TANF and its anti-poverty effect

The current transformation of TANF, which mandates recipients to work in exchange for assistance, aligns with the welfare activation trends observed in various Organisation for Economic Co-operation and Development (OECD) countries during the late 1990s. Welfare activation emphasises reducing entitlements and requiring unemployed recipients to participate in work activities (Eriksen & Molander, Reference Eriksen and Molander2019; Wright, Reference Wright2016). Similar to many other countries, such as the UK and Norway, the United States implemented welfare reform in 1996. Since then, the literature on TANF has predominantly revolved around the repercussions of the welfare reform, generating a substantial body of research investigating how the economic wellbeing of recipients has changed after the welfare reform. This extensive body of work has primarily examined the impact of welfare reform and/or TANF policies – such as work requirements, sanctions, and time limits – on two main aspects: (1) TANF receipt and (2) the economic wellbeing of recipients. However, research findings regarding the effects of welfare reform have been inconsistent.

Some studies have pointed to positive consequences of welfare reform, indicating that welfare reform increased earnings/incomes of families (Schoeni & Blank, Reference Schoeni and Blank2000) or those of female-headed families (Grogger, Reference Grogger2003; Meyer & Sullivan, Reference Meyer and Sullivan2004)Footnote ⁵ while reducing welfare use/caseloads (Blank, Reference Blank2001; Moffitt, Reference Moffitt and Danziger1999). On the contrary, another line of research has argued that the impact of TANF on poverty is relatively modest. Hoynes, Page and Stevens (Reference Hoynes, Page and Stevens2006) suggest that TANF effects on the poverty are not substantial. While they find evidence that TANF program contributes to increasing incomes of low-income families, it is important to note that TANF transfers phase out at income levels below the poverty threshold. In addition, others have demonstrated that a substantial number of TANF recipients remained in poverty, even after the adoption of welfare reform (Brauner & Loprest, Reference Brauner and Loprest1999; Slack et al., Reference Slack, Magnuson, Berger, Yoo, Coley, Dunifon and Osborne2007).

Despite calls for more research on the long-term effect of welfare reform on the economic wellbeing or self-sufficiency of welfare leavers (Blank, Reference Blank2002), less is known about the duration of employment maintenance after leaving TANF and whether their working conditions (e.g. earnings, disposable incomes, and job status) improves as work hours increase. A few studies have indicated that former TANF recipients who left TANF due to employment and earnings were not significantly better off in terms of employment, incomes and earnings (Acs & Loprest, Reference Acs and Loprest2007). Specific research focussing on female householders with children has yielded similar findings, revealing declining employment rates over time and economic hardships following exit from TANF (Cancian et al., Reference Cancian, Haveman, Meyer and Wolfe2002; Cancian & Meyer, Reference Cancian and Meyer2000). It is also reported that former welfare recipients show a high likelihood of returning to the welfare system due to low earnings and unstable job status (Anderson, Halter, & Gryzlak, Reference Anderson, Halter and Gryzlak2004).

In summary, studies on the immediate consequences of TANF (e.g. changes in TANF caseloads or short-term changes in employment and earnings) have yielded inconsistent findings. Furthermore, the long-term effects of TANF on recipients who voluntarily or involuntarily left the program have not been adequately examined. Given this landscape, it remains challenging to definitely conclude that TANF, as a safety net program, has significantly contributed to poverty reduction amongst low-income families.

