Abstract
States redistribute wealth through two mechanisms: spending and taxation. Yet studies of the social determinants of redistribution typically focus exclusively on government spending. This article explores how one determinant of social spending—racial composition—influences preferences for, and the structure of, tax systems. First, analyses of state and local tax burden data indicate that an increasing proportion of Latinos within states is associated with more regressive tax systems. Second, evidence from a nationally representative survey experiment suggests that individual preferences for taxation may be influenced by changes in the racial composition of communities. Finally, analyses reveal that in-group solidarity is a key mechanism through which racial threat shapes preferences for taxation. In demonstrating a relationship between racial change, tax preferences, and tax structures, this article contributes to our under-standing of the determinants of redistribution as well as the broader project of the new fiscal sociology.
INTRODUCTION
States redistribute wealth through two mechanisms: spending and taxation. Yet analyses of the social determinants of redistribution typically focus exclusively on government spending, ignoring the role of taxation. Although the overall level of taxation may be largely a function of spending level, the structure of tax systems can vary dramatically even across states with the same level of spending. The distribution of relative tax burden—the degree to which a tax system is progressive or regressive—is one aspect of tax structure that has important consequences for inequality and individual well-being. Motivated by Joseph Schumpeter’s ([1918] 1991) assertion that the best way to understand a society and its priorities is to analyze how it taxes its citizens, this article furthers the project of the new fiscal sociology (Martin, Mehrotra, and Prasad 2009; Martin and Prasad 2014) in examining how racial composition—shown to be an important determinant of redistributive social spending—influences the progressivity of state and local taxes in the United States.
Where racial heterogeneity is higher, government spending is lower. This association has been documented in analyses of public spending both within and between countries (see, e.g., Alesina, Baqir, and Easterly 1999; Alesina, Glaeser and Sacerdote 2001; Alesina and Glaeser 2004). In the United States, for example, racial composition is an important predictor of cash welfare spending across states, net of a host of other factors (see, e.g., Rodgers and Tedin 2006). Although cash assistance is a critical component of welfare state policy, total welfare spending in the United States is substantially less than the direct cash transfers to low-income households delivered through the tax code in the form of refundable tax credits. Indeed, tax policy is central not only to shaping aggregate levels of inequality but also to comprehending the size and generosity of the welfare state, particularly in the United States (Howard 1999; Garfinkel, Rainwater, and Smeeding 2010). Considering the structure of tax systems is therefore necessary to understanding the overall level of redistribution in a given society. At the same level of revenue, a highly regressive tax system can serve to reduce the net redistributive effect of high levels of social spending: low-income households may benefit from spending on education and welfare, but at the cost of reduced disposable income. Given the impact of tax policy on household income, it is not surprising that tax systems have been shown to influence a range of individual outcomes—including health and educational attainment—even after accounting for government spending (e.g., Strully, Rehkopf, and Xuan 2010; Newman and O’Brien 2011; Evans and Garthwaite 2014; Hoynes, Miller, and Simon 2015). Moreover, recent work using administrative tax data to study trends in intergenerational economic mobility finds a direct link between tax structure and the transmission of (dis)advantage across generations: localities with more progressive tax systems have higher levels of inter-generational economic mobility, net of the overall levels of taxation and spending (Chetty et al. 2014; Chetty and Hendren 2015). Taken together, these studies demonstrate that the structure of tax systems has an effect on life chances independent of government spending.
Beyond their redistributive functions, tax systems also provide a unique lens for analyzing larger social forces, in this case the changing nature of ethnoracial division in the United States and its role in the perpetuation of inequality through public policy. Indeed, historical accounts document the important role of racial division—at times characterized by overt racism—in structuring American tax policy from the colonial era to the present at all levels of government (see, e.g., Einhorn 2001, 2006). Although attributing policy positions to overtly racist motivations is less common today, race continues to shape policies and preferences by serving as a salient marker of group membership (see Bonilla-Silva 2006; Abascal 2015; DiTomaso 2013). Studies of public opinion reveal that support for redistributive policies is partially contingent on whether respondents believe those policies will benefit members of their own (“coethnic”) in-group (Gilens 1999; Luttmer 2001), a process of inequality reproduction Tilly labels “opportunity hoarding” (Tilly 1999). These feelings of in-group solidarity are theorized to be most salient when members of the dominant racial group feel their group position is threatened, which is likely to occur when there is an increase in the number of minorities present in the community (Blumer 1958; Blalock 1967; Enos 2014; Abascal 2015). Given this, we might expect that preferences for redistribution, including the structure of taxation, are strongly influenced by change in the racial composition of states over and above the level of racial diversity present.
This investigation provides three empirical contributions to our understanding of race as a determinant of redistribution and to the social processes that shape the structure of tax systems. First, using fixed effects regression analyses and data on state and local tax systems, I demonstrate that changes in racial composition are associated with changes in the progressivity of state and local tax systems: between 1995 and 2007, an increase in the percentage of Latinos in a state is associated with more regressive state and local tax systems and an increasing tax burden on low-income households. Second, using evidence from a nationally representative survey experiment I find that individual preferences for taxation are actively shaped by the changing ethnoracial composition of the community. Third, additional evidence from the survey experiment reveals that in-group preference—or feelings of “solidarity”—is one mechanism through which racial change shapes tax preferences.
By providing empirical evidence that racial composition influences not only preferences for taxation at the individual level but also the real structure of tax systems at the state and local levels, this investigation makes an important contribution to our understanding of the determinants of redistribution and welfare state generosity. Moreover, in analyzing how social forces shape tax structures, this article contributes to the broader project of the New Fiscal Sociology, a movement enjoying a resurgence as sociologists explore the cultural and historical roots (Morgan and Prasad 2009; see esp. Martin et al. 2009), modern politics (Martin 2008; Pearson 2014; see also Prasad and Deng 2009), and social consequences of tax systems (Newman and O’Brien 2011; Strully et al. 2010; Chetty et al. 2014; Chetty and Hendren 2015).
RACE AND REDISTRIBUTION
Racialized Origins of the American Tax State
Documenting the role of race and racial division in the development—or underdevelopment—of the U.S. welfare state has been and continues to be a focus of researchers from a variety of disciplinary backgrounds (e.g., Quadagno 1994; Lieberman 2001; see also Manza 2000). The historical and social scientific literatures on race and redistribution detail the myriad ways that race has served to not only dampen overall levels of social spending at the federal, state, and local levels but also to shape the contours of the American welfare state—perhaps most notably its decentralized (federalized) structure, which emphasizes local control. One line of research analyzes the role of race in the development of welfare state policy by focusing on how specific policy efforts to increase social spending were either curtailed or wholly blocked by racially motivated politics, most notably during key historical periods from Reconstruction to the New Deal to the Great Society eras (e.g., Quadagno 1994; Goldfield 1997). A second related analytic approach prioritizes the role of race in the creation and perpetuation of institutional barriers to welfare state development, such as super-majority voting requirements and localized control of public programs and public dollars and other “veto points” that serve to impede the implementation of progressive spending policies (e.g., Lieberman 2001).
As government spending is tightly coupled with tax revenues, particularly in earlier epochs, disentangling the role of race in the evolution of the American welfare state from the role of race in the evolution of the American tax state is complicated. In documenting how race shapes the politics of social spending, the existing literature already articulates—either directly or indirectly—one pathway through which race has influenced tax policy at all levels of government (see Alesina and Glaeser 2004). Efforts to block increases in social spending (almost always) by definition also serve to block increases in the tax level. The reverse is also true. Yet beyond the level of taxation, there is also significant historical evidence that the structure of tax systems, namely, the distribution of tax burden, has been and continues to be shaped by race and racialized politics. Race has influenced the structure of tax systems in specific historical moments as well as shaped the distribution of tax burdens over time through racially motivated institutional barriers to progressive taxation.
Analysis of tax and spending policies in the Southern states in the decades before and after the Civil War illustrates how racialized politics shaped not only the level but also the structure of tax systems (Thornton 1982; Newman and O’Brien 2011; Foner 2005). Throughout the colonial and antebellum periods, as the historian Robin Einhorn (2001, 2006) details, Southern slave owners worked to ensure that all property, including both land holdings and slave holdings, be taxed at the same rate, an idea motivated out of fear that voters may want to adopt a more progressive structure that taxed slave wealth—and therefore the rich—at higher rates. According to Einhorn (2001, 2006), these “uniformity clauses” put downward pressure on property taxes in the South, resulting in a poorly funded public sector and an institutional legacy of taxation that gave preferential treatment to property.
In the period following the Civil War, newly enfranchised blacks moved to increase social spending, particularly on education (Woodward 1971; Kousser 1980; Thornton 1982; Foner 2005). With increased spending came increased taxes, often levied with distinctly progressive rate structures, including significant new taxes on property, of which blacks owned very little. This sharp increase in the level of taxation—necessary to fund a public education system that now served blacks in addition to whites—as well as the progressive distribution of the tax burden fueled backlash and resentment among landowning whites culminating in a dramatic reappropriation of power. During this “redemption” era (Foner 2011), both the level and structure of taxation were changed as public spending was cut and the newly instituted progressive property taxes were rolled back in favor of significantly more regressive tax policies such as poll taxes, which had the added benefit of disenfranchising black Americans (Woodward 1971; Thornton 1982; Foner 2005). Efforts to roll back progressive tax policies were pursued through legislatures as well as the courts where the uniformity clauses of the antebellum period were invoked to strike down the newly implemented progressive tax instruments (on income, property, business) (Einhorn 2001, 2006; Newman and O’Brien 2011).
The racialized politics of Reconstruction structured tax policy in the South for decades to come, and not just through the inertia inherent to tax systems. Beginning in the redemption era, white landholders across the South put in place a series of institutional barriers—from supermajority requirements to state constitutional amendments—designed to impede efforts to increase any taxes and particularly progressive taxes such as those on property (Newman and O’Brien 2011; Einhorn 2001, 2006). Similar procedural obstacles to progressive taxation were put in place during the Great Depression, a period when racialized politics served as a major constraining force to President Franklin Delano Roosevelt’s New Deal agenda (Biles 1994; Wright 1986, 2010). And three decades later, as racial tensions flared during the civil rights era, yet more procedural barriers to progressive taxation were put into place: following the passage of the Voting Rights Act of 1965, Louisiana adopted a constitutional amendment requiring a two-thirds majority to increase any tax in the state. Mississippi followed suit with a three-fifths majority. Similar institutional barriers swept states and localities across the country in the decades that followed (Knight 2000; Mullins and Wallin 2004; Waisanen 2008; Newman and O’Brien 2011; see also Pearson 2014).
