Lyle Scruggs and Thomas J. Hayes...Scruggs and Hayes 37total public spending on welfare and are...

Politics & Society2017, Vol. 45(1) 35 –66

© 2017 SAGE PublicationsReprints and permissions:

sagepub.com/journalsPermissions.nav DOI: 10.1177/0032329216683165

journals.sagepub.com/home/pas

Article

The Influence of Inequality on Welfare Generosity: Evidence from the US States

Lyle Scruggs and Thomas J. HayesUniversity of Connecticut, Storrs

AbstractThis article examines the relationship between income concentration and policy outputs that determine the generosity of two major state-level safety net programs: unemployment insurance and cash social assistance. Using a difference in differences framework, it tests the degree to which the top 1 percent share is associated with benefit replacement rates for these programs during the period 1978–2010. The results suggest that higher state income inequality lowers those states’ welfare benefits significantly in ways consistent with a “plutocracy” hypothesis that has been suggested in work by scholars such as Bartels, Bonica, Gilens, and Page. The results are robust to controls for several alternative explanations for benefit generosity, including citizen ideology, party control of government, fiscal pressure on programs, state racial heterogeneity, and public opinion liberalism. The results thus support the notion that growing income concentration at the very top undermines social protection policies.

Keywordseconomic inequality, welfare generosity, plutocracy, top 1 percent, social assistance, unemployment insurance

Corresponding Author:Thomas J. Hayes, University of Connecticut, 365 Fairfield Way, U-1024, Storrs CT 06269-1024, USA. Email: [email protected]

683165 PASXXX10.1177/0032329216683165Politics & SocietyScruggs and Hayesresearch-article2017

mailto:[email protected]

https://us.sagepub.com/en-us/journals-permissions

https://doi.org/10.1177/0032329216683165

https://journals.sagepub.com/home/pas

36 Politics & Society 45(1)

The second half of the twentieth century has witnessed the greatest expansion of eco-nomic prosperity in Western history.1 Yet recently the benefits of this economic expan-sion have been uneven. Since the late 1970s, economic inequality has increased dramatically in the United States and other western democracies, becoming particu-larly more concentrated at the top.2 While scholarship has documented this in the United States for some time, scholars have more recently begun to examine the ways in which economic inequality might influence public policy.

One growing body of research has shown that the government tends to be much more responsive to the interests of very wealthy.3 Some suggest the average citizen has almost no meaningful influence over the political system, which is instead domi-nated by economic elites. An implication of these studies is that greater income concentration shapes policy outputs in ways that are more and more favorable to the preferences of the rich. How much and how is an outstanding question, as Gilens and Page point out in their recent paper: “[We] need to learn more about exactly which economic elites (the ‘merely affluent’? the top 1 percent? the top one-tenth of 1 percent?) have how much impact upon public policy, and to what ends they wield their influence.”4

Standard models of policymaking under majority rule predict that the median voter serves as a major check on inequality via redistributive policies.5 According to this type of model, as inequality increases, the average voter becomes more favorable to redistribution from the rich, pushing democratically responsive leaders to adopt more redistributive economic policies. We examine these contrasting predictions by exam-ining how rising inequality influences the social safety net in the United States.

Such an examination is important. Since the start of Roosevelt’s New Deal, and especially since Johnson’s Great Society, the social safety net in the United States has contributed significantly to poverty reduction.6 Social Security and Medicare, for example, have contributed considerably to reducing poverty rates among the elderly. In the US federal system, however, states play a large role in addressing poverty and social policy for the non-elderly. Three of the ways in which state policies play a big role in assisting the non-elderly who struggle economically are through unemploy-ment insurance, cash and food assistance (TANF and SNAP, respectively), and Medicaid. These programs have historically been important in reducing the financial hardship of the poor and middle class, making state policies central to overall efforts to address poverty during decades of increasing inequality.

State social policy also provides an opportunity to examine the degree to which the income concentration influences public policy. It can do so for at least three reasons. First, new state-level data permit us to track variations in income concentration across the states over a long period. This provides a large number of cases to examine the effect of growing inequality on the social safety net, much more than are available in national level time series. A second important reason to focus on the states as an important laboratory for study is that US federalism provides much policy autonomy to political units that have in common many features that cross-national research sug-gests play an important role in affecting social policy, and the responsiveness of social policy to rising income inequality.7 The states generally account for about a quarter of

Scruggs and Hayes 37

total public spending on welfare and are integral to spending on all means-tested and many social insurance programs.8 This has been true since the New Deal, and perhaps even more true following the passage of the Personal Responsibility and Work Opportunities Act in 1996.9

Gilens’s research provides a third reason to look at the states.10 His research shows that the policy views of the wealthy are overrepresented in Washington; but it is less clear what his results imply about how these findings are affected by increasing inequality. Is unequal representation sensitive to the amount of economic inequality as is often inferred, or is it largely unaffected by the growth of income concentration? Is this unequal representation mainly a feature of national level politics, and thus reme-died by decentralizing political power; or can we see it at work at the subnational policy level? His work also suggests that the rich are least overrepresented in matters of social policy. However, Gilens mainly investigated national program reform pro-posals from which even the quite well off benefit considerably (Social Security and Medicare), and not more redistributive policies. State unemployment insurance and social assistance policies provide some important insights on these questions.