Earned Income Tax Credits (EITC) as an effective tool to reduce poverty

Earned Income Tax Credits (EITC) is a federal cash transfer program in the United States. Introduced in 1975, the EITC aimed to alleviate Social Security tax burdens of low-income working families. Its primary objective is to reduce the reliance on welfare programs, increase self-sufficiency and promote employment amongst low-income working families (Mendenhall MSW, Reference Mendenhall2006). In discussion about the working poor, research has highlighted that this group often does not receive benefits they qualify for in comparison with other groups, such as the non-working poor (Bendick, Reference Bendick1980; Kim & Mergoupis, Reference Kim and Mergoupis1997; Meyers & Lee, Reference Meyers and Lee2003). This argument is supported by analysing participation rates of eligible working poor individuals compared with non-working poor individuals for public assistance programs such as Food Stamps, TANF and Medicaid. Despite significant expansions of welfare program (e.g. EITC, Medicaid and Child Health Insurance Program) in the 1990s, working poor families still faced challenges accessing certain forms of assistance, such as childcare services and health insurance (Meyers & Lee, Reference Meyers and Lee2003; Seccombe & Amey, Reference Seccombe and Amey1995). In these situations where fewer benefits are offered to the working poor, the EITC has emerged as a successful program that both incentivises and rewards the work of low-income working families, becoming the largest cash transfer program in the U.S.

The effectiveness of the federal EITC has been assessed as it gradually expanded. Research findings consistently indicate that the EITC effectively reduces poverty (Haskins, Reference Haskins, Blank and Haskins2001; Gundersen & Ziliak, Reference Gundersen and Ziliak2004).Footnote ⁶ Its anti-poverty effects are particularly pronounced amongst low-income families with children (Kim, Reference Kim2001; Ozawa & Hong, Reference Ozawa and Hong2003), although there are various aspectsFootnote ⁷ that need further modification to alleviate poverty (Eamon, Wu, & Zhang, Reference Eamon, Wu and Zhang2009). Studies analysing the impact of EITC have utilised income/earnings, employment or poverty levels of low-income families to measure economic success or wellbeing. The labour supply incentives of the EITC vary on the basis of individual characteristics (e.g. marital status, working status, gender and education attainment). However, previous studies have offered consistent findings that the EITC encourages labour force participation amongst single parents with children (Eissa &Liebman, Reference Eissa and Liebman1996; Grogger, Reference Grogger2003; Meyer & Rosenbaum, Reference Meyer and Rosenbaum2001) and increases the incomes of low-income families (Hoynes & Patel, Reference Hoynes and Patel2018).

Eissa and Liebman (Reference Eissa and Liebman1996) provide evidence that the EITC encourages labour force participation amongst single women with children, and these effects are greater amongst those with less than a high school education. Unlike most studies focussing solely on the impact of federal EITC, Neumark and Wascher, Reference Neumark and Wascher2000) estimate the impact of both federal and state EITC on transitioning out of poverty by comparing earnings and employmentFootnote ⁸ of families with earnings below and above the poverty level. Their findings confirm that both federal and state EITCs have a positive impact on employment, especially for single mothers who initially were not in the workforce.

The effects of EITC have been mainly discussed in terms of whether it induces employment and whether it increases or decreases work hours. While labour supply disincentives do exist for married women eligible for the EITC (Eissa & Hoynes, Reference Eissa and Hoynes1998, Reference Eissa and Hoynes2004; Hoynes & Patel, Reference Hoynes and Patel2018), the EITC is known to encourage employment amongst single women with children, as mentioned earlier. It also offsets labour supply distortions that might arise amongst female welfare recipients (Eissa & Hoynes, Reference Eissa and Hoynes1998). However, evidence regarding the impact of the EITC on work hours of single mothers has been inconsistent. For instance, Eissa and Liebman (Reference Eissa and Liebman1996) find no evidence that the EITC leads to a decrease in work hours amongst single mothers already in the workforce. However, other studies support the positive effects of the EITC on work hours of single mothers (see Meyer and Rosenbaum, Reference Meyer and Rosenbaum2001; Grogger, Reference Grogger2003). This mixed evidence may result from the income ranges studied. For instance, Eissa and Liebman (Reference Eissa and Liebman1996) specifically concentrate on the phase-out range, wherein EITC benefits decrease as income exceeds a certain threshold. In contrast, other studies seem to consider the entire income spectrum of single mothers.

