Abstract
In this field note, we examine the changing distribution of TANF spending patterns across the states. Given the important shift from the more restrictive categorical grant of AFDC to the more flexible block grant of TANF, examining patterns of spending across program categories is an important facet for understanding states’ welfare efforts. Overall, there has been a general decrease in the percentage of state TANF spending on cash assistance. Using a new measure of party dominance, we find that Democratic states initially spent a higher share of their TANF funds on cash assistance than Republican states. However, by the 2010s this partisan difference basically disappeared with Democratic states spending patterns matching those of Republican states, focusing less on cash assistance and giving greater priority to areas less reflective of TANF’s principal objectives.
Created by the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) in 1996, the Temporary Assistance for Needy Families (TANF) program is one of the country’s safety net programs for providing assistance to needy families and children. Replacing Aid to Families with Dependent Children (AFDC) in 1996, TANF is funded by both federal and state dollars, with the states receiving federal funds through a block grant: the State Family Assistance Grant (SFAG) (Congressional Research Service 2024). 1
In keeping with the block grant funding approach, states are given wide latitude in how they provide benefits and services for families with children, so long as they comply with four statutory goals (Azevedo-McCaffrey and Safawi 2022):
Provide assistance so children can be cared for in their homes (cash assistance)
Reducing parental dependency by promoting work, job preparation, and marriage (facilitating employment)
Reducing the incidence of out-of-wedlock pregnancies
Fostering the two-parent family structure.
Given the decentralized approach of the block grant, an examination of state spending patterns can be instructive in understanding state government approaches regarding needy families and children. In this field note, we examine longitudinal changes in state TANF allocations, with a particular eye toward understanding how state political parties have evolved regarding their programmatic spending priorities. A substantial portion of the state politics literature has been devoted to the questions of whether and how politics matters for public assistance spending in the states. 2 Our approach is to focus on the area of TANF most directly comparable to AFDC (cash assistance) and look for changes in party-related patterns to spending allocations. Drawing on work by Parolin (2021) that classifies state TANF spending into four programmatic categories, we look for patterns within these spending shifts over a multi-year period (2000–2018). Our specific focus is on changes in approaches to cash assistance and the role of party dominance in state government, to see if Democrats and Republicans enacted welfare spending differently during the first two decades of TANF. While we acknowledge differences in spending between the parties, our primary focus is on within-party changes in spending priorities over time. We find that there are differences in how the parties approached TANF, but that changes related to cash assistance are largely asymmetric. In the early years of TANF, Democratic-dominant states were significantly more likely to spend a greater share and amount of their TANF money on cash assistance compared to Republican- dominant states. We note an important shift, however, and find that within the last decade these partisan differences have become negligible, as distribution patterns in Democratic states now more closely resemble those in Republican states.
Broad Patterns of State TANF Spending
A fundamental component of TANF was the shift from the more restrictive categorical grant of AFDC to a more flexible block grant, thereby allowing states significant discretion in how they craft programs to reflect state needs and priorities. The basic federal block grant that funds TANF was originally set at $16.5 billion, and this amount has not changed since the program’s origination (Congressional Research Service 2024). 3 Yet, a recent analysis by the Center of Budget and Policy Priorities indicates that while overall funding levels from the federal government have not changed, how states allocate these funds has shifted considerably. While states initially focused most prominently on the area most directly related to AFDC (cash assistance), the spending data reveal a steady decline, with states spending only $7.1 billion in 2020 (22% of all TANF funds) compared to $14 billion (71%) in 1997. Moreover, there is significant variation across states in terms of cash assistance spending, ranging from a low of 4.4% (Illinois) to a high of 68.3% (Kentucky) in 2020 (Azevedo-McCaffrey and Safawi 2022; U.S. Department of Health & Human Services and Office of Family Assistance 2021).
The flexibility built into the structure of TANF means that examining patterns of spending across program categories is an important consideration for understanding state welfare efforts, as TANF eligibility requirements and spending priorities are, to a large degree, a function of state of residence. Given the significant segment of the state politics literature that addresses the influence of party on political outcomes, our focus is primarily on party-related differences in state approaches to TANF spending. In addition, we examine intra-party variation, detailing shifts in both Republican and Democratic spending priorities over time.