Although more work is needed to elucidate the special role of race in the evolution of tax policy at the federal, state, and local levels, existing research demonstrates that racial division was pivotal to the successful implementation of regressive tax schemes in key historical moments as well as to the institutionalization of barriers to progressive taxation—such as supermajorities and constitutional limits—that continue to shape the structure of tax systems to this day: in 2007 the state of Alabama settled a lawsuit that asserted that the underfunding of public higher education in the state is the direct result of a century-old discriminatory tax system (Walker 2007).
Race, Spending, and Taxation in the Modern Era
Complementing historical analyses of how racial division undermined progressive social policies throughout the 19th and 20th centuries, recent empirical studies of the link between racial composition and levels of redistributive social spending have found a consistent inverse (and arguably causal) relationship (Orr 1976; Cutler, Elmendorf, and Zeckhauser 1993; Ribar and Wilhelm 1996; Poterba 1997; Alesina, Baqir, and Easterly 1999; Alesina, Glaeser, and Sacerdote 2001; Alesina and Glaeser 2004; Rodgers and Tedin 2006). Alesina, Glaeser, and Sacerdote (2001) demonstrate, for example, that redistributive social spending is lower when beneficiaries are disproportionately from minority backgrounds and find that this association holds both within and between countries. In one study of U.S. cities, for example, Alesina, Baqir, and Easterly (1999) find that spending on productive public goods, including education, sewers, and roads, is lower in cities with higher levels of ethnic fragmentation. More recent work analyzing implementation of the Temporary Assistance for Needy Families program following the welfare reforms of the mid-1990s finds racial heterogeneity to be an important predictor of a state adopting more restrictive and punitive rather than inclusive welfare policies (Soss et al. 2001; Fellowes and Rowe 2004).
In his analysis of attitudes toward redistributive spending, Luttmer (2001) finds that levels of support for welfare spending are strongly influenced by the proportion of local welfare recipients that are coethnics. He concludes that support for social programs is higher where welfare spending is perceived to benefit members of one’s own ethnic group. This “racial group loyalty” shapes preferences for welfare spending net of a respondent’s own income level; even low-income households who stand to benefit from greater welfare spending are less likely to support increased spending when they perceive welfare to disproportionately benefit other racial or ethnic groups. In his highly influential study, Gilens (1999) similarly finds that antiwelfare attitudes among whites are motivated by assumptions that this spending disproportionately advantages minorities (specifically blacks) whom whites view as being lazy and lacking work ethics. These stereotypes undermine feelings of solidarity with minority groups and thereby reduce whites’ motivation to mobilize resources for their benefit.
In analyzing determinants of social spending, these studies provide indirect evidence for how racial composition influences the level of taxation, given the relationship between the two. Yet these studies tell us little about how race might influence the structure of tax systems, specifically the distribution of tax burden.
Two articles in the sociological study of taxation attempt to address this gap in the literature by analyzing changes in the federal tax code. Using event history analysis to explore changes in the U.S. federal tax code over the second half of the 20th century, Jacobs and Helms (2001) find that events that reflect positively on racial minorities, such as peaceful civil rights demonstrations, are associated with progressive changes to the federal tax code, whereas events that reflect negatively on racial minorities, such as violent urban riots, are associated with regressive changes to federal taxes. Notably, the authors find no evidence for a direct effect of changes in the racial composition of the population on the structure of the tax system. Earlier work by Jacobs and Waldman (1983) did find some evidence that racial composition influenced the structure of tax systems at the subnational level, specifically that states with a higher proportion of black residents had more regressive tax systems.
A more recent political science study of local tax votes in Texas and Massachusetts suggests that it may not be the level of ethnoracial diversity present that influences taxation but rather the degree of demographic change. Analyzing local tax votes in these two states, Hopkins (2009) finds no evidence of an association between the level of racial heterogeneity present and the likelihood a community votes to increase taxes, yet he does find strong evidence of reduced support for increased taxes in communities that had experienced recent demographic change, specifically an increase in racial heterogeneity. The idea that racial change is the key driver of both shifting public opinion and public policy, as evidenced in this study by Hopkins, motivates both the theoretical model and empirical strategies employed below.
Group Position, Opportunity Hoarding, and Racialized Preferences for Redistribution
How might an increase in the presence of ethnoracial minorities influence preferences for tax policy? According to Blumer (1958), racial prejudice is engendered when the dominant group (e.g., whites) feels its sense of group position—atop the hierarchy and/or in control of economic and political resources—is challenged (see also Bobo and Hutchings 1996; Bobo 1999). Minority groups do not need to actively challenge the dominant group, however, for the feelings of threat to be salient. As Blalock asserts (1967), simply the increased presence or visibility of racial or ethnic minorities may increase the sense of threat felt by the (white) majority over control of economic resources and political power.
The cognitive basis for this sense of threat has been systematically explored by social psychologists. Experimental evidence demonstrates that negative stereotypes held by whites about blacks and Latinos, particularly low-income blacks and Latinos, are derived from reduced feelings of warmth and perceptions of competency (Fiske et al. 2002). Low-income minorities are seen to be in competition with the majority population, evoking a “contemptuous prejudice” (Fiske et al. 2002; Cuddy, Fiske, and Glick 2007) among whites. Therefore, differential group membership premised on ethnoracial distinctions may engender preferences for policies that actively harm the out-group (see, e.g., Enos 2014).
Yet the forces that perpetuate racial inequality today need not be, and indeed may not be, compelled by active feelings of race prejudice (Bonilla-Silva 2006; DiTomaso 2013). Charles Tilly’s (1999) notion of opportunity hoarding may be a more fruitful theoretic lens for understanding how race shapes preferences for and patterns of redistribution. Instead of analyzing redistributive preferences and processes as a site for overt racialized conflict—where dominant whites actively harm subordinate minority populations through reduced welfare spending or higher tax rates—Tilly asserts that the same social outcomes can be produced through a process in which whites acquire power and resources that are used for the benefit of in-group members. There need not be active, deliberate harm to out-group members for inequalities to be perpetuated.
For Tilly, the key to understanding how or why a dominant group (e.g., whites) may serve to perpetuate racial inequality is not the degree to which they fear or hate an out-group (e.g., blacks, Latinos) but rather the degree to which in-group members feel bound by notions of solidarity and loyalty (and often reciprocity, as emblematic of immigrant networks). Where solidarity is higher, the motivation, and perhaps even capacity, for the group to “hoard” opportunities for the benefit of group members is greater (Tilly 1998). Indeed, Tilly points to the politics of inequality in both wealthy and developing countries as a strategic site for the hoarding of political rights and economic spoils by “categorically bounded” groups, with particular emphasis on ethnoracially dominant groups and men (1998, pp. 10, 194–212).
How does Tilly’s notion of opportunity intersect with Blalock’s conception of racial threat? A recent study by Abascal (2015) demonstrates that opportunity hoarding is a dynamic process; the degree to which the dominant group hoards resources for in-group members is in part a function of the degree to which they feel threatened. In her behavioral experiment, Abascal finds that priming respondents with information about recent growth in the Latino population makes whites less generous to blacks in a dictator game setting. In this study, increasing racial threat from one out-group—Latinos—leads whites to hoard resources in a way that reduces the welfare of a different out-group—blacks. Under threat, in-group members seek to benefit their own, which in turn increases inequality between the in-group and all out-groups, regardless of the source of threat.
Solidarity within the categorically bounded group of white Americans—perhaps heightened in areas with increasing populations of ethnoracial minorities—may engender a set of policy preferences that favors a more progressive taxation regime in ethnically homogenous communities and a more regressive taxation regime in increasingly heterogeneous communities. And, importantly, race may influence individual preferences for, and the actual structure of, tax systems distinct from how it shapes preferences for and the actual level of social spending and taxation.
CHANGING RACIAL COMPOSITION AND THE PROGRESSIVITY OF STATE AND LOCAL TAXES
Informed by the above empirical work on the determinants of taxation and redistributive social spending as well as the theoretical understanding of how racial change may influence opportunity hoarding by dominant groups, I posit two guiding hypotheses for how changing racial composition at the state level may influence the progressivity of state and local tax systems:
Hypothesis 1.—As the percentage of Latinos and the percentage of blacks in a state increase, state and local tax systems will become less progressive.
Hypothesis 2.—As the percentage of Latinos and the percentage of blacks in a state increase, the tax burden on low-income households will increase.
Before proceeding to the first set of empirical analyses on the link between racial composition and the changing structure of state and local tax systems, I begin with a discussion of how tax progressivity is operationalized in the current investigation.
Measuring Tax Progressivity
Measuring the progressivity of tax systems can be conceptually problematic and practically difficult. In the most basic sense, tax systems are understood to be progressive when higher-income earners pay a greater proportion of their income in taxes than those with lower incomes. Statements about progressivity, therefore, are inherently comparative, that is, a given tax or tax system is said to be more or less progressive than another tax or tax system.
One approach to measuring the progressivity of a tax system holistically is to analyze the mix of taxes used to fund the public sector. Comparisons of tax systems across nations reveal that, on average, sales and excise taxes are more regressive than income and property taxes (Prasad and Deng 2009). Whereas any tax instruments can be designed to be more or less progressive, greater reliance on the sales tax is, on average, associated with more regressive tax systems. Although understandably useful in making cross-national comparisons, systematic analysis of tax systems at the subnational level necessitates a finer level of detail, particularly in the United States, where the use of specific tax instruments varies across states and localities; for example, a number of states do not levy a sales tax and another subset of states does not levy a personal income tax.
Another way to measure progressivity is to calculate the ratio of the tax burden at two different points in the income distribution, for example, the ratio of the tax burden on the top quintile to the tax burden on the bottom quintile. Although intuitively appealing, measures of this type fail to incorporate any information about the distribution of the tax burden across quintiles. Moreover, this type of measure does not incorporate any information about the actual distribution of income or degree of income inequality. The latter information is central to our understanding of tax progressivity; for example, holding the relative tax burden of the top and bottom earners constant, we may have a very different interpretation of progressivity if the income gap between the top and bottom earners is 10:1 rather than to 2:1.
One measure of tax progressivity that considers inequality in both income and taxes is the Suits index, named for economist Daniel Suits (1977) who developed the measure in the 1970s. Analogous to the Gini index for income or wealth inequality, the Suits index captures the cumulative percentage of taxes paid across households as a proportion of the cumulative percentage of income earned across households. It therefore simultaneously incorporates information on both the distribution of income as well as the distribution of taxes. A Suits index of 1—extreme progressivity—is when all of the taxes are paid by the highest earner. A Suits index of −1—extreme regressivity—is when all taxes are paid by the lowest earner.
