Race, “Deservingness,” and Social Spending Attitudes: The Role of Policy Delivery Mechanism

Abstract

This paper examines how the means through which social benefits are delivered—either through a direct government program, or through a tax expenditure program—affects how citizens view social welfare programs and their beneficiaries. Attitudes toward social spending in the United States are strongly conditioned by both racial considerations and perceptions of the deservingness of recipients. We argue that the political cues given by spending conducted through the tax code differ from those given by direct spending in a way that both de-racializes spending attitudes and changes the lens through which citizens evaluate the deservingness of beneficiaries. Through a series of survey experiments, we demonstrate that social benefits delivered through the tax code are less likely to activate racialized thinking than similar or identical benefits delivered directly. This is true, at least in part, because recipients of tax expenditures are perceived as more deserving than recipients of otherwise identical direct spending.

Introduction

A large body of research has explored the determinants of public support for social welfare programs, particularly those designed to benefit society’s most vulnerable citizens (e.g., Feldman and Zaller 1992; Goren 2008; Jacoby 1994, 2000; Wlezien 1995).^1,2 This work shows that the public is broadly supportive of an active government role in assisting those in need, but resists some specific forms of social spending—particularly that allocated to “welfare” programs—designed to help the poor. This resistance is, in large part, because of perceptions that such spending is aiding the “undeserving” poor: those who are not working to help themselves (Petersen 2012).

These perceptions of deservingness are often tied up with racial considerations: the work of Gilens (1999), for example, suggests that the unpopularity of certain social spending programs is driven in part by prejudices—particularly racial prejudices—toward the perceived recipients of such spending. Racialized attitudes, in other words, affect support for social spending programs (even ones that are ostensibly “race-neutral’), and do so because racial attitudes affect perceptions of how deserving recipients are of government aid.

This line of research has taught us a good deal about how social spending programs are perceived. But this work has focused almost exclusively on opinions toward public social programs funded directly by governments. The United States, however, has a divided system of social spending, where social benefits are delivered not only through direct budgetary spending, but also through incentives and credits situated within the federal tax code (Hacker 2002; Howard 1997; Mettler 2011). Tax expenditures, which comprise roughly 33% of all federal social spending, provide an opportunity for government to deliver benefits to citizens (or incentivize the private provision of social benefits) in a more indirect way.

The largest and most prominent tax expenditures—the mortgage interest tax deduction, exclusions for purchasing health insurance or saving for retirement through employer-sponsored plans—benefit predominantly well-off citizens (Faricy 2015). But some tax expenditure programs redistribute wealth downward, and policymakers in both parties have offered plans to expand tax credits as a way to reduce income inequality and help lower-income citizens. If public attitudes toward tax credits are affected by the same sorts of concerns about race and the perceived deservingness of recipients as are traditional welfare programs, then efforts to expand such programs are likely doomed to failure. If, however, providing aid to low-income citizens through the tax code does not engender the same types of opposition, then using tax credits to address poverty and inequality becomes a feasible policy alternative.

In this paper, we examine how the means through which social benefits are delivered conditions how citizens think about both the worth of social programs and the deservingness of their beneficiaries. We pay particular attention to the role that the spending mode plays in either activating or deactivating racial prejudice as a predictor of spending attitudes. We argue that the policy mechanism through which social benefits are delivered matters to how citizens think about how such benefits are working and whom they are designed to help. We suggest that programs which deliver benefits through the tax code prime different considerations than programs which deliver benefits directly, and that this leads people to view both the worth of the programs and the deservingness of their beneficiaries in different ways.

Through a series of surveys and survey experiments conducted on representative national samples, we explore these ideas in three ways. First, we examine the predictors of support for two major federal anti-poverty programs that currently exist: the Earned Income Tax Credit (hereafter the EITC) and Temporary Assistance for Needy Families (hereafter TANF). We find that attitudes toward the EITC are less strongly predicted by racial considerations than are attitudes toward TANF. These results support the notion that spending through the tax code is more popular than spending through direct means, though with the important caveat that TANF and the EITC have somewhat different goals and serve somewhat different populations.

In the subsequent two studies, we hone in directly on the role of delivery mechanism itself. Our second study conducts a survey experiment to isolate the impact of spending mode on social policy views. We find that support for a hypothetical downward-distributing social benefit is more strongly predicted by racial attitudes when it is designed as a direct spending program rather than when it is delivered through the tax code. Finally, using a series of experimental vignettes, we examine the effect of spending mode on perceptions of beneficiaries of social welfare programs. We generally find that recipients of social benefits are perceived as less deserving, and perceptions of deservingness more strongly conditioned by racial attitudes, when recipients receive their benefits directly rather than through the tax code.