To determine whether the political power of increasingly concentrated wealth pro-duces welfare state retrenchment and reinforces a “plutocratic” model of American government, we utilize a unique data set on income shares going to top earners and new measures of two key aspects of state social policies—unemployment and social welfare benefit replacement rates. Our results are overwhelmingly consistent with a “plutocracy hypothesis,” rather than the more middle-voter-centered characterization of political representation that underlies many democratic politics models. Our use of “middle” is intentional, since our results do not hinge on the redistributive position of the critical actor being exactly in the center. The fact is that even a quite “elitist” notion of a median voter (i.e., much better off than the median) is much closer to the bottom than to the top of the income distribution compared to a decade or two ago. We show that it is variation in the economic power of the rich, not merely the better off, that is important for policy retrenchment.

State Level Inequality and Welfare Policy

The effects of economic inequality on American democracy at the national level have drawn considerable attention in the last decade,11 but research increasingly examines its effect among the states.12 State level inequality is an interesting area of focus because citizens may have greater awareness of the effects of more localized inequal-ity than that at the national level.13 Although some studies have found that cutting social spending is correlated with increasingly inequality, those studies do not exam-ine the alternative association that we do here: that increasing income concentration produces programmatic welfare policy retrenchments.

Figure 1 shows that income concentrations at the very top of the income distribu-tion in 1980 and in 2010 in each of the fifty states. All states lie above a line of equal-ity; that is, inequality rose in all states in this thirty-year period. However, there is considerable variation across states in amount of change. Income concentrations in


Hawaii, Iowa, Delaware, or Oklahoma are not much higher, whereas those in Wyoming, Connecticut, and New York are.

For example, in 1976 in Connecticut the top 1 percent earned a bit less than 10 percent of total state income; by 2010, the top 1 percent earned over 31 percent. In neighboring Rhode Island, the respective shares were 9 percent and 16 percent. In California, the top 1 percent earned about 9 percent of state income in 1976 and about 32 percent in 2010. In neighboring Oregon, the top 1 percent share started at a similar level (8 percent), but increased much less over the ensuing period, to 15 percent. We see similar variations around the country, both across and within all regions.

It is in this context that we investigate the extent to which this rising state level income concentration influences state social welfare programs. Past research has examined whether unequal turnout among social classes influences social policy out-puts; we examine whether income concentration affects policies with disproportionate impact on those with modest or low incomes.15 Also, unlike previous research, we look not at aggregate distributional outcomes—which are only partially determined by changing social policies themselves—but at indicators that more closely reflect aspects of policies over which legislators and executives have more direct control.

Studies of the unequal political influence of the wealthy have long been a focus of scholarship.16 Less well researched are the social policy outputs across political

Figure 1. Top 1 Percent Share of Total State Income in 1980 and 2010.Source: Data come from Estelle Sommeiller and Mark Price, “The Increasingly Unequal States of America: Income Inequality by State, 1917 to 2011.”14


jurisdictions, and positive policy feedback—inequality begetting more inequality—as some “oligarchic theories” of American democracy suggest.

The underlying theory leading to expectations that social transfers will be higher as inequality increases under democratic representation is commonly attributed to Meltzer and Richard.17 They suggest that as relative incomes at the top grow faster than those in the middle, the median voter will become more supportive of redistribu-tion. Meltzer and Richard simply assume that the median voter is decisive. Existing oligarchic theories—which are more correctly called “plutocratic” insofar as they sug-gest rule specifically by or in the interests of the wealthy—do not contest the internal consistency of the median-voter logic. They instead question the assumption of broadly democratic selection and representation under conditions of very high eco-nomic inequality. In other words, rather than ask what happens in a representative democracy when income changes, they ask what happens to representative democracy when inequality increases. Median voter models essentially suggest that inequality, given majoritarian democracy, will tend to produce negative feedback (via voting and then transfers); the oligarchy/plutocracy argument questions whether majoritarian democracy really exists in certain contexts (like high inequality). It further suggests (though this seldom explicit) that high and growing inequality leads to positive feed-back: rising inequality produces policies that further exacerbate inequality, and deviate further from democratic representation.18

Comparative political economists often talk of the “Robin Hood paradox,” the idea that among rich, de jure democracies, more market inequality is associated with less, not more, redistribution. The paradox is sometimes taken as a refutation of the Metzler-Richard model. In fact, it is only paradoxical if one’s starting assumption is a median voter model and no systematic constraints on its operation. Relaxing the equation of democratic selection processes (such as elections with universal suffrage, unregulated press and assembly rights) and representation processes (equal representation or “fair” aggregation of interests in public policy) can also “explain” the paradox.19

Acemoglu et al.20 refer to certain violations of the conventional democratic linkage (i.e., that free and fair elections produce equal representation of individual interests in public policymaking) as “elite capture,” and also refer to a variety of ways in which institutional conditions subvert the assumed democratic linkage between choice of leaders and choice of policy. Another mechanism undermining redistribution is social affinity. Lupu and Pontusson21 argue that the social affinity of the middle class for the rich (and against the poor) is important to explain why the middle class does not sup-port redistribution from the rich. They show that middle earners are more likely to side with lower earners where the earnings gap between the middle and rich is high relative to the earnings gap between rich and poor. Iversen and Soskice22 reconcile the Robin Hood paradox via differences in electoral institutions that cause larger commitment problems between voters in the middle and those on the left.23

All of these arguments predict increasing democratic pressures for redistribution in the United States, given the changing structure of inequality over the last thirty years. Based on median voter, social affinity, or even with an electoral institution bias, we should see more economic redistribution over time nationally if democratic


representation holds. Meltzer and Richard mechanisms have been explained already. Lupu and Pontusson’s argument predicts that the middle class will be more likely to side with the poor as their incomes converge. Finally, Iversen and Soskice’s argument also implies that the probability of a redistributive alliance between the middle and lower income groups grow over time (given fixed electoral rules) as either middle and lower income groups converge, or middle and wealthy groups diverge. (Both are now decades-long trends.) On the other hand, a pronounced shift away from demo-cratic representation plausibly generates retrenchment in both middle class and pov-erty programs.