Ideally, the Suits index of tax progressivity would be constructed using data on the income earned and taxes paid by all households. Given data limitations, however, this information is often approximated using data on the cumulative income earned and taxes paid by households at various points in the income distribution. For the purposes of this investigation, the Suits index will be calculated using data on the income earned and taxes paid from seven different points in the income distribution: bottom 20%, second 20%, third 20%, fourth 20%, next 15%, next 4%, and top 1%.
By incorporating information on the total income earned and taxes paid across the income distribution, the Suits index overcomes many of the limitations of other measures of tax progressivity. Yet it should be noted that this measure—indeed any attempt to measure “progressivity”—fails to capture any information on the overall level of taxation. For this reason, differences in tax “progressivity” across states cannot be used to directly infer differences in tax “burden” across levels of income. The tax burden on the poor could be very high in a state that is still considered progressive (if taxes on everyone are high), and the tax burden on the rich could be very low in a state that is still considered progressive (if taxes on everyone are low). As the actual level of the tax burden on the poor has been shown to be associated with a number of measures of individual and household well-being (Newman and O’Brien 2011), the analysis presented below will additionally explore a connection between racial composition and the actual tax burden of the poor.
Data and Methods
The power to tax is vested at many levels of government. In addition to federal and state governments, myriad and often overlapping subunits within a state—including municipalities, counties, school districts, and even special taxing districts—are empowered to tax their residents. The structure of tax systems therefore varies widely across the states, both in which government entity levies taxes and what they tax. In some states, for example, the general sales tax is reserved as an instrument of the state government, whereas in other states sales tax rates are determined and revenue is collected at the county level for local use, and in still other states the sales tax is used at both the state and local level. The same is true for property and personal income taxes. Given differences both across states and over time in the role of local versus state governments in both taxing and spending, this study aggregates state and local spending and uses state as the unit of analysis.2
Data for this analysis are taken from the Institute for Taxation and Economic Policy (ITEP), a nonprofit, nonpartisan, research organization with an expertise and focus on U.S. tax policy at the federal, state, and local levels. Since 1996, ITEP has maintained and updated a microsimulation model for estimating the net effect of current and proposed laws on overall tax revenues as well as the tax burden of households by income quintile. In addition to routinely updating data on federal, state, and local tax laws, the ITEP micro-simulation model is estimated using data from a number of sources, including hundreds of thousands of individual tax returns from the IRS public use files as well as the Current Population Survey (CPS), American Community Survey, and Survey of Consumer Expenditures.3
Although comparable to microsimulation models used by the U.S. Treasury, the Congress, and the Congressional Budget Office, ITEP’s model is unique in its ability to simulate tax incidence at the state level, including aggregated state and local estimates. The degree of detail in the model is essential for estimating the true “net effect” of tax law on households: published tax rates for personal income, property, or sales taxes can vary substantially from the effective tax rate due to numerous credits, offsets, and exemptions that exist at the federal, state, and local levels. The ITEP model also incorporates estimates of indirect taxes that may be paid by households; most notably, for example, households whose heads do not own their own homes are nevertheless considered to bear some brunt of property taxes in a locality by way of higher rent.
Published estimates of tax burden by income for all 50 states using ITEP’s microsimulation model are available at three points in time: 1995, 2002, and 2007. In each of these years, ITEP calculated the total tax burden paid by households in the first 20%, second 20%, third 20%, fourth 20%, next 15% (80th–95th percentile), next 4% (96th–99th percentile), and top 1%.
To calculate the Suits index for each of the 50 states at each point in time, I combined this tax burden information with the state-specific income distribution information used by ITEP to calculate tax burden. Using the ITEP model results, I calculated the total income earned in each state and then calculated the proportion of income earned by each of the bottom four quintiles, the next 15%, the next 4%, and the top 1%. I then calculated the total amount of state and local taxes paid overall and then calculated the proportion of taxes paid by each of the bottom four quintiles, the next 15%, the next 4%, and the top 1%. Relating the cumulative income earned to the cumulative taxes paid according to the function outlined by Suits generates a state-specific Suits index for the progressivity of the state and local tax systems of each of the 50 states in 1995, 2002, and 2007, the three years for which data are available.
The above calculations yield a Suits index for each of the 50 states in the United States at three points in time, or 150 state-year observations in total. To ease interpretation of results in the regression models, I standardized the Suits index by dividing by the standard deviation. As a reminder, a (hypothetical) Suits index of 1 indicates a tax system is perfectly progressive, whereas a (hypothetical) Suits index of −1 indicates a tax system is perfectly regressive. Therefore, in looking at within-state change in the Suits index over time, an increase in the value of the Suits index reflects a move toward greater progressivity, whereas a decrease reflects a move toward greater regressivity.
The calculated Suits index for each state and each year is reported in the appendix (table A1). Notably, every state-year estimate is below zero, meaning that all state and local tax systems can be characterized as regressive. Note also that states which are generally considered relatively “progressive” or generous in terms of redistributive social spending do not always have progressive tax systems, for example, Massachusetts. Moreover, some states that are often characterized as being laggards in social spending have relatively progressive tax structures, for example, South Carolina. This discrepancy further underscores the need to analyze the social determinants of taxation separately from the determinants of redistributive social spending.
As described above, the primary focus of this investigation is to test for an association between a change in the racial composition of a state and a change in the progressivity of state and local tax systems. Therefore, data are analyzed using a linear model with state and year fixed effects. State fixed effects are employed to net out all time-invariant state-specific characteristics that may be associated with racial composition and the structure of tax systems. All models also include year fixed effects to net out national level trends and federal policy shifts that affect all states equally.
The key predictors for racial composition are %black (non-Hispanic), %Latino, and %Asian (non-Hispanic) in the state. These numbers are derived from the CPS and are scaled to 0–100. I use percentage of the population by ethnoracial category over other commonly used metrics of heterogeneity, such as ethnic diversity or “fractionalization” indices, for several reasons. First, blacks, Latinos, and Asians (and other ethnoracial or nation origin groups) have distinct histories and historical trajectories in the United States; indeed, historically the nature and level of welfare support low-income households receive vary across ethnoracial categories, beyond a white-nonwhite dichotomy (Soss et al. 2001; Fox 2012). Second, the content of race-based stereotypes and racial prejudice varies significantly across these groups (Gilens 1999; Fiske et al. 2002; Cuddy et al. 2007). Third, given that we are particularly interested in looking at change in racial composition over time, having separate indicators for the percentage of each minority group in the population allows for a more nuanced examination of demographic change relative to a summary index or more general measure such as % nonwhite. Although summary indices may be necessary in cross-national studies where minority distinctions vary and may be premised on religion or language (e.g., Alesina, Glaeser, and Sacerdote 2001), using these distinct—but still coarse—measures is important and potentially instructive in the U.S. context.
In addition to state and year fixed effects, models also include a vector of time-varying state-level covariates to account for factors that may be jointly correlated with the changing racial composition of the state and the changing progressivity of the tax system. First, I include a measure of the percentage of state residents who are foreign born, to ensure models are isolating the effect of changing ethnoracial composition net of immigration status. Second, models include a number of measures from the CPS to capture the macroeconomic condition of the state, including labor force participation rate (scaled 0–100), unemployment rate and poverty rate (scaled 0–100), and total income per capita (logged), which is an aggregate measure of all income earned by individuals in the state, adjusted for inflation. Log income per capita is highly correlated with state gross domestic product and, in addition to serving as a proxy for macroeconomic conditions, serves as a control for changes in the taxable resources available to a state, which likely has important implications for the structure of tax systems.4 Political party control—which can both influence tax policy and may be shaped by changing demographics—is accounted for by a continuous measure of Republican control of the state house, taken from Harmon (2011).5 Finally all models include state-specific Gini coefficients, as inequality may be associated with both changing racial composition as well as changing distribution of tax burden.
To correct for potential autocorrelation across waves, standard errors are clustered at the state level. Models are robust to various lag structures; unlagged results are presented in the tables below (see Lynch 2011).
Results
Table 1 presents results from the regression model predicting the progressivity of state and local tax systems, as operationalized by the Suits index. In the basic random effects model, the coefficient for %Latino is negative and statistically significant: across states, those with higher Latino populations have more regressive tax systems, net of all covariates. The coefficients for black and Asian are not significant. Although this cross-sectional evidence is suggestive, the unique (often racialized) histories of state-local tax policy yielded significant differences across states that must be netted out in order to evaluate whether changing racial composition is associated with changing tax structures net of underlying differences in tax systems.
TABLE 1.
Models Predicting Tax Progressivity and Tax Burden on First Quintile by State: 1995, 2002, 2007
| Suits Index (STD)
|
Q1 Total Tax Burden
|
|||
|---|---|---|---|---|
| Model 1: State Random Effects | Model 2: State Fixed Effects | Model 3: State Random Effects | Model 4: State Fixed Effects | |
| %Latino (0–100) | −.05* | −.08* | .02 | .26* |
| (.02) | (.04) | (.05) | (.11) | |
| %black (0–100) | −.00 | .02 | −.02 | .17 |
| (.01) | (.03) | (.04) | (.16) | |
| %Asian (0–100) | −.00 | −.00 | .00 | −.07 |
| (.00) | (.01) | (.02) | (.06) | |
| %foreign born (0–100) | .07 | .08 | .03 | .01 |
| (.04) | (.04) | (.09) | (.20) | |
| Gini index (0–100) | .00 | .00 | .12 | .03 |
| (.02) | (.02) | (.10) | (.10) | |
| Total income per capita (logged) | −4.47** | −5.53** | 4.66 | 11.01+ |
| (1.29) | (1.62) | (3.35) | (6.14) | |
| %unemployed (0–100) | .08+ | .05 | .22 | .35 |
| (.04) | (.05) | (.25) | (.30) | |
| %in poverty (0–100) | −.04 | .00 | .09 | .07 |
| (.04) | (.06) | (.16) | (.19) | |
| %in labor force (0–100) | .05+ | .04 | −.13 | −.10 |
| (.02) | (.04) | (.08) | (.14) | |
| State house | −.06 | −.06 | .05 | .33 |
| (.04) | (.04) | (.12) | (.20) | |
| R2 | 0.15 | 0.92 | 0.14 | 0.76 |
Note.—All models feature year fixed effects; N = 150. SEs (in parentheses) are clustered at the state level.
P < .05, two-tailed tests.
P < .01.
P < .001.