Together, these findings suggest that attitudes toward social benefits delivered through the tax code are less driven by racial prejudice than attitudes toward social benefits delivered through the more traditional welfare program. This is in part because spending through the tax code as it is currently constituted tends to target the “deserving poor:” the EITC is popular, for example, because it is targeted explicitly to those earning taxable income. But we also find that the mode of spending itself matters: otherwise identical benefits are viewed differently depending on how they are delivered to beneficiaries. These findings broaden our understanding of public support for social spending generally, and for the relationship between public perceptions, policy design, and support for aid to the poor more specifically. They also have implications for evaluating the political feasibility of various inequality-reducing social spending programs.

The United States’ Divided Social Welfare System

The United States has a divided social welfare system that finances social benefits both through direct federal programs and through the tax code (Faricy 2015; Hacker 2002). The public welfare state constitutes federal spending on publicly administered services or benefits such as Social Security, Medicare, and Temporary Assistance for Needy Families. The federal government also uses tax expenditures—defined as targeted tax breaks provided to individuals or businesses to encourage the private provision and administration of social benefits—to subsidize the private side of the social welfare state.³

While social tax expenditures ostensibly work toward the same sorts of goals as public social welfare spending (e.g., provision of health services, retirement security, and poverty relief) there are politically important differences in the design of the two types of policy tools. First, tax expenditures are used to incentivize the actions of private-sector business and citizens, either by subsidizing the purchase and provision of private social benefits and services (e.g., subsidies for individuals who contribute to private retirement accounts and health insurance from an employer-sponsored private health care plan). Next, and contrary to most public programs, the vast majority of social tax expenditure programs distribute wealth upward (Faricy 2015).⁴

However, there are some federal tax expenditure programs that distribute benefits progressively. An important example is the Earned Income Tax Credit (EITC), which is a federal tax credit for low- and moderate-income families. The EITC is designed as a subsidy for low-wage work that offsets a household’s federal taxes. Since it is a refundable tax credit, it can lead citizens to receive more in the way of a tax refund check than they have paid in federal taxes.

Assistance to the Poor: A Comparison of Welfare and the Earned Income Tax Credit

Two of the largest means-tested assistance programs in the United States are the Temporary Assistance for Needy Families and the EITC. In 1996, the federal government replaced the Aid to Families with Dependent Children (AFDC) with TANF in an effort to reform a welfare program that many detractors saw as deemphasizing the importance of work, and encouraging dependency on government aid. Both President Clinton and the Republican Congress promoted TANF as a reform that would provide assistance to the poor in a way that incentivized work and supported two-parent family structures (Grogger and Karoly 2009). The EITC was passed with bipartisan support during the Ford administration as part of the Tax Reduction Act of 1975. The EITC is a refundable tax credit that essentially subsidizes low wage work and provides the largest benefits to the neediest recipients.

These programs differ in several ways. The EITC has an explicit work requirement—i.e., a recipient must be earning a wage to receive benefits—while a significant portion of TANF recipients are not employed (although getting recipients back to work is an explicit goal of the program). The benefits of the EITC are also distributed entirely by the federal government through the federal tax code, while states have some leeway, with federal guidance, to distribute TANF benefits as they see fit.

Still, the Earned Income Tax Credit shares most of the same broad policy goals as TANF. These two programs share the same policy goals of encouraging employment, stabilizing families, and reducing the role of the federal government in providing aid to the poor (Hoffman and Seidman 2003, Holt 2006).

EITC and TANF also serve broadly similar populations. Both programs provide most of their assistance to families rather than single adults and have helped augment the income of single-female households (Moffitt 2003). Both mainly support younger households with lower incomes (Athreya et al. 2010). The racial demographics of program beneficiaries are similar as well: a recent analysis of the demographic composition of EITC recipients shows that roughly half of those who claim the credit are white, and roughly half are members of racial and ethnic minority groups (Murray and Kneebone 2017).⁵

However, while the EITC and TANF share similar goals and serve similar populations, they have developed at much different rates over the last four decades. The EITC has expanded its eligibility requirements to adjust for family size, included single adults without children, and was indexed to inflation as part of the Tax Reform Act of 1986. In addition, the actual value of the benefit was increased in 1990, 1993, 2001, and 2009. Conversely, there have been only incremental changes in TANF funding over this same period (CBO data suggests that, in real dollars, welfare spending has flatlined), and since 2010, Congress has extended TANF with short-term bills in lieu of a full reauthorization of the program. Due to these changes, the EITC overtook TANF/AFDC as the federal government’s primary way of providing cash assistance to low-income families in 1993. The government now spends three times as much money on the EITC as it does on TANF, making EITC the largest downward-distributing social program in dollar terms.⁶

Public Attitudes Toward Social Welfare Spending: The Role of Deservingness and Race

There are a number of factors that influence citizens’ support or opposition to social welfare policies: partisanship, trust in government, ideology, and values, to name just a few (e.g., Feldman and Zaller 1992; Ellis and Stimson 2012; Hetherington 2005; Jacoby 1994; Rudolph and Evans 2005). But perhaps the most prominent factor in the social-psychological literature on this subject is a simple consideration of whether particular beneficiaries of social assistance are deserving of the aid that they receive (e.g., Aarøe and Petersen 2014; Slothuus 2007; van Oorschot 2006). This ‘deservingness heuristic’ undergirds how citizens make sense of social policy: above and beyond cultural or partisan predispositions, citizens wish to bestow government aid on those perceived as unlucky or working to better themselves, and wish to withhold benefits from those perceived as lazy or unsympathetic.⁷ Citizens’ attitudes towards government programs are, in other words, heavily influenced by perceptions of beneficiaries’ deservingness (Gilens 1999; Petersen et al. 2011; Van Oorschot 2000, 2006).