TANF fund diversion to state EITCs: its impact on TANF recipients

TANF programs are often referred to as a failure, as evidenced by the decline in TANF receipt and low benefit levels. In addition to these issues, the allowability of TANF expenditures – though legally permissible – has recently been criticised for straying from the program’s intended goals. According to a report by CBPP (2023), states have significantly reduced the portion of their federal and state TANF funds spent on core areas, diverting TANF funds to other programs such as childcare, refundable tax credits or pre-K/Head Start. On average, in FY 2021, states spent only 22.6 percent of their combined (federal and state) TANF funds on basic assistance, with state spending ranging from 4 percent to 75 percent (Shrivastava, Reference Shrivastava2023). These statistics suggest that the diversion of TANF funds may leave out those most in need, instead supporting families with incomes well above the poverty line.

The same report also points out states’ current practices of carrying over TANF fundsFootnote ⁹ and urges policy changes to ensure that a majority of TANF funds are spent on basic assistance, with a recalibrated focus on people with incomes below the federal poverty level. In 2021, states’ unspent TANF funds totalled $8 billion, nearly half of the total annual TANF block grants (Shrivastava, Reference Shrivastava2023). Additionally, sixteen states had unspent TANF funds equal to or exceeding their annual block grants (Shrivastava, Reference Shrivastava2023). States should not be criticised merely for carrying over TANF funds if they have adequately spent their allocations to assist those with incomes below the federal poverty level. In such cases, an increase in carried-over funds could reflect a decline in caseloads. However, given that the average proportion of eligible individuals receiving cash assistance between 2016 and 2018 was estimated at less than 30 percent (Falk, Reference Falk2023), it can be argued that states need to make greater efforts to include more eligible individuals and provide additional cash assistance. The current level of TANF spending may not be sufficient to address the needs of those who qualify assistance.

Given the uncertainty surrounding the effects of TANF on the economic wellbeing of recipients, states may find it more effective to utilise a portion of TANF funds to support the state EITC to reduce poverty for several reasons. First, the state EITC does not discourage work amongst single parents, alleviating concerns about potential welfare dependency (Beverly, Reference Beverly2002); TANF has been criticised for its incentive structure that distorts labour supply (Gruber, Reference Gruber2000). Second, the tax credits offered by the EITC provide beneficiaries with a sense of social inclusion (Sykes et al., Reference Sykes, Križ, Edin and Halpern-Meekin2015), resulting in less stigma compared with other means-tested programs (Martin & Prasad, Reference Martin and Prasad2014).

Unlike the extensive studies that have examined the impact of states’ discretion in designing the TANF program – focussing on eligibility rules, work requirements and sanctions – prior research has paid less attention to how states spend federal and state TANF funds at their discretion or empirically explored the impact of TANF fund diversion on recipients. Although there is no empirical evidence to support that the diversion of TANF funds negatively affects current recipients, it is reasonable to expect harm, given that the majority of adult TANF recipients are unemployed with no earned income, making them particularly vulnerable to reductions in basic assistance. While some TANF recipients, especially those who are working, might benefit from other programs, such as state EITC programs supported by the diverted funds, the overall impact of TANF fund diversion is likely detrimental to current recipients. Conversely, it may benefit some former recipients, particularly those who exited TANF due to employment. This underscores the need for empirical studies to explore the true impact of TANF fund diversion on recipients.

Drawing on prior literature’s concerns about the diversion of TANF funds and the effectiveness of EITC for the working poor, this study hypothesises that state governments’ decision to assist the working poor by utilising TANF funds for state EITC programs could negatively impact current TANF recipients, leading to an increase in changes in the poverty gap, as the majority of these recipients are unemployed. However, the impact on former TANF recipients is less clear, given the diverse reasons for their exit. While some former recipients who exited due to employment may benefit from state EITC, others are less likely to benefit if their exit was due to sanctions or non-compliance with work requirements. As a measure of poverty, this study employs 50 percent of the state median income of households with children as a threshold, as commonly practised by many researchers (Brady, Finnigan, & Hübgen, Reference Brady, Finnigan and Hübgen2017; Gornick & Jäntti, Reference Gornick and Jäntti2012; Parolin, Reference Parolin2019), and assesses the poverty gap by calculating the difference between this threshold and the average income of recipients. To analyse changes in the poverty gap, the disparity between the poverty gap at time points t and t−1 is calculated and used in the analysis.