Model 2 presents results from a model that includes state fixed effects. Here we see evidence that within-state change in the %Latino is associated with changes in the progressivity of state and local tax systems: a one-percentage-point increase in the proportion of the state population that is Latino is associated with a 7.7% standard deviation decrease in the progressivity of state and local taxes over this time period. This association exists net of changes in state macroeconomic characteristics %foreign born, level of income inequality, and political party control. Notably, %black and %Asian do not appear to be significant predictors of changes in the progressivity of state and local tax systems over this period, which is not unexpected given the smaller change in those groups at the state level relative to the changes in %Latino over this time period.
This model provides compelling evidence that racial composition at the state level structures the progressivity of tax systems. Yet, as noted above, changes in progressivity may not directly translate into changes in absolute tax burden. Recall that a tax system can simultaneously become more regressive while actually reducing the net tax burden on those at the bottom if, for example, the tax burden on those at the top is also reduced. Although changing progressivity is an important process of interest in its own right, we might also be interested to know the association between changing racial composition and changes in the absolute tax burden, particularly the burden on low-income households whose well-being is likely to be strongly affected by tax policy (Newman and O’Brien 2011). Using total tax burden on those in the bottom quintile estimated by ITEP as the dependent variable, models 3 and 4 in table 1 test for this relationship using random effects and fixed effects specifications, respectively. Although the random effects model suggests that the overall percentage of Latinos is not associated with higher taxes on the poor, the preferred fixed effects specification provides evidence that an increasing percentage of Latinos in the state is indeed associated an increase in the absolute tax burden on low-income households.
The above analyses demonstrate that an increasing percentage of Latinos in a state is associated with a more regressive state and local tax system and, moreover, this increasing regressivity is being driven in part by increasing the tax burden on low-income households. The fixed effects modeling strategy coupled with the macroeconomic and political covariates helps to rule out competing explanations for this association and to isolate a direct, causal effect for racial composition on the structure of tax systems. Notably, %Latino is a significant predictor of both tax progressivity and tax burden on the poor even after accounting for %foreign born in the state (which in itself is not a significant predictor of tax progressivity or burden on poor).
CHANGING RACIAL COMPOSITION AND PREFERENCES FOR TAXATION
Whereas the above analyses provide compelling evidence that changing racial composition at the state level influences the structure of state and local tax systems, the nature of the data makes it difficult to directly test for mechanisms through which ethnoracial change may shape tax progressivity. A first-order hypothesis is that ethnoracial change may influence the structure of tax systems by shaping the preferences of voters, which in turn shape policy. Individual voter preferences, or mass public opinion, play an important role in shaping policy priorities and policy outcomes, although the strength of this relationship is contingent on many factors (Gilens 2005; Brooks and Manza 2007). Given the role of race in structuring opinions toward redistributive social spending and resource allocation across groups (Gilens 1999; Luttmer 2001; Abascal 2015) and suggestive evidence that this effect may extend to preferences for taxation (Jacobs and Waldman 1983; Jacobs and Helms 2001; Hopkins 2009), it is instructive to test how racial change may shape the tax preferences of individuals.
Is there evidence that changes in the ethnoracial composition of local areas influence preferences for taxation? Motivated by this question, I designed a survey experiment to test three additional hypotheses:
Hypothesis 3.—Whites who are told to imagine that their community has experienced a recent influx of black or Latino residents will be less likely to support higher taxes.
Hypothesis 4.—Whites who are told that their community has experienced a recent influx of black or Latino residents will be more likely to prefer a regressive, flat tax over a more redistributive, progressive tax.
Building on the notional mechanism offered by Tilly (1998) and suggested by Luttmer (2001), the final hypothesis attempts to experimentally test for how feelings of solidarity may influence preferences for taxation and mediate any observed relationship between racial composition and preferences for taxation.
Hypothesis 5.—Respondents’ feelings of solidarity with members of their community will mediate the effect of ethnoracial change on preferences for taxation.
Survey Experiment
For the experimental study I recruited a sample (n = 1,030) of individuals to complete an online survey. The sample was recruited from a privately managed online panel and was designed to be nationally representative of the United States by income, age, and gender.6 Respondents were presented with a vignette that asked them to imagine they were living in a hypothetical county that recently experienced rapid population growth. Respondents were randomly assigned to one of three conditions where they were told the new arrivals to the county were predominately (1) white, (2) black, or (3) Latino. All respondents were told that the new arrivals to the county were generally low-income, in order to reduce the likelihood that tax preferences were based on inferences about the income of the new arrivals. The vignette went on to note that the local government has decided it needs to raise additional revenue and asked the respondents for their opinions about how taxes should be raised (see full vignette text in the appendix).
The analytic sample is limited to non-Hispanic whites who successfully passed the manipulation check at the end of the survey by identifying the race or ethnicity of the new arrivals described in their treatment condition (n = 473).7 A multinomial logistic regression model predicting respondent status in each of the conditions confirms covariate balance was achieved across groups through randomization (see table A2 in the appendix). In addition to analyses of the full sample, I also conducted analyses on subsets of the sample by sex, given evidence in the literature that men and women tend to have substantially different preferences for redistribution, with women consistently favoring higher levels than men (see Alesina and Giuliano 1999; Fong 2001; Pinker 2006). Data were analyzed using ordinary least squares (OLS) models adjusting for basic demographic covariates including income (log), age (linear and squared terms), education, and marital status. Although all results presented are from models with covariates, the results are not sensitive to the inclusion of controls (see Mutz 2011; Gerber and Green 2012).
After reading the initial vignette, respondents were first asked whether they would support or oppose the county council’s efforts to raise taxes in the county (1–6 scale with 1 being “strongly oppose” and 6 being “strongly support”). As illustrated in table 2 and figure 1, support for increased taxation was highest among white respondents who were told that the new arrivals to the county were predominately white. Conversely, white respondents who were told that the new arrivals were black or Latino were less likely to support raising taxes, although notably only the Latino condition was statistically different from the white condition. Being in the Latino condition was associated with a .37 point lower score on a 6-point scale, equivalent to one-quarter of a standard deviation.
TABLE 2.
OLS Regression Predicting Support for Tax Increase by Race Condition
| Full Sample | |
|---|---|
| Experimental condition (ref = white): | . . . |
| black condition | −.23 |
| (.17) | |
| Latino condition | −.37* |
| (.17) | |
| Household income (log) | .03 |
| (.09) | |
| Male | .12 |
| (.13) | |
| Age | −.09** |
| (.03) | |
| Age2 | .00** |
| (.00) | |
| Never married | .10 |
| (.18) | |
| Education (ref = high school): | |
| < high school | .34 |
| (.43) | |
| Some college | .32+ |
| (.18) | |
| College | .25 |
| (.19) | |
| > college | .34 |
| (.25) | |
| Constant | 4.72*** |
| (1.23) | |
| N | 473 |
| R2 | .04 |
Note.—Nos. in parentheses are robust SEs.
P < .10, two-tailed tests.
P < .05.
P < .01.
P < .001.
Fig. 1.
Support for tax increase by race condition (marginal effects from OLS regression; bars are ± 1 SE). +P < .10, * P < .05 denote statistically different from white condition.
The above finding provides supportive evidence for the first hypothesis: increasing the percentage of nonwhites in a local population reduces support for efforts to increase taxation. But does the changing racial composition of an area influence respondents’ preferences for the structure of taxes? To test for this association, respondents were asked the following question: “Overall, would you prefer a ‘flat tax’ (where everyone would pay the same percentage of their income in new taxes) or a ‘progressive tax’ (where the tax rate increases with income, in other words, middle- and higher-income people would pay a greater percentage of their income in new taxes than lower-income people)?” Respondents were asked to report whether they preferred a “flat tax” or a “progressive tax” (1–6 scale, with 1 being “strongly prefer flat tax” and 6 being “strongly prefer progressive tax”). As shown in table 3 and figure 2, analysis of the full sample of respondents found no statistically significant difference between the white condition and either of the minority conditions in preferences for a flat versus a progressive tax. However, when subsetting the sample by sex of respondent, substantial differences emerge. Specifically, male respondents in both the black and Latino conditions were significantly more likely to prefer a flat tax than male respondents in the white condition; when the new arrivals to the county were nonwhite, support among white males for a progressive, redistributive tax was lower. Put another way, white male respondents who were told that the new arrivals to the county were white were more likely to prefer a redistributive, progressive tax than those who were told that the new arrivals to the county were black or Latino. The coefficient point estimates reflect that, relative to those in the white condition, the average response for male respondents in the black condition was about .57 points lower on the 6-point scale (approximately 30% of a standard deviation) and the average response for male respondents in the Latino condition was about .78 points lower (approximately 40% of a standard deviation).
TABLE 3.
OLS Regression Predicting Preference for Progressive Tax by Race Condition
| Full Sample | Men | Women | |
|---|---|---|---|
| Experimental condition (ref = white): | |||
| Black condition | −.12 | −.57+ | .48 |
| (.23) | (.32) | (.33) | |
| Latino condition | −.32 | −.78* | .24 |
| (.23) | (.32) | (.33) | |
| Household income (log) | −.44*** | −.38* | −.52** |
| (.12) | (.17) | (.18) | |
| Male | −.29 | — | — |
| (.19) | |||
| Age | −.02 | −.07 | −.01 |
| (.05) | (.07) | (.06) | |
| Age2 | .00 | .00 | .00 |
| (.00) | (.00) | (.00) | |
| Never married | .09 | .45 | −.34 |
| (.25) | (.36) | (.32) | |
| Education (ref=high school): | |||
| < high school | −.02 | −1.27+ | .99 |
| (.68) | (.65) | (.95) | |
| Some college | .24 | .20 | .25 |
| (.25) | (.40) | (.32) | |
| College | −.01 | −.04 | .09 |
| (.27) | (.41) | (.36) | |
| > college | .33 | .20 | .80 |
| (.35) | (.49) | (.56) | |
| Constant | 8.68*** | 9.43*** | 8.73*** |
| (1.64) | (2.38) | (2.24) | |
| N | 473 | 244 | 229 |
| R2 | 0.06 | 0.09 | 0.07 |
Note.—Nos. in parentheses are robust SEs.
P < .10, two-tailed tests.
P < .05.
P < .01.
P < .001.
Fig. 2.
Preference for progressive taxation by race condition (marginal effects from OLS regression; bars are ± 1 SE). +P < .10, * P < .05 denote statistically different from white condition.