In the U.S., deservingness is also often viewed through the lens of racial prejudices and considerations. Many white citizens believe that African Americans are less committed to the American value of a strong work ethic (Bobo 1998; DeSante 2013). Sniderman and Piazza (1995) argue that whites’ opposition to programs designed to help racial minorities is driven largely by perception of group effort—that black citizens as a group lack a work ethic, and therefore, are undeserving of government aid.⁸ The widespread belief that African Americans lack a commitment to work ethic plays an important part in structuring white attitudes towards welfare spending: broadly speaking, this research shows that anti-poverty programs are less popular than they otherwise would be because they both activate racial prejudices explicitly and because they prime conceptions of deservingness that are also conditioned by racial prejudices.

Program Delivery Mode, Race, and Perceptions of Deservingness

The social construction of a target group influences the feasible range of policy options available to assist such groups. American policymaking occurs in an environment where some groups (e.g., small children, the elderly, veterans) are viewed as sympathetic or deserving, while other groups (e.g., those with criminal records, the unemployed not seeking work, some minority populations) are stigmatized as undeserving (Schneider and Ingram 1993). A group’s social construction is, in turn, strongly connected to the types of options that might be realistically available to assist them: both citizens and policymakers are more willing to invest resources to assist groups viewed positively rather than negatively (Schneider and Ingram 1993; Link and Oldendick 1996; Schroedel and Jordan 1998).

Many if not most of the differences in perceived deservingness of target groups are driven by deep-seated and fairly universal factors—whether a group is perceived as working hard, playing by the rules to get ahead, whether they have the ability to help themselves, and so on. We argue, though, that while perceptions of beneficiaries of social programs can condition the range of policy tools available to assist them, the causal arrow also can go in the other direction: that the means through which policy benefits are delivered can help condition perceptions about beneficiaries themselves.

When citizens are faced with thinking about whether the beneficiaries of a particular social benefit are deserving of aid, a wide variety of different considerations may come into play: personal experiences with the program or its beneficiaries, positive or negative stereotypes about beneficiaries, an evaluation of whether the program is likely to encourage positive (e.g., work) or negative (e.g., laziness) behavior, and views of the social goals promoted by the program itself (Hedegaard 2014). For many people, these considerations may be in tension with one another: someone may abstractly support the goal of reducing poverty in the United States, for example, while at the same time thinking that poor citizens who receive government aid are lazy and unsympathetic (Ellis and Stimson 2012). Given the public’s low levels of political knowledge generally and knowledge of particular policies specifically (Achen and Bartels 2016; Converse 1964, Lupia 2016), classic theories of public opinion formation (e.g., Zaller 1992) tell us that the types of considerations that are most prominent in the minds of citizens when they form opinions on these sorts of matters will matter most to the opinions that they ultimately express on them.

Given this, if the means through which a social benefit is delivered affects the kinds of characteristics that citizens are thinking about when evaluating both programs and beneficiaries, then we should expect that citizens will evaluate programs delivered through different means in different ways—even if the ultimate goals and beneficiaries of the programs are similar. We argue the way in which benefits are delivered, in other words, helps to fill an informational void that influences how citizens form judgments about the deservingness of a program’s recipients. More specifically, we argue that because citizens at least in part think about different considerations when evaluating the beneficiaries of a tax expenditure program than they do when evaluating the beneficiaries of direct spending programs, that programs delivered through the tax code should be more popular, and less strongly associated with racial attitudes, than otherwise similar programs delivered through direct spending.

There are two primary reasons why we hold these expectations. The first reason deals with how a delivery mechanism affects how people view the role of “the government” in providing benefits. Scholars have noted a long-standing ambivalence in Americans’ attitudes toward social welfare spending: Americans, even large numbers of conservatives, are generally supportive of a society that is more equal and does more to assist those in need (Page and Jacobs 2009; McCall 2013; Piston 2018). But at least in the abstract, symbolic level, “government” is viewed as wasteful and inefficient, and citizens generally distrust government’s ability to deliver benefits efficiently or effectively (e.g., Jacoby 2000; Hetherington 2005). And as Gilens (1999) and others have noted, the notion of “receiving checks” from the government can also prime racial attitudes: recipients of government spending are both portrayed in the media and perceived by citizens as disproportionately black and undeserving (Iyengar 1990; Will 1993; Hancock 2004).