Data and method

Data

To examine the impact of states’ TANF diversion on changes in the poverty gap for both current and former TANF recipients, this study uses multiple data sources, including the U.S. Census and various TANF reports provided by the Office of Family Assistance (OFA). The dependent variable is the change in the poverty gap, where the poverty gap is defined as the difference between 50 percent of the state median income for households with children and the average income of TANF recipients. The change in the poverty gap is calculated by subtracting the poverty gap at time point t−1 from the poverty gap at timepoint t.

The average income of TANF recipients includes both TANF incomes (i.e. TANF benefits) and non-TANF incomes, as some recipients are employed. For current TANF recipients with both TANF and non-TANF income, their combined income is used, while for former TANF recipients, who only have non-TANF income, their average non-TANF income is employed in the calculation. The U.S. Census Bureau provides data on the median income of households with children, while information regarding TANF and non-TANF incomes for current and former recipients is derived from the OFA’s annually released Characteristics and Financial Circumstances of TANF Recipients report.

In addition to the primary models using 50 percent of the state median income for households with children, a widely used standard for calculating the poverty gap, the ‘poverty threshold’ issued by the Census Bureau – used for the official poverty measure – is also used as a benchmark. This is because 50 percent of the state median income for households with children may be too high a threshold for TANF recipients.Footnote ¹⁰ The main models using 50 percent of the state median income for households with children are reported in models 1 and 2, while the additional analyses are reported in models 3 and 4.

Each model incorporates the same independent and control variables. The primary independent variable is a dummy variable indicating whether states allocate TANF funds to state EITC programs. A positive (or negative) coefficient of TANF fund diversion indicates greater instability (or improved stability) and a widening (or narrowing) poverty gap over time. Additionally, the percentages of EITC funded by federal sources (federal TANF funds) and EITC funded by state MOE funds are included as two other independent variables to assess whether the funding source impacts changes in the poverty gap in distinct ways. However, their impacts are not hypothesised, as it is unclear whether states’ use of either federal TANF funds or state MOE funds to expand state EITC programs would lead to differential spending behaviours that could benefit or harm current and former TANF recipients. States utilise either federal TANF funds or state MOE funds, or sometimes a combination of both, to offer state EITC benefits. Only a few states employ both federal and state MOE funds for their EITC program. As presented in Table 1, there is huge variation in states’ diversion of TANF funds to state EITC programs, ranging from 0 percent to 60.37 percent of federal TANF funds and from 0 percent to 69.41 percent of state MOE funds, although the average is less than 10 percent for both funding sources.

Table 1.

Descriptive statistics

To address cases in which some states allocate more state MOE funds than required, a dummy variable is included to identify states that report exceeding the minimum requirements, termed ‘excess MOE’.Footnote ¹¹ Noting the concern that states are reducing their spending on core activities, an annual percentage of TANF expenditures allocated to basic assistanceFootnote ¹² is included in the analysis. The Office of Family Assistance (OFA) provides annual TANF Financial Data containing information on expenditures across various categories. Data on states’ annual TANF expenditures and the number of households receiving TANF benefits are derived from the same TANF financial data provided by OFA.

Furthermore, other TANF-related variables that might influence changes in the poverty gap are taken into account. First, the percentage of TANF-receiving families amongst low-income families is included. Low-income families are defined as those with incomes below 100 percent of the poverty level estimated by the U.S. Census Bureau. There is considerable variation amongst states in the proportion of TANF-receiving families amongst low-income families. For example, in 2013, this proportion ranged from approximately 87.1 percent in Maine to 2.8 percent in Wyoming. Second, the maximum monthly initial earnings a family of three is allowed to have and still be eligible for TANF benefits are included. The amounts of initial earnings that states do not count when determining TANF applicants’ eligibility vary across states, indicating different levels of generosity in the TANF program. Empirical evidence shows mixed effects: stringent TANF policies lower the odds of working poor women transitioning to TANF recipients, but also discourage them from becoming working non-poor (Cheng, Reference Cheng2010). Information on the number of families with incomes below 100 percent of the poverty level is gathered from the U.S. Census Bureau, and the data regarding the maximum initial earnings are obtained from the Urban Institute website.