In the United States, the structure of relative inequality has, in fact, moved deci-sively during the last thirty or forty years. The “typical” household (around the median income) was previously closer to the top than to the bottom, but is now closer to the bottom than to the top. This is displayed clearly in Figure 2 using data from the Congressional Budget Office.24 Between the late 1970s and late 1980s, median house-hold income was around 3.4 times the average income in the bottom quintile. Median

Figure 2. “Social Distance” between Income Strata: Upper and Lower Relative Income Gap, 1979–2011.Note: The bars represent the difference in the ratio of median income and quintile 1 and quintile 5 and the median. Positive values suggest that there is a larger ratio for median/Q1 than for Q5/median. Dark bars are based on pretax income, light bars on post-tax income. Income estimates include value of employer-paid and government health care premiums.Source: Congressional Budget Office (CBO), “Distribution of Household Income and Federal Taxes, 2011”; https://www.cbo.gov/publication/49440.

https://www.cbo.gov/publication/49440


income fell considerably in the 1990s, and then again after 2004. In 2011, it was at the lowest point recorded. Meanwhile the average income in the top quintile was about 2.3 times the median through the early 1980s. By the late 2000s (i.e., even before the Great Recession), the average income in the top quintile was about 3.5 times the median. We see the same pattern in after-tax data, so this shift is not compensated by (real) shifts in federal tax burdens toward higher incomes.25 We see a similar, but sub-tler, trend if we compare the median to the average in the neighboring second and fourth quintiles. The ratio of the median to second has remained about the same, while it has become slightly more distant from the fourth quintile.

The CBO information extends only to 2011, but Saez’s research suggests that the gap between the very top and the middle probably worsened through at least 2014.26 He estimates that between 1993 and 2014, the top 1 percent reaped 58 percent of all income growth in the country. Moreover, the remaining 42 percent of income growth accrued almost entirely to the next highest 4 percent of earners. Growth in centiles 6–10 was stagnant, and the bottom 90 percent probably lost ground: median household income was flat; real average income fell 6 percent in that period. The upshot is that during the last four decades, middle earners have converged toward lower income groups, and away from higher incomes groups.

Plutocracy in America?

In this section we lay out an argument for how growing inequality at the very top, under conditions of plutocracy, can generate social policy retrenchment. We do not argue that it is logically impossible for a middle income voter to prefer retrenchment under these conditions, or that plutocracy implies that the non-rich have no political influence; but we do hope to show that, echoing Gilens and others,27 a marked decline in the representativeness of democracy might plausibly produce retrenchment, and is a more consistent and plausible story than one that has middle-income voters indepen-dently deciding to abandon redistribution because they are falling further and further behind.

First, not only have the wealthy long been more likely to participate in politics and exercise significant political power relative to their numbers,28 but they have become, increasingly likely to be active in important political activities in the United States. Bonica et al.29 show, for example, that donations by the top .01 percent of earners account for over 40 percent of all campaign contributions in 2012 (up from about 20 percent in the 1980s).30 Of course, not all of this is directed specifically at reducing welfare benefits, or directly furthering economic interests of the very rich; but even rich, liberal donors (who compose an increasingly important part of the Democratic Party constituency) are typically not noted for championing income transfers to the poor and middle class. Income concentration drives donations by rich Republicans more than by rich Democrats; and corresponds to a sizable, conservative shift in the voting records of Republican lawmakers since the mid-1970s that is not matched by cycles in public mood.31


Second, the ideological leanings of the rich tend to be more anti-welfare than the middle class and poor, even when all groups are skeptical of welfare programs.32 Bonica, Rosenthal, and Rothman33 found a very close correlation between income and partisanship among American physicians, a generally high earning group. Page, Bartels, and Seawright34 find the very wealthy have a much more conservative ideo-logical outlook on taxation, regulation, and, most especially, social welfare. Kuo and McCarty35 find that Democratic donors were more likely to support cuts in social spending (such as the Simpson Bowles plan) than non-donor Republicans. Gilens36 draws on a wide variety of policy issues from taxes to foreign policy and finds large opinion differences between income groups. Scholzman, Brady, and Verba37 also show that higher SES is associated with less support for issues like expanding access to health insurance. Studies in social psychology also suggest that increasing rela-tive economic well-being is associated with less prosocial behavior,38 increased nar-cissism and entitlement,39 and lower empathy.40 Thus, not simply pocketbook but also psychological motivations might make the wealthy more likely to support wel-fare retrenchment.