These competing considerations lead to a tension in attitudes toward many types of direct spending: people want government to solve important social problems, but also prefer a government that is smaller and less directly intrusive (Stimson 2016). By incentivizing the private provision of social benefits, tax expenditures reduce this tension, aligning with views of government distrust while also working to achieve specific popular social goals (Haselswerdt and Bartels 2015).⁹ By delivering benefits indirectly, tax expenditures should minimize the role that negative attitudes toward “the government” play in evaluating the worth of particular policies, and—by not priming a direct role of government—should serve to weaken some of the connections between racialized attitudes and support for government spending.

The second reason deals with how a program’s delivery mechanism directly affects perceptions of beneficiaries. Past work has shown that programs delivered through the tax code prime the notion of “taxpayer” rather than “government beneficiary” (see Faricy and Ellis 2014). While Americans have complex views on the role of taxes in facilitating social spending, taken in a vacuum the notion of “taxes” are inherently negative, and “taxpayers” are inherently viewed positively.¹⁰ People receiving money through the tax code are seen as those who have contributed to society twice over – once to the economy through their labor and again to the government through paying taxes. Knowing that the beneficiary has paid taxes meets one of van Oorschot’s (2000, p. 356) conditions of “deservingness:” knowing that a beneficiary has paid taxes implies that they can be expected to contribute to society and the economy in the future.

Because of this, there is also reason to think that priming the notion of “taxpayer” also serves to activate counter-stereotypical perceptions of African Americans that weaken the role that negative racial stereotypes might play in evaluating recipients of government aid. Many anti-poverty programs are unpopular in large part because they prime negative racial stereotypes, particularly those that suggest African-Americans are lazy or otherwise not full contributors to society (Gilens 1999; Kellstedt 2003). This is true to some extent regardless of how a particular program is structured: the very idea of distributing benefits toward “the poor” can prime racial considerations (e.g., Clawson and Trice 2000; Hancock 2004).

However, not all anti-poverty programs—even those which are thought by the public to benefit mostly minority groups—are subject to the same types of negative stereotyping of their beneficiaries. As Gilens (1999) writes, “programs that are seen as enhancing the ability of the poor to support themselves, rather than rewarding the lazy with government handouts, do not evoke the same negative imagery” as programs such as welfare. White Americans are not necessarily opposed to social programs that disproportionately help racial minorities, as long as the programs are perceived as benefiting the deserving. Since “taxpayers” as a group are generally considered to be hardworking contributors to society (Williamson 2017), priming the notion of taxpayer may weaken the connections between “beneficiaries of government aid” and “negative racial stereotypes” found to be so common in direct spending programs.

Social benefits delivered through the tax code, then, negate two negative signals that are commonly associated with direct spending designed to assist the poor: the inefficiency and distrust of “the government”, and the negative, often racially-charged, views of “recipients of government aid.” We believe that these differences in program delivery mechanism should affect the lenses through which citizens view both government aid programs and the beneficiaries of such programs.

Our expectations are straightforward. First, we expect that attitudes toward programs that deliver benefits through the tax code will be less likely to be affected by racial prejudices than programs delivered through direct means. Second, we expect that attitudes toward the beneficiaries of such programs will be less racialized if they are receiving benefits through the tax code than if they are receiving benefits through direct means.

In what follows, we test these expectations in three different ways. First, we examine attitudes toward two of the most prominent real-world programs designed to assist the poor— welfare and the EITC. Second, we isolate the impact of program delivery mechanism on attitudes toward social benefits, conducting a survey experiment where we hold the structure of a hypothetical social benefit program constant, but vary the means through which it is delivered. Third, we run a survey experiment examining the impact of delivery mechanism on perceptions of beneficiaries of social spending, testing whether people who receive benefits through the tax code are viewed differently than people who receive otherwise identical benefits through direct means. The first of these studies is meant to test whether the two most prominent “real world” anti-poverty programs are viewed through different lenses. The second of these is designed to isolate the impact of delivery mechanism itself, testing whether otherwise identical programs are viewed differently conditional on how they are delivered. The third moves from understanding perceptions of programs to perceptions of beneficiaries: it is designed to see whether recipients of otherwise identical programs are viewed differently based on how they receive their benefits.

Data and Analysis

Study 1: Attitudes Toward Welfare and the Earned Income Tax Credit

In our first set of analyses, we explore the connection between racial attitudes and support for two types of prominent downwardly-distributing social programs. To do this, we turn to data from a subsample of 1000 respondents from the 2014 Cooperative Congressional Election Study (CCES) conducted by Yougov. We restrict our analyses here and in what is to come to nonblack respondents only. The Common Content Module of the CCES asked respondents whether spending on welfare programs should be increased, kept about the same, or decreased.¹¹ This module of the CCES asked respondents, on a three-point scale, for their preferred levels of spending for “tax breaks to help subsidize the incomes of low-wage earners,” a colloquial description for the EITC. We use these responses to measure public support for direct (welfare) and tax-code (the EITC) spending on social welfare programs.¹²

To measure racial attitudes, we employ an abridged version of the symbolic racism scale developed by Henry and Sears (2002). We asked respondents for their level of agreement (on a five-point scale, ranging from “strongly agree” to “strongly disagree”) with two statements: “Irish, Italians, Jews and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors,” and “Generations of slavery and discrimination have created conditions that make it difficult for blacks to work their way out of the lower class.” Responses to these questions were added together (the second of the questions was reverse-coded) to create a scale of symbolic racism ranging from zero to eight, with higher values indicating higher levels of symbolic racism.