Interestingly, female respondents in this sample reported no statistically significant differences in their preference for flat versus progressive taxation across the three conditions, suggesting that women may view tax policy differently from spending policy when considering redistributive policies (see Alesina and Giuliano 1999; Fong 2001; Pinker 2006). It is also notable that the order of conditions is consistent in both figures 1 and 2: the point estimates are lower for the Latino condition than the black condition in both instances, although it is important to note the difference between these two conditions is not statistically significant in either case. Taken together, these results provide some support for hypothesis 4: increasing the presence of racial minorities reduces support for progressive tax structures, although notably only for male respondents.
Overall support for taxation—and for male respondents, support for progressive taxation—is lower for whites who are told their community has recently experienced an influx of nonwhites. But what drives this association? Following Tilly (1998), I hypothesized that respondents would be more likely to support raising taxes and, specifically, to support raising taxes progressively, when they felt “solidarity” with the new arrivals, that is, when they felt the new arrivals were like them and therefore more readily identified as being part of the “in-group.” To test for this potential mechanism, respondents were asked the degree to which they believed the new arrivals to the county were like them or not like them (6-point scale with 1 being “not at all like me” and 6 being “just like me”). As shown in table 4 and figure 3, those in the black and Latino conditions were significantly less likely to report that the new arrivals to the county were like them than those in the white condition. In other words, experimentally manipulating the race of the new arrivals significantly altered the degree to which respondents felt solidarity with the group, with the average response for those in black and Latino conditions being .6 and 1 point lower on the 6-point scale, respectively (which translates to approximately 44% of a standard deviation lower for the black condition and 78% lower for the Latino condition, relative to the white condition). Notably, respondents in the Latino condition felt less solidarity with the new arrivals than those in the black condition, a difference that is statistically significant (P < .01).
TABLE 4.
OLS Regression Predicting Solidarity with New Arrivals by Race Condition
| Full Sample | |
|---|---|
| Experimental condition (ref= white): | |
| Black condition | −.60*** |
| (.15) | |
| Latino condition | −1.06*** |
| (.14) | |
| Household income (log) | −.20* |
| (.09) | |
| Male | .00 |
| (.13) | |
| Age | −.01 |
| (.03) | |
| Age2 | .00 |
| (.00) | |
| Never married | −.30+ |
| (.16) | |
| Education (ref = high school): | |
| < high school | −.23 |
| (.58) | |
| Some college | .07 |
| (.18) | |
| College | −.06 |
| (.18) | |
| > college | −.34 |
| (.25) | |
| Constant | 5.96*** |
| (1.08) | |
| N | 473 |
| R2 | .13 |
Note.—Nos. in parentheses are robust SEs.
P < .10, two-tailed tests.
P < .05.
P < .01.
P < .001.
Fig. 3.
Solidarity with new arrivals by race condition (marginal effects from OLS regression; bars are ± 1 SE). +P < .10, * P < .05 denote statistically different from white condition.
But do these differences in expressed solidarity by racial condition account for any of the observed differences in preferences for taxation across the three conditions? Results from a mediation analysis are presented in table 5. Model A reproduces results from the earlier analysis that found that support for taxes was significantly lower among respondents in the Latino condition relative to the white condition. To test whether respondents’ expressed feelings of solidarity with the new arrivals mediate the association between the race of the new arrivals and support for taxation, in model B I added the measure of solidarity to the equation. Results demonstrate that solidarity is a significant predictor of support for taxation—the higher a respondent’s feeling of solidarity toward the new arrivals, the greater the support for increased taxes. At the same time, when I include the measure of solidarity in the model, the coefficients for both the black and Latino conditions shrink toward zero. Indeed, once solidarity is included in the model, the average response for those in the Latino condition shrinks toward zero and is no longer statistically different from the white condition. Changes in the Latino coefficient between the two models suggest that variation in feelings of solidarity can account for approximately 57% of the observed association between being in the Latino condition and reduced support for taxation (Sobel mediation test significant at P < .01).
TABLE 5.
Solidarity as Mediator of Preferences for Taxation across Race Conditions (OLS Regression)
| Support for Taxes (Full Sample)
|
Preference for Progressive Tax (Men Only)
|
|||
|---|---|---|---|---|
| Model A | Model B | Model C | Model D | |
| Solidarity | — | .20*** | — | .26** |
| (.06) | (.10) | |||
| Experimental condition (ref = white): | ||||
| Black condition | −.23 | −.11 | −.57+ | −.41 |
| (.165) | (.17) | (.32) | (.32) | |
| Latino condition | −.37* | −.16 | −.78* | −.50 |
| (.17) | (.18) | (.32) | (.34) | |
| Household income (log) | .03 | .07 | −.38* | −.34 |
| (.09) | (.09) | (.17) | (.17) | |
| Male | .12 | .12 | — | — |
| (.13) | (.132) | |||
| Age | −.09** | −.09** | −.07 | −.08 |
| (.03) | (.03) | (.07) | (.07) | |
| Age2 | .00** | .00** | .00 | .00 |
| (.00) | (.00) | (.00) | (.00) | |
| Never married | .10 | .16 | .45 | .49 |
| (.18) | (.17) | (.36) | (.35) | |
| Education (ref = high school): | ||||
| < high school | .34 | .38 | −1.27+ | −1.30+ |
| (.43) | (.50) | (.65) | (.77) | |
| Some college | .32+ | .31+ | .20 | .20 |
| (.18) | (.18) | (.40) | (.39) | |
| College | .25 | .27 | −.04 | .02 |
| (.19) | (.19) | (.40) | (.39) | |
| > college | .34 | .40+ | .20 | .33 |
| (.25) | (.24) | (.49) | (.48) | |
| Constant | 4.72*** | 3.53*** | 9.43*** | 8.19*** |
| (1.22) | (1.22) | (2.38) | (2.35) | |
| N | 473 | 473 | 244 | 244 |
| R2 | 0.04 | 0.08 | 0.09 | 0.12 |
Note.—Nos. in parentheses are robust SEs.
P < .10, two-tailed tests.
P < .05.
P < .01.
P < .001.
Solidarity is a meaningful predictor of support for taxation and, as evidenced by comparing coefficients in model A and model B, serves to mediate the association between the race conditions and support for taxation. But do feelings of solidarity influence respondent preferences for a progressive tax structure?
Table 5, model C, reproduces the previous analysis, which found a strong effect of race condition on preferences for progressive taxation among male respondents: support for progressive taxation was lower among males who were told the new arrivals to the county were black or Latino. Model D adds the measure of solidarity to the model. Results demonstrate that solidarity is a significant driver of support for a progressive, rather than flat, tax structure. Moreover, after accounting for solidarity, the differences in preferences for a progressive tax among male respondents in the black and Latino conditions relative to the white condition shrink toward zero and are no longer statistically significant. Changes in the coefficients for the black and Latino conditions between models C and D suggest that this measure of solidarity accounts for about 28% and 36%, respectively, of the association between being in the black condition and Latino condition on reduced support for progressive taxation (Sobel mediation test significant at P < .05).
These findings provide support for hypothesis 5—observed differences in support for increased taxes and preferences for progressive tax systems by race condition can be explained, in part, by differences in the degree to which white respondents feel solidarity with the new arrivals to their hypothetical county. This provides some evidence for the individual-level process through which changing racial composition may lead to a real change in tax structures (Hopkins 2009).
EXTENSIONS AND COMPETING HYPOTHESES
Ethnoracial Divisions or Nativity and Citizenship Status?
The regression analysis of the change in the progressivity of state and local tax systems found that as the percentage of Latino residents in a state increased, the tax system became more regressive. Notably, models yield no evidence for an association between %black or %Asian and the progressivity of tax systems, likely in part because few states saw an appreciable change in the proportion of residents who were black or Asian between 1995 and 2007. The experimental results demonstrated a statistically significant difference among white respondents in their feelings of solidarity toward blacks as well as Latinos, with respondents feeling less similar to Latinos than to blacks. One reason whites may feel less solidarity with Latino arrivals than with black arrivals involves assumptions about the immigration and citizenship status of the new arrivals—Latinos may be more different because they were born in another country and may not be citizens or even legal residents of the United States.
It could be argued, therefore, that any observed empirical association between ethnic composition and the structure of tax systems is confounded by assumptions about the immigration or legal status of new residents. In other words, it may not be the ethnicity of new arrivals that is driving preferences for regressive tax systems but instead their immigrant status. In addition to including a control for %foreign born in the regression models, I further tested for this potential confounding explanation directly in the experiment by asking respondents both how likely it is that the new arrivals were foreign born and how likely it is that the new arrivals were citizens or legal residents. Although respondents in the Latino condition were more likely to report the new arrivals were foreign born and noncitizens than those in the white or black conditions, this measure was not associated with respondents’ tax preferences and did not serve to mediate the association between the increase in minorities and tax preferences (see table A3 in the appendix).
These analytic techniques provide some evidence that these results are not driven by a pure “immigrant” effect. At the same time, attitudes toward Latinos in the contemporary United States are certainly influenced by notional assumptions about immigrants and noncitizens. Indeed, the content of the racialized category of Latinos in the United States has evolved and continues to be shaped within discourses of immigration, foreignness, and citizenship (Padilla 1984; Martinez 2008). Nevertheless, I posit that these findings suggest a more complex mechanism for how a changing Latino presence may be influencing tax preferences and tax policies than simply a generalized response to immigrants.
Race and Assumptions of Public Spending
Taxation and spending are inherently related; therefore, we might expect individual preferences for how governments should tax to be strongly influenced by preferences for how government should spend. One could therefore argue that the relationship between race and tax progressivity documented in this investigation is merely an artifact of the already established link between race and social spending, that is, individual support for increased taxes or progressive tax structures is driven by race-based assumptions of how that money will be spent. In other words, it may be that racial composition shapes tax preferences by shifting assumptions about what the money will be spent on. Individuals may prefer to use progressive tax instruments to fund education and regressive tax instruments for highways, for example. Given that minority status is often conflated with poverty and further associated with welfare dependency and criminality (Gilens 1999), it could therefore be argued that the observed association is being driven by attitudes toward certain types of spending that are typically associated with minority populations.
I attempted to account for this potential feedback loop between race, assumptions about government spending, and preferences for taxation in a variety of ways. First, the experimental study was designed to prevent the conflation of minority status with welfare dependence—and the conflation of race and class more generally—by explicitly noting that the new arrivals were generally low-income irrespective of race. Moreover, in follow-up questions, respondents were asked how likely it was that the county needed additional revenue in order to increase spending in each of the following areas: health and hospitals, welfare and income assistance, parks and recreation, police and public safety, and education. Notably, I found no evidence of a difference in expectations for how the tax revenue would be spent across the three race conditions; respondents in the black and Latino condition were no more likely to report that they believed the county needed revenue to increase spending on welfare, policy, parks, education, or health than respondents in the white condition (results available on request). This null finding suggests that attitudes toward redistributive taxation are being shaped directly by attitudes toward race, not by differential expectations about what the race of the new arrivals may mean for how the new tax dollars are spent.