Nationally representative surveys support such findings, most specifically related to government spending on programs meant to provide benefits to those struggling eco-nomically. Figure 3 shows information from two national opinion polls that demonstrate that rich and poor generally disagree sharply on welfare policies. The opinion item on

Figure 3. Net Opposition to Social Safety Net Spending by Income Group.Source: GSS cumulative file 1972–2008 and ISPP 2006.


welfare is from the General Social Survey (GSS) cumulative file (1972–2008) while the items for unemployment benefit spending and government spending are from the 2006 International Social Survey Programme (ISSP) survey on the role of government. Each question was recoded to the dichotomous “support” or “oppose” (exact question word-ing can be found in Appendix A). The figure displays net opposition (oppose minus support) by income group for spending in each area so that positive values in the figure indicate majority opposition while negative values indicate majority support.41 The fig-ure shows increasing opposition to spending on each safety net program as individual incomes increase. While welfare and government spending in general may be unpopular, they are more unpopular as household income increases. Notably, there is net support for unemployment benefits for all but the highest income group.

One objection to the results in Figure 3 and many national surveys is that they do not really differentiate the wealthy from the upper middle class. A typical top household income category ($150,000) includes more than 10 percent of households today. The top 1 percent starts at over $400,000.42 Two surveys (in 2010 and 2014) conducted by the ABC/Washington Post provide a top income category of $250,000. Each contains a question about support for extending the duration unemployment benefits in the wake of the Great Recession.43 Figure 4 shows that opposition to benefit extension increased noticeably at very high incomes. Unsurprisingly, net opposition is higher among Republicans on this issue, but rises among high income Democrats in 2010 and 2014.

The final element of our argument is that policymakers (in America at least) over-weight the interests of higher economic strata. In other words, not only do interests and psychology of greater wealth appear to reduce support for social welfare pro-grams going to the middle and lower ends of the distribution, but the wealthy carry considerably more weight with elected politicians. This result is due, among other things, to dependence on individual financing to run for office. Gilens and Bartels both show convincingly that the rich are much more likely to see their preferences represented in policy decisions.45 Rigby and Wright also conclude that economic policy outcomes are more responsive to the preferences of the rich than the poor in the states.46 Gilens concludes that the representational biases are so stark that they “call into question the very democratic character of our society.”47 Barber finds that politician’s votes in the US Senate are much more closely correlated with the ideo-logical preferences of (rich) donors than with those of the voters.48 Gilens and Page clearly expresses a disconnect between America’s electoral and representative democracy:

Americans do enjoy many features central to democratic governance, such as regular elections, freedom of speech and association, and a widespread (if still contested) franchise. But we believe that if policymaking is dominated by powerful business organizations and a small number of affluent Americans, then America’s claims to being a democratic society are seriously threatened.49

To paraphrase Schattschneider, increasing inequality appears to be increasing the his-torically upper class accent of American politics.50


But how does this shift in influence result in cuts to social welfare and unemploy-ment insurance? Increasing income concentration mobilizes more wealthy individuals—the rich themselves—who have increasing relative power over the electoral fates of politicians and the ear of politicians in office. Holding total income fixed, an increase in income concentration shifts resources away from middle class or poor voters who are (empirically) much more likely to support strong unemployment or welfare protec-tion presumably out of both self-interest and “fellow feeling” for the plight of the poor and middle class, to individuals who tend to be less empathetic in general, have less economic interest in supporting such policies, and accrue to people (the poor and middle class) to whom they are increasingly socially disconnected. Given the existing evidence that relative poverty tends to reduce participation among the poor, rising inequality shifts the political agenda in the direction of less social protection, while reducing the participation of those most likely to vote for more social protection. After the election, growing inequality further favors the interests of those with concentrated income. The evidence suggests that this class skew exists for partisans of both major political parties, even though supporters of one party (Republicans) tend to be more anti-welfare independent of their income.51

Hypotheses

As the earlier reference to Schattschneider makes clear, the idea of class bias in representation in America has historic roots. But Schattschneider’s era was one of

Figure 4. Opposition to Unemployment Benefit Extension by Income Level and Political Party Identification, 2010 and 2014.Source: ABC News / Washington Post polls; Roper Center for Public Opinion Research.44


declining inequality. The contemporary question is: what are the implications of rising inequality? We hypothesize that it will tend to produce social welfare retrenchment: the percentage of income going to the top 1 percent leads states to reduce the generos-ity of welfare and unemployment benefits. We label this the plutocracy hypothesis. As inequality rises, the interests of the rich become (more) separated from those of the non-rich. But it is also the case that increasing inequality depresses participation by and representation of the poor and middle class.52 More than one study finds this spe-cifically for turnout and preference representation in US state elections.53 Depressed participation at the bottom implies that the median voter may actually get richer as social inequality rises, thus blocking the negative feedback effects assumed in the full turnout Metzler-Richard model. Some research implies that this scenario is more likely in the US case, because the party system has no viable political parties on the far left (or right), and that it may shift the terms of the policy competition to differences among those with higher incomes. If we add that electoral competition in the US is particularly dependent on fund raising from large, wealthy donors, more concentration of income at the very top makes candidates for office (from both parties) more beholden to economic preferences of the (increasingly) wealthy.

The alternative, the democracy hypothesis, predicts that states will increase wel-fare or unemployment insurance programs as the concentration of income increases at the top. The precise mechanism underlying the democracy hypothesis is not specified, but, as we discussed, several different theoretical mechanisms point toward expanding social protection in recent decades given functioning representa-tive institutions.

Data and Methods

Our data set encompasses annual information on individual US states from 1978 to 2010. All variables used in the analyses are described and summarized in Appendix B, Tables B1 and B2. Looking at variation within US states allows us to control for confounding factors that frequently confound estimates of the effects of inequality on redistribution and welfare state policy outputs in cross-national research. Inequality is high in the United States by the standards of most rich countries, but there is also considerable variation among the states. Unemployment and social assistance benefits also vary considerably, though not as much as they do internationally.