Do racial considerations predict attitudes toward these programs? Columns 1 and 3 of Table 1 present results of regression models predicting welfare and EITC attitudes as a simple function of our symbolic racism scale. The models for each policy area are estimated using seemingly unrelated estimation, allowing for a comparison of coefficient estimates across models. The models show, consistent with what might be expected, that support for both programs is negatively correlated with higher levels of symbolic racism: all else equal, those who score at the highest value on the symbolic racism scale are roughly 1.5 points (on a three-point scale) less supportive of welfare spending, and roughly 1.2 points less supportive of EITC spending, than those at the lowest level of symbolic racism.

Table 1 Predicting welfare and EITC attitudes

This is a very simplistic model without any control variables, but it suggests that, at least correlationally, racial attitudes are predictive of attitudes toward all manner of spending to help the less fortunate. Importantly, though, we see that the coefficient for symbolic racism in the welfare spending model is larger than the coefficient in the EITC model [(and the difference between these coefficients is statistically significant (X² = 6.78, p < 0.01)]. What this tells us is that welfare attitudes are more closely connected than are EITC attitudes to symbolic racism.

Columns 2 and 4 of Table 1 add standard demographic and political controls to these basic models, adding in variables for party identification (as measured on a 7-point scale), income, race, and gender. To this, we add two attitudinal variables known to have a particularly powerful role in shaping attitudes toward government and social spending (see, e.g., McCloskey and Zaller 1984; Hetherington 2005; Hetherington and Rudolph 2015): trust in government (as measured by a four-point response scale to the question “How often do you trust the government to do what is right?”) and a short battery of questions designed to measure egalitarianism.¹³ The inclusion of these variables does a good deal to diminish the independent effect of racial considerations for both types of spending, as racial prejudices are highly correlated with some of the other predictors (particularly egalitarianism). Some of these variables have different effects across spending mode, while some do not. But the relative magnitude of the symbolic racism coefficient across spending types, if anything, becomes larger in the presence of these controls: once controls are added, the coefficient for symbolic racism is roughly twice what it is in the EITC model, a difference that is again statistically significant (X² = 5.59, p < 0.05).

Figure 1 provides an indicator of the substantive impact of these differences, showing respondents’ expected level of support for welfare and EITC spending as a function of symbolic racism, holding the other variables constant at their mean values. Support for the EITC is higher than support for welfare regardless of a respondent’s symbolic racism score. But since symbolic racism matters more to attitudes toward welfare than it does for attitudes toward the EITC, the differences in relative levels of support for these two programs becomes greater as one moves up the symbolic racism scale. For those highest in symbolic racism, the difference in relative levels of support for the two programs is roughly twice as large as it is for those scoring lowest.¹⁴

Fig. 1

Expected levels of support for welfare and EITC spending, by symbolic racism score

The results, in general, are consistent with our assertion that social benefits delivered through the tax code are less likely to prime racial considerations than benefits delivered through direct government spending. The fact that differences in the importance of racial attitudes exist in attitudes toward “real world” social spending programs is supportive of our expectations, and sheds light on why all else equal, the EITC is more popular than welfare.

Of course, there are several limitations to this finding. Most importantly, welfare and the EITC are not directly comparable: they are, in fact, different programs with different target populations. Though (at least for the past 20 years) most direct government-assistance programs have as a requirement that recipients are either working or searching for work, and the EITC is explicitly targeted toward people who are currently in the workforce. And, of course, the word “welfare” itself is racially charged. It is certainly possible that the EITC is less likely to activate racial considerations than welfare spending because it is explicitly designed to benefit those who are working. In any case, it is clear that welfare and the EITC differ in ways that go well beyond the program delivery mechanism. To more directly isolate the effects of delivery mechanism, we turn to two survey experiments.

Study 2: Program Delivery Mechanism and Racial Considerations

Our second study uses a survey experimental approach to try to isolate the effect that program delivery mechanism has on how respondents view social benefits. This experiment was conducted using an original survey from Yougov on a nationally representative sample during September, 2015.¹⁵ It asks respondents for their level of support (as measured on a 1–5 scale, ranging from “strongly support” to “strongly oppose”) for a hypothetical downwardly-distributing social policy that provides assistance to low-income citizens. We develop a short summary of this program that provides basic information about its intent, likely beneficiaries, and cost. We vary the description of the program such that one subset of randomly-selected respondents will learn that the program is to be delivered through direct government spending, while the other subset of respondents learns that the program will be delivered through the tax code. Aside from this “delivery mechanism” treatment, all other information about the program remains essentially identical, allowing us to isolate the impact of delivery mechanism itself on program support.¹⁶ The wording for the two delivery mechanism frames are:

• Direct spending Some have endorsed a program that would provide low-income citizens with assistance in paying for groceries and other necessities. Under this program, certain citizens would receive monthly checks from the government that could be used to purchase groceries and other necessities. The total cost of this program is expected to be $65 billion per year.