Examining TANF Spending
Initial research indicated that early TANF policies reflected longstanding state approaches related to AFDC (Rodgers and Tedin 2006) with eligibility strictness and lower benefits being related to more conservative state governments and a higher prevalence of Black citizens in state populations (or TANF caseloads) during at least some phases of the TANF time frame (Bentele and Nicoli 2012; Fellowes and Rowe 2004; Gais and Weaver 2002; Rodgers and Tedin 2006; Soss et al. 2001). More recent work examines how the broad discretion given states in determining how their allocations are spent reflects where TANF spending is directed and how effectively outcomes related to TANF’s stated goals are achieved. Fusaro (2021a, 2021b) finds further evidence of racial influences, as states with larger minority populations direct fewer resources to cash assistance. In addition, Fusaro shows that racial attitudes also matter for TANF allocations, with cash assistance receiving lower priority in states where White populations tend to hold negative stereotypes of Black citizens. Parolin (2021) finds that reducing racial differences in allocations to cash assistance (as opposed to prioritizing programs directed toward reducing unwed motherhood and single-parent families) could reduce the black-white child poverty gap. Yoonsook et al. (2022), on the other hand, find that state spending on family and marriage promotion and work activities is related to lower levels of child neglect, as having more income related to a two-parent family may reduce the impacts of poverty-related aspects of neglect. 4
Finally, an additional thread of research examines how changes in TANF spending reflect state-level responses to the broader economic environment. Ewalt and Jennings (2014), for example, note the impact of the 2007–2009 recession on several welfare programs, and find that while spending is generally higher during better economic conditions, the recession triggered a contraction of spending on TANF as states looked for ways to bolster state budgets damaged by the economic downturn. Haskins, Albert, and Howard (2014) place the TANF performance during the Great Recession within the context of previous recessions occurring between 1980 and 2007. They find wide state-level variation in TANF caseload responsiveness even though unemployment rates rose substantially across all states. 5 Overall, TANF was the least responsive of the safety net programs examined (TANF, Medicaid, Unemployment Insurance, SNAP). Work by Anderson, Butcher, and Schanzenback (2015) also shows a lack of TANF responsiveness in terms of program participation, particularly when compared to other programs. Subsequent research by Albert (2016) links this to the more stringent eligibility and time limit restrictions of TANF as well as benefit cuts during the recession.
It is against this backdrop that we add to this discussion, examining changes in party approaches to TANF from 2000 to 2018. Our specific focus is on how states are utilizing the wide discretion they have in terms of program activities, and how this varies depending on party strength in a state. We also look for (and find) important intra-party changes in approaches to state TANF priorities, particularly with regard to the utilization of cash assistance.
Data and Methods
For TANF spending, we use data from The Center on Budget and Policy Priorities (CBPP) and supplement the data with TANF spending reports from the Department of Health and Human Services. The dataset includes yearly spending by state from 1997 to 2018 for all federal dollars sent to the states, state spending or maintenance of effort spending (MOE), and transfers to other federal block grants (Social Services Block Grant and Child Care and Development Block Grant). Following Parolin’s approach, we break the data down into four conceptual spending categories “based on (a) the particular social outcome that the spending aims to influence; (b) the programs or policy tools implemented to achieve the desired outcome; and (c) the implications for the financial security of low-income families” (Parolin 2021, 1008). These categories are cash assistance, work related activities to encourage employment, programs that encourage two-parent families, and an “other” category. The first three categories reflect those areas “identified and aligned with the core purposes explicitly listed in the TANF legislation,” while the other category is spending on services only “tangentially related to the program’s core purposes” (Parolin 2021, 1008).
For this analysis, we focus on the percentage of all spending, federal and state, in the cash assistance category as well as total per capita spending on cash assistance. 6 This allows us to track changes in spending patterns over time to determine if there are any partisan trends to cash assistance expenditures under TANF. We focus on cash assistance for a couple of reasons. Theoretically, this is the category most closely linked to traditional redistributive politics, such as AFDC, as it is a direct cash payment to families and is the one with the clearest partisan expectations for spending. Methodologically, the Department of Health and Human services amended the state reporting form beginning in fiscal year 2015 which makes comparing spending in the other categories prior to 2014 to 2015 and beyond problematic. The cash assistance category remained the same allowing us to make comparisons across the entire time frame. We track spending from 2000 to 2018. While the data go back prior to 2000, we start the analysis in 2000 as the CBPP advises the data is not reliable in the initial years of the program. 7
While there are many options for studying party control, from simply accounting for which party controls government to measures that tap into electoral competition (e.g., Ranney index), here we employ a new measure of party dominance that accounts for party control of the branches of government and levels of partisan competition as well as public opinion, in the form of two-party presidential vote (Parry et al. 2022). This measure provides a more aggregate view of party influence as it combines several components, often examined individually, to capture the broader context of party control and competition in a state. Here we use the four-year index of party dominance that combines the previous four years of the measure as a moving average. The measure can range from 0 (complete Republican dominance) to 100 (complete Democratic dominance) with a 50 indicating partisan balance. 8
Results
Before getting to the impact of party dominance, we first examine overall trends in spending on cash assistance. Figure 1 shows spending as the mean percentage of cash assistance of all TANF spending by year across all states as well as the mean per capita spending in 2018 constant dollars by year.