Dependent Variables

The main dependent variables are posttax benefit replacement rates for a notional individual.54 The replacement rate is the ratio of (after tax) benefit to (after tax) aver-age wage in each state. An advantage of using program replacement rates over spend-ing levels is that the latter track the joint outcome of both policy rules and demand. They are less subject to the control of policymakers (e.g., because macroeconomic shocks may increase or lower demand for benefit). Program replacement rates are widely thought to function as a superior gauge of policymakers’ intentions.55


Unemployment Insurance Replacement Rate (UIRR). We calculate this rate based on the benefit due to a qualified single individual earning the average wage net of federal and state taxes.56

UIRR net unemployment insurance benefit net averastate year= − % gge wagestate year −

State unemployment insurance (UI) provides weekly benefits to the qualified unem-ployed. Information used to compute the benefit comes primarily from tables pub-lished by the Department of Labor’s Employment and Training Administration, entitled “Significant Provisions of State Unemployment Insurance Laws.”57 We com-pute a weekly benefit for a single worker in each state earning the monthly average covered wage in the state during the two years preceding unemployment. Average wage in each state is taken from the Employment and Training Administration’s Unemployment Insurance Financial Data Handbook.58 State and federal taxes and employee FICA (Social Security and Medicare) contributions are computed on tax rules for each state (starting in 1977) from TaxSim 9.0.59 Benefits are usually payable for up twenty-six weeks, so we double the benefits to facilitate the tax computations. Net unemployment benefit replacement rates vary considerably, from a low of 28 per-cent of the state’s average wage (Alaska 1980 and Indiana 1991) to over 80 percent (Oregon mid 1980s). The overall distribution of replacement rates in the data set is single-peaked, and slightly left skewed. Nationally, the percentage of the labor force covered by UI is very high (over 90 percent of civilian employment). There is some variation in coverage across states, but these differences are relatively small, as cover-age conditions are largely determined at the national level. There is considerably more variation in the recipiency rate (beneficiaries/total unemployed). Recipiency varies less within states than across them over long periods.60

Social Assistance benefit replacement rate (SARR). We combine benefits from Temporary Aid for Needy Families (TANF), formerly Aid to Families with Dependent Children (AFDC), and benefit entitlements from the Supplemental Nutritional Assistance Pro-gram (SNAP), formerly Food Stamps, to compute a social assistance benefit for a family of three—for example, single mother and two children—with no other income. We combine AFDC/TANF and SNAP because recipients of AFDC/SNAP are almost always automatically qualified for SNAP. Furthermore, Moffitt, Ribar, and Wilhelm used AFDC/TANF plus SNAP since the former rates are set based on known informa-tion about national SNAP benefits.61

As with UIRR, the state’s average net-of-tax wage is the denominator. We assess taxes on benefits and wages as if the recipient household has one adult and two chil-dren, and receives only benefits (or wages) for the entire year.62

SARR SNAP TANF benefit net average

wage sin

state year %= +( ) −

ggle parent two childrenstate year

,( ) −


The benefits data used in these calculations come from published and unpublished information from the US Department of Health and Human Services, and the Urban Institute’s Welfare Rules Databook.

Covariates

Although cross-state measures of “plutocracy” are scarce, Winters and Page64 suggest that income concentration at the top is an reasonable indicator of both the preferences and power of plutocrats. Our measure of inequality is the percentage of income going to the Top 1 percent of earners.65 Values are estimated from income tax records following meth-ods similar to estimation from national tax records in Piketty and Saez.66 We use the share of income going to the top 1 percent because much of the income gain over recent decades has gone to this group as the pretax share of income (excluding capital gains) going to the top 1 percent of tax filers more than doubled from 1979 to 2011.67 This is the best measure available to capture the amount of income going to the top group and thus captures plu-tocracy more effectively than other measures such as the Gini coefficient.

Our other independent variables include factors that are commonly included in econometric models—both in the comparative politics and state politics literatures—on determination of welfare benefit generosity. To control for the influence of state economic conditions, we use (logged) real Income per Capita in each state. In order to control for fiscal pressures, which might motivate cuts in benefits independently of income concentration, we control for the reserve ratio of the state’s unemployment benefit trust fund68 or, in the case of social assistance benefits, the ratio of the state’s general fund revenue to spending.69

To capture differences in state electoral and institutional characteristics, we include Gubernatorial Election, a dummy variable (coded 1 in the year of a gubernatorial elec-tion, 0 otherwise). To address the possibility that inequality and welfare state cuts are jointly correlated with more antigovernment sentiment in state political culture, we also include a measure of citizen Ideology: the average state government and citizen ideology developed by Berry et al.70

Literature on US state politics often finds an important role for political parties in shaping outcomes (like the literature in comparative politics). Since partisanship might be correlated with both inequality and benefits—Republicans appear to be more likely support lower government benefits and are less concerned about inequality—we use a dummy variable to measure party control of state government: GOP Control (1 if the governor and a majority of both houses of the state legislature are held by the Republican Party, 0 if not). The data on the partisan composition of state legislatures were complied from State Partisan Composition.71

We also control for state racial characteristics (percentage of Nonwhite Population) because variations in state social policy and inequality have been correlated with racial animosity.72 Elderly population share (Population over 65) is included because retired Americans rely less on state unemployment and social assistance benefits. We also include state unionization rates to account for the possibility that inequality and social


policy are jointly correlated with labor’s organizational power in the state. Each of these measures is taken from Kelly and Witko,73 and they are based largely on Census data.