• Tax expenditure Some have endorsed a program that would provide low-income citizens with assistance in paying for groceries and other necessities. Under this program, citizens would be eligible to deduct the amount of money that they pay for groceries from their taxable income. These tax credits would reduce the amount of federal income tax that these citizens pay each year, and can result in a citizen receiving a tax credit from the federal government instead of paying federal income taxes at all. The total cost of this program is expected to be $65 billion per year.

Columns 1 (direct spending) and 3 (tax expenditure) of Table 2 shows results predicting attitudes toward these programs as a function of the same two-question symbolic racism scale used in Table 1. As in Table 1, models are run separately for each spending frame using seemingly unrelated regression. And as in Table 1, we see evidence that racial considerations are significantly (χ² = 4.20, p < 0.05) more strongly predictive of attitudes toward this program when it is portrayed as a direct spending program than when it is portrayed as a tax expenditure program. Again, these results show that while symbolic racism depresses program support regardless of spending type, the impact of racial considerations is stronger for the direct spending frame.

Table 2 Attitudes toward social welfare spending as a function of spending frame

The results here do not show that overall support for a direct spending program is lessened as a function of symbolic racism. Unlike in Study 1, here we see that the mean levels of support do not vary across spending frame type: the mean level of support (on the 1–5 scale) is 3.43 for the direct spending frame, and 3.45 for the tax expenditure frame. This does not comport with our expectation that social programs delivered through the tax code will be more popular than programs delivered through direct means: both frames lead to a program that is equally well-supported by the American public. But the results are consistent with the expectation that the connection between racial attitudes and program support is stronger when the program is delivered through direct spending than when it is portrayed as delivered through the tax code.

Columns 2 and 4 add in the same battery of demographic and political controls as in Table 1. These results show that support for social welfare programs goes well beyond racial considerations and further highlight how spending mode affects the relative impact of racial considerations. After controlling for other factors, the impact of symbolic racism on attitudes toward spending in the tax expenditure frame essentially goes to zero, while symbolic racism remains a substantively important predictor of attitudes in the direct spending frame. The difference in relative magnitude of these coefficients remains statistically significant (χ² = 4.17, p < 0.05).

Figure 2 gives a sense of the magnitude of these differences, showing expected support as a function of symbolic racism and spending frame type, again holding the other variables constant at their means. This graph illustrates that while the mean support for this program does not differ significantly by spending mode, racial attitudes strongly affect the type of spending that respondents would prefer: those lowest in symbolic racism are expected to prefer the direct spending program to the tax expenditure program, while those high in symbolic racism are expected to prefer the opposite. Again, the core implication is that direct spending programs activate racial considerations in a way that otherwise identical tax expenditures do not.

Fig. 2

Expected support for government assistance, by spending type and symbolic racism

Study 3: Delivery Mechanism and Recipient Deservingness

Our final study explores the issue of recipient deservingness directly, attempting to see whether program delivery mode affects how citizens perceive the recipients of downward-distributing government aid. We build from the work of Aarøe and Petersen (2014), who use a vignette-based approach to understand how citizens view the deservingness recipients of government assistance. To do this, we commissioned a second nationally-representative survey, also, conducted by Yougov, this one during January 2016.

In these vignettes, we ask respondents to think about three different hypothetical recipients of government aid: a man who receives aid to pay for basic necessities in lieu of having a regular job, a married couple which receives aid to purchase health insurance for their low-income family, and a man who receives aid to subsidize his low-wage employment. In each of these three vignettes, we vary information about how government benefits will be provided to citizens: either through direct payments from the government, or through the tax code. Again, the goal is to hold constant all information about the aid that beneficiaries receive, save for the means through which those benefits are delivered. Wording for these vignettes is as follows:

• “Paying for necessities”

  • Direct spending: Think of a young man who has been out of work for some time. He has a high school degree and once had a regular job, but has not been able to find work that suits his expectations. The man receives federal assistance in the form of monthly checks that help him pay for groceries and basic necessities.

  • Tax expenditure: Think of a young man who has been out of work for some time. Imagine a young man who has been out of work for some time. He has a high school degree and once had a regular job, but has not been able to find work that suits his expectations. The man receives federal assistance in the form of a federal tax refund that helps him to pay for groceries and basic necessities.

• Health care

  • Direct spending: Think of a lower middle-class family of four. Neither parent works at a job that provides health insurance benefits. The family receives federal assistance in the form of a health insurance plan that is paid for in part using subsidies from the federal government.

  • Tax expenditure: Think of a lower middle-class family of four. Neither parent works at a job that provides health insurance benefits. The family receives federal assistance in the form of a tax break that allows them to deduct their health insurance premium from their taxable income.