Percent of total TANF spending.
In 2000, states spent 37%, on average, of TANF money on cash assistance and $48 per capita. Both have steadily declined in the years since with only a few minor increases throughout the years. By 2010, the average percentage had dropped to 26% and by 2018 this had declined to just under 18%. For per capita spending, by 2010, this had fallen to an average of $27 and was down to an average of $15 by 2018.
Moreover, the changes in spending patterns have not been uniform across the states. Figure 2 shows the percent each state spent on cash assistance from 2000 to 2018 with the national average included for comparison purposes. Most states had a decline in cash assistance during this time similar to the national average trend. Several states saw drastic decreases in cash assistance while a few actually had increases throughout the past 20 years. Forty-three states decreased the percentage of spending on cash assistance from 2000 to 2018 with an average decline of 24%. Six states, Hawaii, Maine, New Jersey, New Mexico, Rhode Island, and Vermont, had over a 35% drop in spending on cash assistance. While only seven states, Idaho, Kentucky, Montana, Oregon, South Dakota, Wisconsin, and Wyoming, increased the percent spent on cash assistance by an average of only 8.9%. The maximum increase was Wisconsin’s 12.7% increase.
Percent spending on cash assistance, 2000–2018, by state.
Additionally, Figure 3 shows there is not much correlation (r = 0.24) between percentage spending on cash assistance at the beginning of TANF and in 2018, indicating states have not stayed on the same spending path as when the TANF program started.
Comparison of percent spending on cash assistance in 2000 and 2018 by state.
As shown in Figures 4 and 5 the drops were even more dramatic for per capita spending on cash assistance. Almost every state decreased per capita spending on cash assistance from 2000 to 2018 with only three (Idaho, Oregon, and Wisconsin) having per capita dollar increases. The average decrease was $35 per capita with five states (Alaska, California, Hawaii, Rhode Island, and Vermont) having a decrease of $70 per capita or less. Overall, Figure 5 shows greater consistency (r = 0.72) in state per capita spending with those states spending more in 2000 being more likely to still be spending more in 2018; albeit at lesser amounts than in 2000. Moreover, while research has indicated differences in state responses to the Great Recession (Ewalt and Jennings 2014; Haskins, Albert, and Howard 2014), it is important to note that the decline in cash assistance spending, while exacerbated by the economic downturn, began before the recession and continued thereafter.
Per capita total spending on cash assistance, 2000–2018, by state.
Comparison of per capita spending on cash assistance in 2000 and 2018 by state.
To examine the effect of party dominance on cash assistance spending, we start by showing the mean spent by party over time. For illustrative purposes, we categorize party dominance, similarly to Parry et al. (2022), as states with a score less than 40 to have Republican dominance, and states with a score greater than 60 to have Democratic dominance. We categorize states in between a score of 40 and 60 as competitive states.
Figures 6 and 7 show the patterns of spending by party dominance for percent and per capita cash assistance spending. In 2000, Democratic states spent, on average, just under 50% of their total TANF dollars on cash assistance compared to 36% by competitive states, and 33% by Republican states. By 2006, Democratic states were spending 17% less on cash assistance and looked similar to Republican states. The initial partisan divide in spending on cash assistance was erased completely by 2008 with states, regardless of partisan strength, continuing a slow trend of shrinking percentages over the next decade.
Percent cash assistance by party dominance.
Per capita cash assistance by party dominance.
Figure 7 shows a similar pattern for per capita spending. In 2000, Democratic states average $84 per capita compared to $43 for competitive states and $42 for Republican states. By 2006, Democratic states had dropped by nearly half to $46 per capita while competitive states declined to $31 and Republican states to $28. From 2008 to 2014, Democratic states per capita spending held fairly steady with another decline starting in 2015. By 2018, Democratic states were spending, on average, $30 per capita. Republican states had a sharp decline after 2010 and have remained, on average, under $20 per capita throughout most of the 2010s. Overall, competitive states have followed a similar trajectory as Republican states throughout this period. Regardless of whether we examine percent of funds directed to cash assistance or per capita expenditures, the overall story is one of a significant separation between the parties in the early years of TANF. In subsequent years Democratic states shifted to resemble the approach of Republican states.