Finally, significant federal reforms to unemployment benefits and social assistance benefits occurred in 1987 and 1996, respectively. In 1987, unemployment benefits were made federally taxable, and taxable by most states with income taxes. In 1996 Congress passed the Welfare Reform Act.74 We expect that the shift to tax UI benefits will have a large negative impact on the change in replacement rates in 1987. Consequently, we include a dummy variable for that year. While we are less concerned that welfare reform would have the same level of impact on social assistance replace-ment rates—states had long been cutting welfare replacement rates, and the Welfare Reform Act initially prevented states from making large cuts to benefits—we never-theless include a dummy variable for 1996 in the social assistance models.

Statistical Methods

The estimation data set consists of information on fifty states measured annually over thirty-two years: 1978–2010. Our data structure is best characterized as cross-sectional time series (TSCS), rather than traditional panel data, because we have a moderate number of observations per unit.75 We use a first difference specification, and estimate the models with what is essentially a difference in difference approach. It is reasonable to expect changes in inequality to impact programs on a regular basis: benefit amounts and tax provisions to pay for them are frequently changed on an annual, or even semi annual basis, often to adjust (or not) for purchasing power or wage growth.76 Estimates of inequality effects on unemployment benefits using non-differenced data, for example, fixed effects on levels, suggest even larger and more precise negative impacts than are reported here.

The combined theoretical model is of the basic form:

∆ ∆Y = X +it it-1 tβ εj i∑

We use a one-year lag of the right-hand side variables in order to reduce the prospect of reciprocal effects of policy (benefit cuts) on top income shares such as those suggested in Kelly and Witko.77 (Lagged election year is a dummy variable = 1 if there was an elec-tion two years prior, because with only a few exceptions, new governments take power in the year after the election). Finally, in estimating the parameters of the models, we allow for clustering of errors within panels and for the possibility of different national trends in unemployment and welfare replacement rates with fixed state effects.78

Results

We begin with the basic relationship between welfare generosity and inequality, as measured by the percent of income going to the top 1 percent. Figure 5 shows the bivariate relationship between the level of income going to the top 1 percent and each of our measures of state welfare benefit generosity. The results show a clear negative correlation between inequality and benefit generosity. This negative correlation


squares precisely with the prediction of the oligarchy hypothesis, but runs counter to the democracy hypothesis.

Of course, we are ultimately interested in the relationship taking into account other factors that might also be causing changes in benefit generosity. Table 1 shows four sets of estimates. The first two have annual change in state unemployment insurance replacement rate as the dependent variable; the second two have change in the social assistance (TANF + SNAP) replacement rate as a dependent variable.

For unemployment, the estimates suggest that a one point increase in the top 1 per-cent income share in a state is associated with a .10 to .13 point decrease in the state’s UI benefit in the next year. The effect is larger (i.e., more negative) once we account for other factors that affect the benefit.79 The impact of rising income concentration on social assistance benefits also suggests that more inequality means more retrenchment in benefit rates for the neediest families. A one point increase in the income concentra-tion at the top is associated with a .08 point decrease in social assistance replace in the full models.80

Other results in Table 1 also support previous research on the causes of welfare state retrenchment. The introduction of UI benefit taxation was associated with a reduction in replacement rates of over six points. This suggests that the states did not compensate for the imposition of federal benefit taxation starting in 1987. Other results in Column 2 of Table 1 suggest that unemployment benefit cuts are more likely after elections, following increases in the nonwhite population, and following increases in the over sixty-five population share.

Other results in Column 4 of Table 1 suggest that faster income growth, and increases in the nonwhite population are associated with large social assistance replacement rate cuts. Moreover, despite provisions to prevent benefit cuts following the Welfare Reform Act’s implementation, there was a sizable decline in benefits on implementation: the first year of welfare reform (b =−.85, p < .001), log of income

Figure 5. Bivariate Relationship between Top Income Share and State Welfare Generosity.Source: Authors’ data.


growth (b = 2.6, p < .061), and the share of the nonwhite population (b = −.136, p < .003). All of these effects are in the expected direction; that is, welfare reform, income and a larger share of racial minorities results in larger cuts to social assistance, and faster income growth leads to increased benefits.

As checks on the robustness of the results (available on request), we also estimated the models with controls for (lagged difference of) the pretax Gini coefficient and the income share of the ninety to ninety-nine percentiles of the distribution. The resulting estimates for top 1 percent were largely identical to those reported here. This provides some reassurance that Top 1 is identifying a correlation between retrenchment and income concentration among the rich, and not simply picking up the general increase in inequality.81

Table 1. Fixed Effects Regression for Program Replacement Rates.