• Wage subsidies

  • Direct spending: Think of a working father of two. This man earns $18,000 a year working at a large retail chain. His also benefits from a federal program that sends him money designed to supplement his income. He receives about $3000 in payments per year from the federal government.

  • Tax expenditure: Think of a working father of two. This man earns $18,000 a year working at a large retail chain. He also benefits from a federal program designed to supplement his income by providing him a refund on his federal tax bill. He receives about $3000 in federal tax credits per year.

For each question, respondents are asked how deserving (very, somewhat, not very, or not at all) recipients are of the aid that they receive. Respondents are randomly selected into one of these two “delivery frames” separately for each of the three programs.¹⁷ Our expectations for this section are twofold. First, we expect that those who receive government aid through the tax code will be perceived as more deserving than those who receive it through direct means. Second, we expect that attitudes toward recipients of aid through the tax code will be less strongly predicted by racial attitudes than attitudes toward recipients of direct spending.

Table 3 shows mean levels of perceived deservingness by spending frame for each of the three questions. As might be expected, those receiving aid to supplement low wage work or to purchase health care were perceived as generally more deserving than the aid recipient who was not working. These differences illustrate how closely the notion of work is tied to deservingness. But for two of these three questions (“basic necessities” and health care), people who received aid through the tax code were on average perceived as modestly but significantly (p < 0.05) more deserving than people who received aid through direct spending. Those who receive aid through the tax code are perceived as more deserving than those who receive identical packages of benefits through direct means. At least at the margins, these differences support the idea that how a social program is delivered provides information about how to think about the deservingness of its beneficiaries.

Table 3 Perceived deservingness of government aid recipients, by spending mode

The models in Table 4 test the impact of racial considerations on perceptions of deservingness, using the same symbolic racism battery as above, to test whether racial considerations affect perceptions of deservingness, and whether the effects of racial considerations vary across spending type. The results show that for all three spending programs, racial attitudes matter to how citizens perceive beneficiaries: those who score higher in symbolic racism perceive government aid recipients as less deserving of the aid that they receive, and the effect of symbolic racism is robust to demographic and attitudinal controls.

Table 4 Predicting deservingness of government aid recipients

We also see general support for the idea that spending type conditions the impact of symbolic racism on perceptions of deservingness. When it comes to health care, the results show that symbolic racism matters toward perceptions of deservingness when benefits are delivered through direct means, but do not matter when they are delivered through direct means, and the differences across models are statistically significant at the 0.05 level. A similar pattern emerges for the “paying for necessities” issue—the coefficient for symbolic racism is only about half as large in the tax expenditure frame as it is in the direct spending frame: the differences in the magnitude of the coefficients here are significant at the 0.10 level.

The results point in the same direction for the wage subsidy issue, though the effects are far weaker and not statistically meaningful. Although racial attitudes predict how deserving citizens view recipients of wage subsidies, in other words, this effect does not differ substantially across spending mode. This result runs counter to our expectations, and we have no clear explanation for why this might be the case. It may be the case, though, that since this vignette primes “work” so directly, the explicit statement that the respondent is working might render less important some of the other considerations—including spending mode—that affect how citizens view the deservingness of recipients.¹⁸

Discussion and Conclusion

This paper set out to examine if differences in how the federal government delivers social benefits influences both how citizens view both the programs and their beneficiaries. The general point of our argument was that spending through the tax code is less likely to prime respondents to think the stigma of “government spending” (and related attitudes toward distrust of government), and the racially- charged stereotypes of “recipients of government aid” than spending conducted through direct means. Citizens would thus evaluate such programs using different criteria than they would social spending delivered through more direct means. Our analyses first explored the two main ways in which government delivers social welfare benefits to lower-income citizens in the real world: welfare spending and the Earned Income Tax Credit. We found that the EITC was more popular than welfare spending, and that opinions toward the EITC were less strongly predicted by both racial prejudices than opinions toward welfare.

Our next two analyses worked to isolate the effect of delivery mechanism itself on attitudes toward benefits and beneficiaries. The results here showed that while spending through the tax code was not necessarily more popular than spending conducted directly, those who received aid via tax expenditures were (at least for two of three social programs) perceived as more deserving of what they receive than recipients of identical benefits delivered through direct means. In addition, they showed that attitudes toward spending via the tax code (and attitudes toward beneficiaries of spending through the tax code) were less driven by racial prejudices than attitudes toward direct spending.

These findings occur at the margins: we do not make the case that the public evaluates tax code spending and direct spending in fundamentally different ways. The same sorts of concepts long known to predict attitudes toward direct spending policies—trust, values, income, racial attitudes—generally predict attitudes toward spending through the tax code as well. But even so, these results point to differences in how the two modes of spending are evaluated: in general, tax code spending is evaluated through a less racialized lens. Given how racialized public support for social spending of all types continues to be (e.g., Winter 2006; DeSante 2013; Tesler 2016), even modest differences in racialization across delivery mode could have significant consequences (Stimson 2011).