Next, we model these outcomes to see if these partisan patterns hold up once we control for important demographic, economic, and TANF related factors. Of note, here we use the full range of the party dominance variable rather than the categories presented in the figures above. To test for any significant partisan patterns over time, we interact party dominance and year in the models and focus on the results of the interaction term. For demographic controls, we include the percentage of the state’s population that is African American as well as the percentage of unwed mothers in the state. We also account for the state’s poverty rate and three important TANF related factors: TANF to poverty ratio, TANF caseload, and the workforce participation rate. We lag all the independent variables, except for year, since spending in the current year reflects budgetary decisions made in the previous year.
Table 1 shows the model results. Both models are OLS regressions with state and year fixed effects with Driscoll-Kraay standard errors (Clark and Linzer 2015; Hoechle 2007). 9 Appendix 1 shows the descriptive statistics and details on the variables included in the models. Figures 8 and 9 show the relationship of interest: the marginal effects in terms of predicted probabilities of partisan dominance over time as this provides a straightforward way to interpret interaction effects (Brambor, Clark, and Golder 2006). Given the model includes the full range of the party dominance measure, Figure 8 (percent cash assistance) and Figure 9 (per capita dollar cash assistance) show the predicted probabilities across years for a competitive party state (mean party dominance), a strong Republican state (−1 standard deviation), and a strong Democratic state (+1 standard deviation) with all other variables held at their appropriate means or modal category. 10
Cash Assistance Model Results, 2000–2018.
Note: OLS regression with fixed effects by state and year.
Driscoll and Kraay standard errors in brackets.
p < .05. **p < .01. ***p < .001.
Predicted percent spending: Cash assistance.
Predicted per capita spending: Cash assistance.
As Figure 8 shows, once we add in controls for demographics, economics, and TANF related factors along with state and year fixed effects and account for robust standard errors, we find a pattern like that in Figure 6. For percentage spending (Figure 8), Democratic states were predicted to spend a significantly greater share on cash assistance than Republican states in the early years of the program, but any statistically significant differences were gone by 2008. From 2012 to 2018, the model predicts the Democratic and Republican states spending to be largely equivalent. The pattern looks very similar for per capita spending (Figure 9) with Democratic states spending like Republican states after 2004. However, post-2012, the Democratic state is actually predicted to spend a significantly lower amount of per capita dollars on cash assistance than the Republican state. Overall, the marginals reflect the descriptive patterns documented earlier. In the initial years of TANF Democratic and Republican spending were significantly different. Over time, Democratic states have come to resemble Republican states regarding their utilization of cash assis-tance.
Since there exists clear declines in spending on cash assistance, we examine the other categories of spending for any partisan patterns that may reflect the movement away from cash assistance. Figures 10 and 11, similar to Figures 4 and 5, show the percentage and per capita spending by the Work, Family, and Other categories from 2000 to 2014 by partisan dominance. We do not extend the analysis beyond 2014 due to the changes in categories from HHS starting in 2015. In terms of percent spending, there is volatility among Democratic states in the work and family categories compared to Republican states, without any clear patterns evident. The clearest patterns emerge in the Other category with Republican states outpacing other states throughout the years and with a sharp increase post-2010. Democratic states have steadily increased spending in these tangentially related areas with a sharp increase post-2008. For per capita spending, Democratic states have always outpaced Republican states in the work category with some divergence post-2010 with Democratic states increasing dollars spent on work-related spending and Republican states spending less. Democratic states had a sharp increase in per capita spending after 2008 while spending in Republican states has remained fairly steady.
Percent spending across the work, family, and other categories.
Per capita spending the work, family, and other categories.
Given the pattern evident in the Other category, we replicate the multivariate analysis (Table 2) with percent Other and per capita dollars spent in the Other category as the dependent variable. Figures 12 and 13 show the predicted spending outcomes. For percent spending (Figure 12) in the other category, the model predicts the Republican state will outpace the Democratic state from 2000-2014; however, the Democratic state closed the gap with steady increases in predicted spending in the other category over the years. Figure 13 shows for per capita spending in the other category, the Republican state spent a significantly greater amount from 2000 to 2008. However, during this time, the Republican state was spending less each year while the Democratic state was spending more. By 2010, the states were spending similar amount in the other category regardless of partisan strength. Taken together, the patterns shown in Figures 8–13 provide broad evidence that Democratic and Republican states are becoming increasingly alike when it comes to TANF spending priorities. Importantly, the change is largely asymmetric, with Democratic states generally converging to the same patterns as Republican states.