This study employs a set of state economic and demographic factors as control variables to examine the influence of the internal state economy and demographic composition on changes in the poverty gap. Given that TANF clients are predominantly female householders with children, the models include a percentage of female-headed households with children under 18 years. To capture the economic conditions and the fiscal capacity of states, this study utilises the previous year’s unemployment rate and the total taxable resources per capita, with data obtained from the Department of Labor and the Department of Treasury (DOT), respectively. For demographic composition, variables such as the percentage of non-white individuals and the percentage of individuals aged 25 years or older with a high school degree or less are included. All data were collected from the U.S. Census Bureau. Lastly, the government ideology, as defined by Berry et al. (Reference Berry, Fording, Ringquist, Hanson and Klarner2010), is incorporated into the model, as the ideology of state governments may influence the prioritisation of assistance to the working poor over the non-working poor. Higher values of the government ideology variable indicate a more liberal orientation of the state government. Descriptive statistics of all variables used in this study are presented in Table 1.

Method

The dataset used in this study is a balanced panel of all fifty states over a span of 9 years from 2008 to 2016. The primary models (models 1 and 2) presented in Table 2 are analysed using a fixed-effects estimation method on the basis of the Hausman test. The test results show that all models significantly reject the null hypothesis with p-values less than 0.01, indicating that fixed-effects estimation is preferred over random-effects estimation. The model specification is as follows:

$$\eqalign{ & \Delta {y_{it}} = {\beta _{0{\rm{\;}}}} + {\beta _1}{\rm{\;}}TANF{\rm{\;}}Diversio{n_{it - 1}} + {\beta _2}{\rm{\;}}Proportion{\rm{\;}}of{\rm{\;}}Federally{\rm{\;}}funded{\rm{\;}}EIT{C_{it - 1}} \cr & + {\beta _3}{\rm{\;}}Proportion{\rm{\;}}of{\rm{\;}}State{\rm{\;}}MOE{\rm{\;}}funded{\rm{\;}}EIT{C_{it - 1}} + {\beta _4}{\rm{\;}}{X_{it}} + {\alpha _i} + {\delta _t} + {u_{it}} \cr} $$

where $\Delta {y_{it}}$ denotes changes in the poverty gap within state i at time of t, and X _it represents time-varying covariates related to the TANF program and demographics that could potentially influence changes in the poverty gap. State-fixed effects (α _i ) have been incorporated to account for time-invariant factors unique to each state that might impact changes in the poverty gap. Additionally, the parameter δ _t corresponds to year-fixed effects, accounting for unobserved effects that change over time while remaining constant across states.

Table 2.

The impact of states’ TANF fund diversion on changes in the poverty gap

Note: ***p < 0.01, **p < 0.05, *p < 0.1; standard errors are clustered by states. Models 3 and 4 are analysed using random-effects estimation on the basis of the results of the Hausman test.

^a The poverty gap is measured by the difference between 50 percent of the median income of households with children and the average incomes of TANF recipients. In terms of changes in the poverty gap, the difference between the poverty gap at timepoint t and t−1 is calculated and then subjected to logarithmic transformation for analytical purpose.

^b In Models 3 and 4, the threshold used to calculate the poverty gap is the ‘poverty threshold’ issued by the Census Bureau. For the final models, the poverty threshold for a family size of two with one child is applied.

As presented in Table 1 in the Appendix, the correlations between the variables are all below 0.6, except the correlation between the EITC dummy variable and the proportion of state MOE funded EITC. The correlation between these variables is 0.734, which slightly exceeds the generally accepted threshold of 0.7. As a diagnostic to detect multicollinearity, variance inflation factor (VIF) values are calculated. VIF values for each variable are less than 4, with an average VIF value of 2.43. Considering correlations and VIF values, it can be inferred that the model is not substantially impacted by serious multicollinearity issues. To address concerns related to heteroscedasticity and autocorrelation, standard errors clustered by states are provided.