Finding no difference in the expected use of tax revenue by race provides additional support for the argument that the association is being driven by a more basic notion of solidarity with the new arrivals than by assumptions about their deservingness for redistribution, at least in terms of criminality and welfare dependence. Although it may be a stretch to conclude that respondents believed blacks and Latinos were no more likely than whites to rely on welfare or be involved in criminal activity, the results are suggestive and raise a number of interesting questions for future research.
DISCUSSION AND CONCLUSION
Understanding the social determinants of taxation is essential to understanding the social determinants of redistribution. This study therefore offers three contributions to the existing literature on race and redistribution. First, it demonstrates that increasing racial heterogeneity, specifically the increasing proportion of Latinos in a state, is associated with more regressive tax systems, including higher taxes on the poor. Second, in so doing, this analysis demonstrates that racial change, and not simply the level of ethnoracial diversity present, is a key driver of tax policy and preferences. Third, evidence from the survey experiment highlights one mechanism through which changing racial composition may influence preferences for taxation: lower feelings of solidarity among whites toward new arrivals of color. This mechanism, motivated by Tilly’s notion of opportunity hoarding, may prove to be more useful than conflict-based theories of race prejudice in future explorations of the relationship between race and redistribution.
Although the state-level analyses presented above find %Latino—and not %black—to be associated with tax progressivity, it is important to underscore that this discrepancy may be in part a function of the time period analyzed, 1995–2007. Indeed, as described above, racialized conflict between whites and blacks has played a central role in the evolution of subnational tax structures in the United States since the colonial era. For this reason, it is worth emphasizing that much of the “effect” of blacks on tax structure has already been “locked in” to state and local tax systems by 1995, the first year of data used in the current analysis. Moreover, given that there is considerably less within-state variation in %black compared to % Latino during this time period, it is no surprise the models yield no significant association between %black and tax progressivity in these data, despite historical evidence detailing how racialized conflict centered on African-Americans has influenced subnational tax systems.
Although the conclusions of this study are broadly consistent with the existing literature, the particular impact of increasing Latino presence on tax preferences and tax outcomes and the gendered differences in sensitivity to racial change warrant further exploration in studies of social spending. Moreover, given the limitations inherent to using state as the unit of analyses, future work should explore the link between racial change and the distribution of tax burden at lower levels of government, for example, counties or municipalities. This would permit more detailed exploration of how the changing composition of blacks, Asians, and subsegments of the Latino population (e.g., Mexican immigrants) shapes tax policy. Local-level analyses may also provide more analytic leverage for parsing how racial change is shaping public discourse as well as the political and policymaking processes that lead to reduced social spending and more regressive tax systems. Studies focusing on municipalities, specifically, would also permit analysis of non-tax sources of government revenue, including fines and fees, which are absent from the current investigation.
The relationship between taxation and spending is complex and contingent on a wide range of historical, social, political, and economic considerations. Regressive tax policies do not always correlate with low levels of redistributive social spending just as high spending does not always correlate with progressive tax structures. Indeed, across U.S. states we see examples of high-spending states with generally regressive tax systems (e.g., Massachusetts) as well as states with relatively progressive tax systems that spend relatively less on social investments (e.g., South Carolina). Cross-nationally, Prasad and Deng (2009) find that European countries with high levels of redistributive social spending—and often characterized by high levels of social solidarity—typically rely on regressive tax instruments such as consumption taxes to fund generous social programs. Conversely the United States has a comparatively progressive tax system but less generous social spending. This underscores the need to consider both taxation and spending in analyses of redistributive social spending as well as the potential for new insights to emerge in the study of welfare states and political economy from grappling with the intersection of these two policy areas. Under what social, political, economic, and demographic conditions is social spending funded through progressive versus regressive tax structures? Does the distribution of the tax burden influence public support for redistributive policies and programs or is it the reverse?
The structure of a tax system has important consequences for inequality within and between groups. Given recent evidence of the independent effect of tax systems on everything from health (e.g., Strully et al. 2010) and educational attainment (Newman and O’Brien 2011) to intergenerational economic mobility (Chetty et al 2014; Chetty and Hendren 2015), social scientists analyzing the role of policy in structuring life chances must consider the role of taxation. Net of social spending, taxes can serve to ameliorate or exacerbate existing inequalities by reducing the real income of some households through taxation while increasing the income of others through tax transfers. Future work should strive to account for both taxes and social spending when estimating the net effect of redistributive policy on everything from measures of individual and household well-being to aggregate levels of inequality. In tandem with efforts to better integrate analyses of taxation into studies of redistribution, more work must be done to quantify and contextualize the structure of tax systems in contemporary societies. The study of taxation—from empirically modeling tax structures and detailing their historical determinants to isolating their causal effects on state capacity, inequality between groups, and individual life chances—has the potential to generate new understandings across subfields on the causes and consequences of welfare state policy.
APPENDIX
Vignette from Survey Experiment
Imagine you are a longtime resident of Chestnut County. Chestnut County is a midsized county on the outskirts of a large American city. The county has experienced rapid population growth in recent years, and most of the people moving to the area are [white/black/Latino] and generally low income. Recent changes in the county have placed increased demands on the county’s finances. Therefore, the county council has decided it needs to raise additional revenue by increasing taxes. Next are a series of questions designed to better understand how you, a longtime resident of Chestnut County, would prefer to see the county raise taxes. [end vignette]
TABLE A1.
Suits Index and Racial Composition by State, 1995–2007
| State | Suits Index
|
Racial Composition 2007
|
Δ 1995–2007 (% points)
|
|||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 1995 | 2002 | 2007 | Δ95–07 | Black | Latino | Asian | Black | Latino | Asian | |
| AK | −.08 | −.04 | −.11 | −.03 | 3.11 | 3.69 | 4.59 | −1.91 | 1.46 | 2.04 |
| AL | −.10 | −.12 | −.13 | −.03 | 26.38 | 3.46 | 1.13 | 1.15 | 2.54 | .82 |
| AR | −.06 | −.06 | −.10 | −.03 | 16.05 | 5.82 | .92 | .28 | 4.96 | .45 |
| AZ | −.07 | −.10 | −.13 | −.06 | 3.51 | 32.32 | 1.97 | .25 | 9.13 | 1.19 |
| CA | −.02 | −.05 | −.02 | .00 | 6.13 | 36.11 | 11.87 | −.22 | 4.41 | 3.19 |
| CO | −.07 | −.10 | −.11 | −.04 | 3.77 | 19.39 | 2.13 | −.13 | 8.74 | .42 |
| CT | −.10 | −.14 | −.12 | −.02 | 9.26 | 11.48 | 3.21 | 2.20 | 4.43 | 2.31 |
| DE | −.03 | −.01 | −.04 | −.01 | 19.82 | 7.17 | 3.87 | 3.56 | 4.14 | 2.98 |
| FL | −.16 | −.18 | −.27 | −.11 | 14.95 | 21.77 | 1.73 | −1.20 | 5.93 | .59 |
| GA | −.06 | −.10 | −.09 | −.03 | 29.44 | 8.03 | 2.96 | −4.97 | 5.97 | 2.25 |
| HI | −.05 | −.07 | −.09 | −.03 | 2.01 | 7.64 | 41.66 | −.64 | 5.72 | −21.92 |
| IA | −.06 | −.07 | −.06 | .00 | 2.41 | 4.81 | 2.32 | −.13 | 3.94 | 1.93 |
| ID | −.03 | −.05 | −.05 | −.01 | .36 | 9.08 | .95 | −.56 | 1.92 | .54 |
| IL | −.11 | −.12 | −.15 | −.04 | 14.80 | 13.12 | 5.49 | −.19 | 4.42 | 3.85 |
| IN | −.08 | −.09 | −.09 | −.01 | 8.53 | 5.13 | .71 | 1.05 | 2.94 | .55 |
| KS | −.06 | −.07 | −.06 | .00 | 5.47 | 7.73 | 2.65 | −5.01 | 5.04 | .85 |
| KY | −.06 | −.07 | −.07 | −.01 | 7.17 | 2.10 | .98 | 1.04 | 1.46 | .49 |
| LA | −.09 | −.09 | −.09 | .00 | 3.95 | 2.99 | 1.10 | 1.45 | 1.48 | −.03 |
| MA | −.06 | −.09 | −.11 | −.05 | 6.44 | 6.71 | 5.83 | 1.07 | 3.06 | 4.29 |
| MD | −.07 | −.07 | −.07 | .01 | 28.55 | 7.48 | 4.92 | .42 | 4.09 | 2.39 |
| ME | −.03 | −.03 | −.04 | −.01 | .76 | .59 | .48 | .76 | .26 | −.16 |
| MI | −.09 | −.10 | −.07 | .01 | 13.83 | 3.66 | 1.93 | −.93 | 2.19 | .88 |
| MN | −.04 | −.06 | −.06 | −.02 | 4.22 | 4.67 | 3.86 | .57 | 3.20 | 3.23 |
| MO | −.07 | −.07 | −.08 | −.01 | 11.35 | 2.61 | 1.52 | 1.55 | 1.13 | .78 |
| MS | −.07 | −.08 | −.09 | −.02 | 37.11 | 2.11 | .79 | .25 | 1.19 | .31 |
| MT | −.02 | −.03 | −.04 | −.02 | .40 | 2.32 | .76 | −.08 | 1.63 | .30 |
| NC | −.05 | −.07 | −.05 | .00 | 21.08 | 6.95 | 1.91 | .36 | 5.47 | .88 |
| ND | −.06 | −.06 | −.08 | −.03 | .74 | 2.29 | .82 | .53 | 1.61 | −.12 |
| NE | −.07 | −.05 | −.07 | −.01 | 4.26 | 8.46 | 1.74 | 1.61 | 6.61 | 1.04 |
| NH | −.09 | −.13 | −.15 | −.06 | .92 | 2.94 | 1.70 | .66 | 1.78 | 1.14 |
| NJ | −.07 | −.07 | −.03 | .04 | 13.23 | 16.11 | 9.00 | .39 | 4.83 | 6.03 |
| NM | −.07 | −.07 | −.11 | −.04 | 2.10 | 39.90 | 1.41 | 1.05 | −.37 | .67 |
| NV | −.16 | −.18 | −.24 | −.08 | 7.31 | 23.21 | 6.19 | −1.36 | 8.20 | 3.37 |
| NY | −.07 | −.10 | −.07 | −.01 | 14.53 | 17.15 | 8.28 | −.81 | 4.71 | 5.09 |
| OH | −.06 | −.05 | −.07 | −.02 | 11.54 | 3.11 | 1.08 | .25 | 2.17 | .49 |
| OK | −.07 | −.07 | −.10 | −.03 | 7.60 | 5.26 | .61 | .86 | 2.23 | −.19 |
| OR | −.04 | −.04 | −.03 | .00 | 1.66 | 8.41 | 3.41 | −.59 | 3.19 | 1.51 |
| PA | −.09 | −.12 | −.12 | −.02 | 9.87 | 3.95 | 1.63 | .90 | 1.90 | .66 |
| RI | −.05 | −.07 | −.07 | −.02 | 5.32 | 1.43 | 3.04 | 2.17 | 6.10 | 1.86 |
| SC | −.03 | −.05 | −.04 | −.01 | 28.94 | 3.64 | .80 | −.73 | 3.00 | .58 |
| SD | −.12 | −.15 | −.18 | −.06 | .46 | 2.78 | .96 | −.16 | 2.00 | .34 |
| TN | −.12 | −.15 | −.17 | −.05 | 16.79 | 3.93 | 1.39 | −3.99 | 3.16 | .95 |
| TX | −.11 | −.15 | −.18 | −.06 | 11.26 | 36.31 | 3.23 | −2.54 | 5.83 | 1.29 |
| UT | −.07 | −.09 | −.09 | −.02 | .86 | 12.53 | 2.09 | .12 | 6.90 | 1.11 |
| VA | −.06 | −.07 | −.07 | .00 | 18.83 | 6.90 | 5.49 | −.96 | 4.07 | 3.70 |
| VT | −.03 | −.04 | −.03 | .01 | .61 | .62 | 1.85 | .18 | −.62 | 1.32 |
| WA | −.13 | −.21 | −.21 | −.08 | 3.21 | 8.36 | 6.69 | .94 | 6.48 | 2.47 |
| WI | −.07 | −.04 | −.06 | .01 | 5.22 | 4.96 | 1.73 | −1.88 | 3.26 | .84 |
| WV | −.04 | −.04 | −.04 | .00 | 3.05 | .43 | .49 | .74 | −.05 | .20 |
| WY | −.11 | −.17 | −.25 | −.14 | .85 | 7.53 | .63 | .07 | 2.40 | .41 |
Source.—Author’s calculation based on ITEP data and the U.S. Census Current Population Survey.
