulation tax model, the first year the Census computed these estimates for SPM units was 2010 (Engel and Shantz 2024). Moreover, beginning with the 2004 CPS-ASEC, the Census instituted updates to its tax model to more accurately estimate state and federal taxes (O’Hara 2004). However, we find that the Census’s estimates prior to 2004 include only estimates of state tax liabilities and not credits. To both overcome this lack of tax data and draw on a single source of estimates for all three of our tax mechanisms, we use TAXSIM to estimate federal and state income taxes and liabilities, as well as the payroll tax contributions, for all years of our data. Comparing state income tax estimates from the Census tax model and TAXSIM for years 1994–2003 for working-age single-mother families with no other adults that fall below the SPM poverty threshold, we find that the two vary substantially. While TAXSIM estimates an average net tax credit for this group that ranges from $64 to $226 in 2003 inflation-adjusted dollars across years, the Census model’s estimates range from an average tax liability ranging from $167 to $621. The use of any tax simulation program, including TAXSIM, assumes all families eligible for tax credits claim them, which risks overestimating the anti-poverty impact of tax credits.
↵11. For years 2010–2019, we use SPM thresholds available from the CPS-ASEC, accessible via IPUMS (Flood et al. 2025). We apply the Historical SPM data from the Center on Poverty and Social Policy at Columbia University for data years 1995–2009 (Wimer et al. 2023). The issues of which unit to use for poverty measurement (households, families, or individuals), how to define income or household resources, whether and how to account for expenses or geographic cost-of-living differences, how to determine poverty thresholds, and whether these thresholds should be based on a relative or absolute poverty conceptualization remain debated in poverty scholarship. Many scholars use the SPM instead of the Census Bureau’s official poverty measure (OPM) because it takes into account a larger array of income sources including transfers and taxes, is adjusted for geographic variations in cost of living, and takes into account expenses including whether a family rents or owns its home (for example, Baker et al. 2022; Sullivan and Ziegert 2021). Our measure of disposable-income poverty incorporates transfers and taxes and, consistent with the SPM, we deduct medical out-of-pocket, work, and child support expenses from families’ market and disposable income prior to our poverty estimations. For an extended discussion of how the SPM is calculated, as well as its benefits and limitations, see National Academies of Sciences, Engineering, and Medicine (2023a).
↵12. Sequence-independent decompositions calculate the magnitude of each mechanism’s absolute poverty reduction by averaging each component’s impact across all the potential sequences by which that component could be added to a family’s income package. Applying a Shapley-value decomposition addresses a noted drawback of traditional additive decompositions, namely that the order in which components are added to the equation affects the estimated poverty reduction attributable to each component (Caminada et al. 2021). We employ a Shapley-value decomposition with the downloadable Additive Decomposition by Components (ADECOMP) application using the Stata statistical package (Azevedo et al. 2012). Our estimates adapt the ADECOMP procedure, designed to estimate changes in poverty between two points in time, to calculate changes in poverty across income definitions at a single point in time. We measure the average marginal poverty reduction attributable to the addition of transfers and taxes to families’ market-income.
↵13. The revenue from payroll taxes may, in the long run, be used to transfer resources to these families, but in the short run, they can only reduce disposable income.
↵14. The receipt and value of school-lunch and housing subsidies are imputed by the Census. The receipt and value of SSI, SNAP, and TANF are self-reported, but have been adjusted for underreporting using Parolin’s TRIM3-based method (see note 9).
↵15. Families may contain several tax units; however, we use the family as the tax unit and the filing status of the head when simulating taxes using TAXSIM. TAXSIM requires all tax units to possess a filer status of household head, married filing jointly, married filing separately, or dependent, and thus cannot calculate liabilities or credits for non-filers. Rather than exclude all families that are non-filers from our TAXSIM calculations, we assign all single-mother families the filing status of single. Our decision to include potential non-filers in our TAXSIM calculations may overestimate the poverty reduction attributable to federal and state income taxes.
↵16. Marked variation in average levels of support from WC (ranging $380–$35,100) is connected to very low rates of receipt; because fewer families receive this income, outlier values can skew estimates.
↵17. Our federal transfers mechanism excludes programs, such as SNAP and SSI, where the federal and state governments share discretion over their financing, administration, or rulemaking.
↵18. Dramatic variation in average levels of support from Veterans’ Benefits is the result of outlier values.
↵19. As of 2024, thirty-two states provide a state-level EITC, and the credit is refundable in all states except Missouri, Ohio, Oklahoma, South Carolina, Utah, and Virginia. Moreover, the state-level EITCs enacted in Missouri, Montana, Utah, and Washington went into effect after the final year included in our analyses, while North Carolina repealed its state-level EITC in 2014. For more information, see Tax Policy Center (2024).
- © 2026 Russell Sage Foundation. van der Naald, Joseph, Sarah K. Bruch, and Janet C. Gornick. 2026. “Poverty Disparities and the Reconfiguration of Social Provisioning: White, Black, and Latino Single-Mother Families, 1996–2018.” RSF: The Russell Sage Foundation Journal of the Social Sciences 12(1): 67–95. https://doi.org/10.7758/RSF.2026.12.1.03. We thank Zachary Parolin and Christopher Wimer for their assistance with the historical Supplemental Poverty Measure and the participants at the RSF “Three Decades Since Making Ends Meet” conference, as well as three anonymous reviewers, for their feedback on an earlier version of this manuscript. This research benefited from funds awarded to the CUNY Graduate Center’s Stone Center on Socio-Economic Inequality by the James M. and Cathleen D. Stone Foundation. Direct correspondence to: Joseph van der Naald, at jvandernaald{at}gradcenter.cuny.edu, Program in Sociology, CUNY Graduate Center, 365 Fifth Avenue, New York, NY 10016, United States.
Open Access Policy: RSF: The Russell Sage Foundation Journal of the Social Sciences is an open access journal. This article is published under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
REFERENCES
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.