∆ Unemployment InsuranceReplacement Rate (UIRR)

∆ Social AssistanceBenefit Rate (SARR)

∆ Top 1 t-1 −.103** −.134** −.212*** −.0839***(.0466) (.0637) (.0263) (.0208)

Unemployment Benefits First Taxed (1987)

−6.250*** −6.232*** (.414) (.439)

∆ Ideology t-1 .0144 .00170 (.0141) (.00561)

∆ State GOP Control t-1 −.0259 −.170 (.196) (.102)

∆ Gubernatorial Election t-2 −.327** −.269*** (.144) (.0783)

∆ Per Capita Income (log) t-1 −.878 2.592* (2.711) (1.350)

∆ % Nonwhite t-1 .110* −.136*** (.0633) (.0432)

∆ Population over 65 t-1 −.928** .329 (.420) (.208)

∆ Unionization Rate t-1 −.00987 −.0466 (.0396) (.0305)

∆ Fiscal Pressure t-1 −.174 .0683 (.228) (.652)

Welfare Reform (1996) −.838*** −.855*** (.107) (.108)

Constant .0913*** .218** −.367*** −.481***(.0170) (.102) (.00617) (.0460)

Observations 1650 1470 1650 1470

Note: Standard errors in parentheses; *p < .10, **p < .05, ***p < .01.Source: Authors’ data; see Appendix Table B1.


Discussion and Conclusion

Our analysis suggests that the pattern of changing social protection benefits at the state level is responsive to changes in the concentration of income at the top of the income distribution in ways that contradict the general assumption of democratic representation. Over the last several decades, faster growth in income concentration has tended to lead to faster retrenchment, despite the fact that various “democratic” models predict increasing redistribution due to the increase in and changing struc-ture of inequality. An important reason for this is that wealthier individuals of both political parties are more opposed to social welfare programs. Although politicians have always relied more on donations from wealthier individuals, the increasing reliance on fundraising and donations from an increasingly lopsided income distri-bution drives politicians to retrenching policies of regressive redistribution. Our findings are consistent with a growing body of research that suggests that represen-tation in America may have veered quite far from the ideal of one-person, one-vote in recent years.

This study makes several unique contributions to this literature. First, we show that developments at the state level, for specific policy outputs over which state politicians exercise significant political control follow a pattern of overrepresentation of the inter-ests of the very wealthy. Second, unlike results estimated on cross-sectional models or pooled models that fail to adjust for country fixed effects, we show that larger changes in inequality at the top are negatively associated with subsequent changes in benefits, even after taking into account partisan, demographic, budgetary or macro-economic conditions that also affect such programs. Third, our paper focused on an important, if under explored, feature of state welfare policy: the income replacement rates for wel-fare assistance and for unemployment insurance. Unlike budgetary outcomes like spending levels or spending per person which typically are not directly controllable by policymakers, replacement rates are closer to policy “outputs” and under more direct political control from year to year.

Our research contributes to the literature on US state welfare policy by focusing on the effect that state level policies can have on individuals as well as the intentions of policymakers. Replacement rates have decreased over time and it is necessary to investigate the reasons, especially as incomes remain stagnant for many Americans. Our findings suggest that state-level inequality influences the declining replacement rates and that plutocratic tendencies of the American political economy may be a con-tributing factor.

Our analysis provides some initial answers to why and how economic elites may exercise their oversided influence on policy. But our research remains circumstantial. For example, while we can conclude that it is specifically income concentration at the very top, and not, say, overall or in the top quintile, that is linked to program cuts, we cannot say whether the connection is based on an even smaller concentration among the wealthy. We also lack information about specific instances of retrenchment that we can positively confirm are due to more concentrated income. For example, one


implication of our research is that if increasing concentration of income is important to policy influence, given increasing inequality, we should expect to see increasing evidence of misrepresentation.

The extent to which the wealthy have unequal influence over the political system is a key concern in evaluating the degree of de facto democracy in a given nation.82 Additional research should also consider ways to ameliorate political inequality so that it does not reinforce growing economic inequality of opportunity. State governments have become increasingly important for welfare and unemployment policies for a large number of struggling citizens, and it is important that policy reflects a more democratic balance.

Appendix A. Survey Question Wording for Figures 3 and 4.

Survey Question Wording

GSS cumulative file (1972–2006)

We are faced with many problems in this country, none of which can be solved easily or inexpensively. I’m going to name some of these problems, and for each one I’d like you to tell me whether you think we’re spending too much money on it, too little money, or about the right amount…are we spending too much, too little, or about the right amount on welfare? (Incomes are in 1986 dollars.)

ISSP 2006 Listed below are various areas of government spending. Please show whether you would like to see more or less government spending in each area. Remember that if you say “much more”, it might require a tax increase to pay for it. Unemployment Benefits.

ISSP 2006 Here are some things the government might do for the economy. Please show which actions you are in favor of and which you are against. Government Spending.

ABC News/Washington Post (2010)

Because of the economic downturn, Congress has extended the period in which people can receive unemployment benefits, and is considering doing so again. Supporters say this will help those who can’t find work. Opponents say this adds too much to the federal budget deficit. Do you think Congress should or should not approve another extension of unemployment benefits?

ABC News/Washington Post (2014)

Congress is considering whether or not again to extend the period In which people who’ve lost their jobs can get unemployment benefits. Supporters say this will help those who can’t find work. Opponents say this adds too much to the federal budget deficit. Do you think Congress should or should not approve another extension of unemployment benefits?


Appendix B. Variables and Summary Statistics

Table B1. Variables Used in the Analysis.