TABLE A2.
Multinomial Logistic Regression Predicting Assignment to Race Condition in Survey Experiment
| Black Condition | Latino Condition | |
|---|---|---|
| Household income (log) | .12 | −.01 |
| (.16) | (.16) | |
| Male | −.25 | −.20 |
| (.25) | (.25) | |
| Age | .03 | .00 |
| (.06) | (.06) | |
| Age2 | .00 | .00 |
| (.00) | (.00) | |
| Never married | .54 | .38 |
| (.33) | (.33) | |
| Education (ref = high school): | ||
| < high school | −.23 | −.12 |
| (.97) | (.97) | |
| Some college | .34 | .52 |
| (.32) | (.33) | |
| College | .20 | .40 |
| (.34) | (.35) | |
| > college | −.08 | .20 |
| (.44) | (.44) | |
| Constant | −2.27 | −.58 |
| (2.16) | (2.14) | |
| Pseudo R2 | .01 | |
Note.—N = 473.
P < .10.
P < .05.
P < .01.
P < .001.
TABLE A3.
Immigrant and Legal Status as Mediators of Preferences for Taxation across Race Conditions (OLS Regression)
| Support for Taxes (Full Sample)
|
Preference for Progressive Tax (Men Only)
|
|||
|---|---|---|---|---|
| Model E | Model F | Model G | Model H | |
| Likely foreign born | — | .09 | — | .05 |
| (.10) | (.17) | |||
| Likely “illegal” | — | −.00 | — | −.06 |
| (.10) | (.16) | |||
| Experimental condition (ref = white): | ||||
| Black condition | −.23 | −.21 | −.57+ | −.56+ |
| (.17) | (.16) | (.32) | (.32) | |
| Latino condition | −.37* | −.49* | −.78* | −.78* |
| (.17) | (.19) | (.32) | (.37) | |
| Constant | 4.72*** | 4.51*** | 9.43*** | 9.43*** |
| (1.22) | (1.23) | (2.38) | (2.42) | |
| N | 473 | 473 | 244 | 244 |
| R2 | .04 | .05 | .09 | .09 |
Notes.—Nos. in parentheses are robust SEs.
P < .10, two-tailed test.
P < .05.
P < .01.
P < .001.
Footnotes
I would like to thank Sara McLanahan, Viviana Zelizer, Katherine Newman, Scott Lynch, David Pedulla, Sarah Brayne, Trevor Bača, and Jeremy Levine as well as the AJS reviewers and seminar participants at Princeton University, Harvard University, the Massachusetts Institute of Technology, the University of Wisconsin–Madison, the University of Toronto, the University of Texas–Austin, and the University of California–Davis for their thoughtful comments and suggestions.
It should be noted that aggregating data to the state level—which is necessary in a national analysis given variations in the structure of substate tax regimes and basic data limitations—masks potentially important variations across localities and within localities over time. Indeed, racial change may be more likely to influence policy at a local level than at a state level in the short run.
For more information, see http://www.itep.org/about/itep_tax_model_full.php.
Separate analysis used state gross domestic product instead of total income. Results are unaffected by choice of measure.
From Harmon: “State House and Senate variables were constructed by centering the percent Republican around 50% so that Republican control represents positive deviations from 50%, while Democratic control represents negative deviations. The absolute values of the deviations were then logged with the negative sign returned to the Democrats to create a logarithmic scale with positive and negative deviation from zero to represent the diminishing returns of political party concentrations” (2011, p. 103, n.20).
The survey was administered to participants in an online panel recruited and maintained by Qualtrics, a research firm. The general adult population in the Qualtrics sample was invited to take the survey, and quotas were used to make the sample representative across income, age, and gender. Although this sampling strategy is inferior to a true national probabilistic design, the range of respondents recruited through this method is likely more readily generalizable than convenience samples typically used in experimental designs (e.g., university students).
Of the 793 white respondents, 320 failed the manipulation check asking them to recall the race/ethnicity of the new residents described in the prime. A logistic regression model predicting a correct answer to the manipulation check finds no differences in the odds of answering correctly by respondent sociodemographic characteristics (income, age, sex, or education). As these respondents did not receive the “treatment” being tested here, they add “noise” to the data and are excluded from the analyses that follow (see Oppenheimer, Mayvis, and Davidenko 2009). When respondents who failed the manipulation check are included in the sample (in their assigned conditions), coefficients for the Latino and black conditions in the model predicting support for taxes attenuate to zero and are nonsignificant (P > .9 and P > .4, respectively). The results for the main finding of preferences for flat vs. progressive taxation are more robust; inclusion of those who failed the manipulation check reduces the coefficients on being in the Latino or black conditions by half (P = .22 and P =.13, respectively). Coefficients on the Latino and black conditions in the model predicting solidarity also shrink but remain statistically significant (P < .01 and P < .01, respectively).