Other Spending Model Results, 2000–2014.
Note: OLS regression with fixed effects by state and year.
Driscoll and Kraay standard errors in brackets.
p < .05. **p < .01. ***p < .001.
Predicted percent spending: Other.
Predicted per capita spending: Other.
Conclusion
The flexibility in state spending that is afforded by the TANF block grant underlies the complexity of both how those funds are categorically distributed and the justification for doing so. Although the formula for determining the amount that a state receives is standard, the actual amount of the grant per state varies, and the distribution of a high percentage of those funds is subject to the changing political and economic conditions of individual states. Broad federal guidelines mean a narrow understanding of why shifts in spending occur. Once the minimum federal requirements are met, the lingering question, which this paper has attempted to answer, is how states decide to distribute the remaining funds among designated TANF categories.
Our findings show that there has been a general decrease in cash assistance spending across states in recent years, both in share of TANF funds spent and in per capita dollar amounts. Democratic states initially spent a much greater share and per capita amounts on cash assistance. However, over time, the spending patterns in Democratic states came to mirror spending more closely in Republican states. While Republican states initially spent more money, both as a share and in per capita dollars, in the “Other” category, Democratic states started to follow suit as they shifted money away from cash assistance. The initial partisan differences have decreased considerably and the partisan explanation for this convergence appears to be asymmetric with Democratic states shifting spending away from traditional Democratic spending priorities. One potential explanation may be that stronger Democratic states responded differently to the 2007 recession, utilizing the flexibility of the TANF block grant to shift funds away from areas that had been a higher priority to help manage budget difficulties and programmatic needs in other areas. However, the shift away from cash assistance began before the recession, suggesting the need for greater flexibility to meet changing programmatic priorities was not a direct result of the Great Recession.
Given the findings of this study, future research should continue to track spending patterns over time, particularly in Democratic states, to provide an explanation for why these partisan patterns have changed over time. One approach might be to explore more direct public opinion data on the statutory goals of TANF (i.e., cash assistance, job preparation, reduction of out-of-wedlock pregnancies, supporting two-parent families) to see how differences across states align with priorities, approaches, and goals regarding public assistance.
Supplemental Material
sj-pdf-1-slg-10.1177_0160323X241262047 – Supplemental material for State Partisan Dominance and the Distribution of TANF Funds, 2000–2018
Supplemental material, sj-pdf-1-slg-10.1177_0160323X241262047 for State Partisan Dominance and the Distribution of TANF Funds, 2000–2018 by Jonathan Winburn, Robert Brown and Nichole Gligor in State and Local Government Review
Footnotes
Appendix
Variable Descriptive Statistics.
| Variable | Mean | Standard deviation | Data source |
|---|---|---|---|
| Dependent variables | |||
| % Cash assistance | 27.52 | 13.62 | Center on Budget and Policy Priorities and Department of Health and Human Services |
| $ Cash assistance | 29.08 | 25.94 | |
| % Other | 21.78 | 16.43 | |
| $ Other | 20.24 | 17.69 | |
| Independent variables | |||
| Party dominance (lag) | 48.31 | 13.17 | Parry et al. (2022) |
| % African American population (lag) | 11.07 | 9.59 | Correlates of State Policy (Grossman, Jordan, and McCrain 2021) |
| State poverty rate (lag) | 12.52 | 3.33 | Correlates of State Policy |
| TANF to poverty ratio (lag) | 30.89 | 19.39 | Center on Budget and Policy Priorities |
| TANF caseload (lag) | 38,569.36 | 77,914.84 | Correlates of State Policy |
| Workforce participation rate (lag) | 39.31 | 16.39 | Department of Health and Human Services |
| % Unwed mothers (lag) | 37.61 | 6.92 | Centers for Disease Control and Prevention |
Note: % Cash Assistance and $ Cash Assistance: 2000–2018.
% Other and $ Other: 2000–2014.
$ Cash Assistance and $ Other: Reported in constant 2018 dollars and are per capita for the state’s population.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplemental material for this article is available online.
Notes
Author Biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