These results, of course, come with limitations. First and perhaps most importantly, all of the social benefits that we presented to respondents are benefits that, in some form, exist in the real world. Though most citizens are unfamiliar with the details of federal policy programs, our experimental results may be affected to some extent by citizens’ actual experiences or perceptions of how these sorts of social benefits are delivered. To use one of the vignettes from our third analysis as an example, a “federal tax credit for low-wage work” does exist in the real world, but a “monthly payment to supplement low-wage work” does not. It would be useful in the future to replicate these results in areas where the federal government plays a less currently visible role—perhaps, areas such as paid family leave, or pre-K education—to see if the impact of delivery mechanism on attitudes also exists for programs that are purely hypothetical.

In addition, future work can do more to uncover the precise mechanisms that link choice of policy tool to public attitudes. We have argued, for example, that citizens assume that the beneficiaries of tax expenditure programs are workers and taxpayers, and therefore are perceived as more deserving. We have also argued that these heightened perceptions of deservingness occur, at least in part, because priming the notion of “taxpayer” serves to negate some of the negative racial stereotypes that come to mind when citizens think about recipients of aid. Our results are consistent with these claims, but addressing them fully calls for more direct evidence. In addition, public perceptions of “racialized” policies do not occur in a vacuum, but are also shaped, in large part, by the strategic actions of political elites to either racialize or deracialize policies in ways that affect how citizens think about these issues (see, e.g., Kellstedt 2003).

Related to this, future work could do more to explore whether the effects that we have seen here vary across types of citizens. For example, Page and Jacobs (2009), Ellis and Stimson (2012) and others find strong abstract support for government programs to reduce inequality, even among people, such as ideological conservatives and those who distrust government. But such programs, when funded directly, are also the sorts of programs subject to racial backlash, or the perception that government money will be wasted or used ineffectively, from many of the same sorts of people. If this is the case, then perhaps funding such programs through the tax code would be better able to “thread the needle,” in effect, meeting popular demand for solving certain types of social problems while minimizing the concerns that many have about specific kinds of direct efforts to solve such problems.

Our results also have a number of implications for the study of public opinion and the American welfare state. First, these results may be of use in helping to explain the rise of EITC spending in particular and social tax expenditures in general. The EITC and other tax-credit programs have grown markedly over the past few decades without much high-profile partisan disagreement (several Republican candidates in the 2016 presidential election, for example, proposed expansions of the Earned Income and Child Tax Credits), while traditional welfare programs have changed only incrementally and with much partisan discord (a group of Republicans in 2018 made efforts to tie the passage of an annual farm subsidy bill to significant cuts in the federal Supplemental Nutrition Assistance Program). Past work has suggested a number of different reasons for this divergence: the role of parties, interest groups, policy entrepreneurs, and even the different methods through with tax expenditure and direct spending policies are created (e.g., Pierson 1993; Gitterman 2010; Mettler 2011).

Our analysis points to another possible explanation: public opinion. If tax expenditure programs are more popular, or at least less racialized, than direct spending programs, policymakers will generally be freer to expand or change such programs with less fear of public backlash. It is, of course, an empirical question whether public opinion has driven the expansion of programs such as the EITC, but given the strong dynamic relationships between public opinion and government action on matters of direct spending policy (e.g., Soroka and Wlezien 2010), it is an empirical question worth testing.

More broadly speaking, our results show that the social construction of actual or potential beneficiaries of government aid is not fully exogenous to government programs themselves, but is instead influenced by the design of public policy. Public policies, by and large, tend to earn popular support when their beneficiaries are perceived as deserving. But we argue that a policy design can bestow certain virtues (such as deservingness) on a target group, and might help explain how political parties work to target government spending to electoral constituencies that are less popular with the general public.

Finally, these results have implications for understanding the political feasibility of inequality-reducing public policies, showing that, at least to some extent, tax expenditure programs can be used to deliver benefits to lower-income citizens while priming fewer of the racial stereotypes that often accompany such programs. As currently constituted, most major tax expenditures distribute benefits upward. But provided that tax credits are refundable, there is no a priori reason that tax expenditures must be regressive: we can certainly conceive of any number of downward-distributing social benefits—to finance child care, pre-K education, or job training, for example—that could be delivered through the tax code.

Tax expenditures are often targets of criticism from economists because of their relative inefficiency (see, e.g., Surrey 1970; Burman and Phaup 2012): dollars delivered directly through the federal government might be spent more accountably, effectively, and transparently than those delivered through tax expenditures. Tax expenditures also create issues with administration and compliance that are different than those created by direct spending. There is certainly reason to think that this would be the case of tax credits for (for example) child care, social insurance, or other items were greatly expanded at the expense of direct spending on these programs. The greater political palatability of tax expenditures must certainly be weighed against these other sorts of practical and economic considerations. But for policymakers seeking to reduce inequality, tax expenditures may have one critical advantage over direct spending programs: they are in all likelihood more politically feasible.