References
- Abascal Maria. Us and Them: Black-White Relations in the Wake of Hispanic Population Growth. American Sociological Review. 2015;80(4):789–813. [Google Scholar]
- Alesina Alberto, Glaeser Edward. Fighting Poverty in the U.S. and Europe. Oxford: Oxford University Press; 2004. [Google Scholar]
- Alesina Alberto, Baqir Reza, Easterly William. Public Goods and Ethnic Divisions. Quarterly Journal of Economics. 1999;114(4):1243–84. [Google Scholar]
- Alesina Alberto, Giuliano Paola. Working Paper No. 14825. National Bureau of Economic Research; 2009. Preferences for Redistribution. [Google Scholar]
- Alesina Alberto, Glaeser Edward, Sacerdote Bruce. Why Doesn’t the United States Have a European-Style Welfare State? Brookings Paper on Economics Activity. 2001;2:187–278. [Google Scholar]
- Biles Roger. The South and the New Deal. Lexington: University of Kentucky Press; 1994. [Google Scholar]
- Blalock Hubert M. Toward a Theory of Minority Group Relations. New York: Wiley; 1967. [Google Scholar]
- Blumer Herbert. Race Prejudice as a Sense of Group Position. Pacific Sociological Review. 1958;1(1):3–7. [Google Scholar]
- Bobo Lawrence D. Prejudice as Group Position: Microfoundations of a Sociological Approach to Racism and Race Relations. Journal of Social Issues. 1999;55(3):445–72. [Google Scholar]
- Bobo Lawrence, Hutchings Vincent L. Perceptions of Racial Group Competition: Extending Blumer’s Theory of Group Position to a Multiracial Social Context. American Sociological Review. 1996;61(6):951–72. [Google Scholar]
- Bonilla-Silva Eduardo. Racism without Racists: Color-Blind Racism and the Persistence of Racial Inequality in the United States. Lanham, Md: Rowman and Littlefield Publishers; 2006. [Google Scholar]
- Brooks Clem, Manza Jeff. Why Welfare States Persist: Public Opinion and the Future of Social Provision. Chicago: University of Chicago Press; 2007. [Google Scholar]
- Chetty Raj, Hendren Nathaniel. Working Paper. Harvard University; 2015. The Impacts of Neighborhoods on Intergenerational Mobility: Childhood Exposure Effects and County-Level Estimates. [Google Scholar]
- Chetty Raj, Hendren Nathaniel, Kline Patrick, Saez Emmanuel. Where Is the Land of Opportunity? The Geography of Intergenerational Mobility in the United States. Quarterly Journal of Economics. 2014;129(4):1553–1623. [Google Scholar]
- Cuddy Amy J, Fiske Susan T, Glick Peter. The BIAS Map: Behaviors from Intergroup Affect and Stereotypes. Journal of Personality and Social Psychology. 2007;92(4):631–48. doi: 10.1037/0022-3514.92.4.631. [DOI] [PubMed] [Google Scholar]
- Cutler David M, Elmendorf Douglas W, Zeckhauser Richard J. Working Paper No. w4283. National Bureau of Economic Research; 1993. Demographic Characteristics and the Public Bundle. [DOI] [PubMed] [Google Scholar]
- DiTomaso Nancy. The American Non-Dilemma: Racial Inequality without Racism. New York: Russell Sage Foundation; 2013. [Google Scholar]
- Einhorn Robin L. Species of Property: The American Property-Tax Uniformity Clauses Reconsidered. Journal of Economic History. 2001;61(4):974–1008. [Google Scholar]
- Einhorn Robin L. American Taxation, American Slavery. Chicago: University of Chicago Press; 2006. [Google Scholar]
- Enos Ryan D. Causal Effect of Intergroup Contact on Exclusionary Attitudes. Proceedings of the National Academic of Sciences. 2014;111(10):3699–3704. doi: 10.1073/pnas.1317670111. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Evans William N, Garthwaite Craig L. Giving Mom a Break: The Impact of Higher EITC Payments on Maternal Health. American Economic Journal: Economic Policy. 2014;6(2):258–90. [Google Scholar]
- Fellowes Matthew C, Rowe Gretchen. Politics and the New American Welfare States. American Journal of Political Science. 2004;48(2):362–73. [Google Scholar]
- Fiske Susan T, Cuddy Amy JC, Glick Peter, Xu Jun. A Model of (Often Mixed) Stereotype Content: Competence and Warmth Respectively Follow from Perceived Status and Competition. Journal of Personality and Social Psychology. 2002;82(6):878–902. [PubMed] [Google Scholar]
- Foner Eric. Forever Free: The Story of Emancipation and Reconstruction. New York: Knopf; 2005. [Google Scholar]
- Foner Eric. Reconstruction: America’s Unfinished Revolution, 1863–1877. New York: Harper Collins; 2011. [Google Scholar]
- Fong Christina. Social Preferences, Self-Interest, and the Demand for Redistribution. Journal of Public Economics. 2001;82(2):225–46. [Google Scholar]
- Fox Cybelle. Three Worlds of Relief: Race, Immigration, and the American Welfare State from the Progressive Era to the New Deal. Princeton, N.J: Princeton University Press; 2012. [Google Scholar]
- Garfinkel Irwin, Rainwater Lee, Smeeding Timothy. Wealth and Welfare States: Is America a Laggard or Leader? New York: Oxford University Press; 2010. [Google Scholar]
- Gerber Alan S, Green Donald P. Field Experiments: Design, Analysis, and Interpretation. New York: W. W. Norton; 2012. [Google Scholar]
- Gilens Martin. Why Americans Hate Welfare. Chicago: University of Chicago Press; 1999. [Google Scholar]
- Gilens Martin. Inequality and Democratic Responsiveness. Public Opinion Quarterly. 2005;69(5):778–896. [Google Scholar]
- Goldfield Michael. The Color of Politics: Race and the Mainsprings of American Politics. New York: New Press; 1997. [Google Scholar]
- Harmon Mark G. The Imprisonment Race: Unintended Consequences of ‘Fixed’ Sentencing on People of Color over Time. Journal of Ethnicity in Criminal Justice. 2011;9(2):79–109. [Google Scholar]
- Hopkins Daniel J. The Diversity Discount: When Increasing Ethnic and Racial Diversity Prevents Tax Increases. Journal of Politics. 2009;71(1):160–77. [Google Scholar]
- Howard Christopher. The Hidden Welfare State. Princeton, N.J: Princeton University Press; 1999. [Google Scholar]
- Hoynes Hilary, Miller Doug, Simon David. Income, the Earned Income Tax Credit, and Infant Health. American Economic Journal: Economic Policy. 2015;7(1):172–211. [Google Scholar]
- Jacobs David, Helms Ronald. Racial Politics and Redistribution: Isolating the Contingent Influence of Civil Rights, Riots, and Crime on Tax Progressivity. Social Forces. 2001;80(1):91–121. [Google Scholar]
- Jacobs David, Waldman Don. Toward a Fiscal Sociology: Determinants of Tax Regressivity in the American States. Social Science Quarterly. 1983;63(3):550–65. [Google Scholar]
- Knight Brian. Supermajority Voting Requirements for Tax Increases: Evidence from the States. Journal of Public Economics. 2000;76:41–67. [Google Scholar]
- Kousser J Morgan. Progressivism—for Middle-Class Whites Only: North Carolina Education, 1880–1910. Journal of Southern Economic History. 1980;46(2):169–95. [Google Scholar]
- Lieberman Evan S. Race and Regionalism in the Politics of Taxation in Brazil and South Africa. Cambridge: Cambridge University Press; 2003. [Google Scholar]
- Lieberman RobertC. Shifting the Color Line: Race and the American Welfare State. Cambridge, Mass: Harvard University Press; 2001. [Google Scholar]
- Luttmer Erzo FP. Group Loyalty and the Taste for Redistribution. Journal of Political Economy. 2001;109(3):500–528. [Google Scholar]
- Lynch Scott M. How Many Lags of X? In: Newman Katherine S, O’Brien Rourke., editors. Taxing the Poor: Doing Damage to the Truly Disadvantaged. Berkeley: University of California Press; 2011. pp. 163–74. app. 1. [Google Scholar]
- Manza Jeff. Race and the Underdevelopment of the American Welfare State. Theory and Society. 2000;29(6):819–32. [Google Scholar]
- Martin Isaac W. The Permanent Tax Revolt: How the Property Tax Transformed American Politics. Stanford, Calif: Stanford University Press; 2008. [Google Scholar]
- Martin Isaac W, Mehrotra Ajay K, Prasad Monica., editors. The New Fiscal Sociology: Taxation in Comparative and Historical Perspective. Cambridge: Cambridge University Press; 2009. [Google Scholar]
- Martin Isaac W, Prasad Monica. Taxes and Fiscal Sociology. Annual Review of Sociology. 2014;40(1):331–45. [Google Scholar]
- Martinez Lisa M. ‘Flowers from the Same Soil’: Latino Solidarity in the Wake of the 2006 Immigrant Mobilizations. American Behavioral Scientist. 2008;52(4):557–79. [Google Scholar]
- Morgan Kimberly J, Prasad Monica. The Origins of Tax Systems: A French-American Comparison. American Journal of Sociology. 2009;114(5):1350–94. [Google Scholar]
- Mutz Diana C. Population-Based Survey Experiments. Princeton, N.J: Princeton University Press; 2011. [Google Scholar]
- Mullins Daniel R, Wallin Bruce A. Tax and Expenditure Limitations: Introduction and Overview. Public Budgeting and Finance. 2004;24(4):2–15. [Google Scholar]
- Newman Katherine S, O’Brien Rourke L. Taxing the Poor: Doing Damage to the Truly Disadvantaged. Berkeley: University of California Press; 2011. [Google Scholar]
- Oppenheimer Daniel M, Meyvis Tom, Davidenko Nicolas. Instructional Manipulation Checks: Detecting Satisficing to Increase Statistical Power. Journal of Experimental Social Psychology. 2009;45(4):867–72. [Google Scholar]
- Orr Larry L. Income Transfers as a Public Good: An Application to AFDC. American Economic Review. 1976;66(3):359–71. [Google Scholar]
- Padilla Felix M. On the Nature of Latino Ethnicity. Social Science Quarterly. 1984;65(2):651–64. [Google Scholar]
- Pearson Elizabeth. Saying Yes to Taxes: The Politics of Tax Reform Campaigns in Three Northwestern States, 1965–1973. American Journal of Sociology. 2014;119(5):1279–1323. doi: 10.1086/675386. [DOI] [PubMed] [Google Scholar]
- Pinker Steven. The Blank Slate: The Modern Denial of Human Nature. New York: Penguin Books; 2006. [Google Scholar]
- Poterba James M. Demographic Structure and the Political Economy of Public Education. Journal of Policy Analysis and Management. 1997;16(1):48–66. [Google Scholar]
- Prasad Monica, Deng Yingying. Taxation and the Worlds of Welfare. Socio Economic Review. 2009;7(3):431–57. [Google Scholar]
- Quadagno Jill. The Color of Welfare: How Racism Undermined the War on Poverty. New York: Oxford University Press; 1994. [Google Scholar]
- Ribar David, Wilhelm Mark O. Welfare Generosity: The Importance of Administrative Efficiency, Community Values and Genuine Benevolence. Applied Economics. 1996;28(8):1045–54. [Google Scholar]
- Rodgers Harrell R, Tedin Kent L. State TANF Spending: Predictors of State Tax Effort to Support Welfare Reform. Review of Policy Research. 2006;23(3):745–59. [Google Scholar]
- Schumpeter Joseph A. In: 1991 The Crisis of the Tax State. Schumpeter Joseph., editor. 1918. pp. 99–140. [Google Scholar]; Swedberg Richard., editor. The Economics and Sociology of Capitalism. Princeton, N.J: Princeton University Press; [Google Scholar]
- Soss Joe, Schram Sanford F, Vartanian Thomas, O’Brien Erin. Setting the Terms of Relief: Explaining State Policy Choices in the Devolution Revolution. American Journal of Political Science. 2001;45(2):378–95. [Google Scholar]
- Strully Kate W, Rehkopf David H, Xuan Ziming. Effects of Prenatal Poverty on Infant Health: State Earned Income Tax Credits and Birth Weight. American Sociological Review. 2010;75(4):534–62. doi: 10.1177/0003122410374086. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Suits DanielB. Measurement of Tax Progressivity. American Economic Review. 1977;67(4):747–52. [Google Scholar]
- Thornton J Mills., III . Fiscal Policy and the Failure of Radical Reconstruction in the Lower South. In: Morgan Kousser J, James M, editors. Region, Race and Reconstruction Essays in Honor of C. Vann Woodward. McPherson. New York: Oxford University Press; 1982. pp. 349–94. [Google Scholar]
- Tilly Charles. Durable Inequality. Berkeley: University of California Press; 1999. [Google Scholar]
- Waisanen Bert. Report. Washington D.C: National Conference of State Legislatures; 2008. State Tax and Expenditure Limits, 2008. [Google Scholar]
- Walker Rebecca. Flora-Ala Student Newspaper. 2007. Feb 22, After 26 Years, Knight-Sims vs. Alabama Case Settled at Last. [Google Scholar]
- Wright Gavin. Old South, New South: Revolutions in the Southern Economy since the Civil War. New York: Basic Books; 1986. [Google Scholar]
- Wright Gavin. The New Deal and the Modernization of the South. Federal History Journal. 2010;2:58–73. [Google Scholar]
- Woodward C Vann. Origins of the New South, 1877–1913. Baton Rouge: Louisiana State University Press; 1971. [Google Scholar]