Variable Name Source Coding

Unemployment Insurance Net Replacement Rate

SWED (State Welfare Entitlement Data set)

After-tax benefit of percentage of income replaced by program .277 (min) to .763 (max)

Net Social Assistance/Net Minimum Wage Replacement Rate

SWED (State Welfare Entitlement Data set)

After-tax benefit of percentage of income replaced by program .191 (min) to .763 (max)

Top 1 Sommeiller and Price, “The Increasingly Unequal States of America”

Percent Income going to top 1 percent in each state

Ideology Berry et al., “Measuring Citizen and Government Ideology”

Average of state NOMINATE scores and citizen ideology measure

Republican Control of Government

Statistical Abstract of the United States, Green Papers

1 = Governor and Majority of Both Houses State Legislature Controlled by Republican Party (0 otherwise)

Gubernatorial Election Berry and Berry, “Tax Innovation in the States”)

1 = Year of Gubernatorial Election (0 otherwise)

Unionization Kelly and Witko, “Replication Data for: Federalism and American Inequality”

Proportion of workers who are members of unions in a state

PC Income (logged) Berry and Berry, “Tax Innovation in the States”)

Change in real disposable income per capita

Nonwhite Population Kelly and Witko, “Replication Data for: Federalism and American Inequality”

Proportion of population nonwhite

Population over 65 Kelly and Witko, “Replication Data for: Federalism and American Inequality”

Proportion of population aged 65+

Fiscal Pressure Unemployment Insurance Financial Data Handbook; Klarner, "State Economic and Government Finance Data”

Reserve ratio of the state’s unemployment benefit trust fund (UIRR models)

Ratio of the state’s general fund revenue to spending (SARR models)

Welfare Reform Welfare Reform Act 1996 1 = 1996, 2009Unemployment Insurance

TaxUnemployment benefits

federally taxable (1987)1 = 1987

Source: Authors.


Appendix C. Alternative Model Specifications

Table B2. Summary Statistics.

Variable Mean Std. Dev. Min Max.

Unemployment Insurance Net RR 56.02 7.8 27.7 84.18Net Social Assistance/Net

Minimum Wage RR35.2 8.5 19.1 75.9

Percentage Income to Top 1 % 13.3 4.4 4.3 35.9Ideology 50.4 11.8 19.5 82.6State GOP Control .15 .35 0 1Gubernatorial Election .26 .44 0 1Per Capita Income (log) 9.9 .25 9.2 10.5% Nonwhite 19.1 13.7 .54 71Population over 65 12.05 2.16 2.3 18.5Unionization Rate 15.6 7.2 2.3 38.7Fiscal Pressure 1.44 1.13 −.53 5.4

Source: Authors’ data; from sources in Table B1.

Table C1. Dynamic Fixed Effects Error Correction Models (Strongly Balanced Panels).

∆ UIRR ∆ UIRR ∆ SARR ∆ SARR

Long RunTop Income Share −.345*** −.246* −1.278*** −.698***

(.108) (.130) (.111) (.180)State Ideology −.119 .143**

(.0775) (.0613)Republican Control −4.414* .760

(2.323) (.769)Income per Capita (nlog) 2.040 −15.92***

(4.611) (4.145)% Nonwhite .0583 −.224*

(.120) (.114)Population over 65 −.602 −1.267**

(.730) (.546)Unionization .238 −.345**

(.290) (.170)Fiscal pressure

(reserve ratio or gen. fund deficit)

.439 .606 −5.783 −5.943(.514) (.455) (4.387) (4.890)

Short RunReversion Rate −.182*** −.205*** −.118*** −.131***

(.0402) (.0435) (.0133) (.0180)

(continued)



Top Income (short run) −.115*** .0216 −.146*** −.00841(.0352) (.0401) (.0163) (.0228)

Ideology (short run) −.0550*** −.0161* (.0201) (.00948)

Republican Control −.848*** −.256** (.224) (.125)

Election (prior year) −.125 .0994** (.0869) (.0449)

Income per Capita −.331 .609 (2.077) (1.530)

% Nonwhite −.00711 −.102** (.0639) (.0500)

Population over 65 −.286 .448 (.862) (.310)

Unionization .0560 −.0124 (.0619) (.0365)

Fiscal Pressure (reserve ratio or gen. fund deficit)

−.649*** −.839*** −1.639*** −.837(.233) (.300) (.397) (.527)

Constant 10.82*** 9.604 5.763*** 28.38***(2.640) (10.44) (.755) (6.559)

Observations 1650 1421† 1650 1421†

Start Year 1978 1978 1978 1978End Year 2010 2007 2010 2007

Note: Standard errors in parentheses; *p < .10, **p < .05, ***p < .01.†Nebraska excluded because of unicameral legislature.Source: Authors’ data.

Table C1. (continued)

(continued)

Table C2. Fixed Effects Regression for Program Replacement Rates (Differenced Model).


∆ Top 1 −.285*** −.226*** −.0882*** .0312*(.0376) (.0423) (.0168) (.0179)


−6.934*** −6.503*** (.413) (.435)

∆ Ideology −.0322* −.0136 (.0168) (.00901)

∆ State GOP Control −.372* −.345*** (.193) (.105)



∆ Gubernatorial Election .221*** −.0195 (.0810) (.0428)

∆ Per Capita Income (log) −1.983 .396 (2.466) (1.782)

∆ % Nonwhite .00793 −.0974* (.0589) (.0501)

∆ Population over 65 .858 .466* (.563) (.250)

∆ Unionization Rate .0491 −.00312 (.0497) (.0340)

∆ Fiscal Pressure −1.059*** −.478 (.257) (.414)

Welfare Reform (1996) −.891*** −.856*** (.110) (.124)

Constant .163*** .157** −.400*** −.560***(.0168) (.0637) (.00500) (.0538)

Observations 1657 1428 1700 1470

Note: Standard errors in parentheses; *p < .10, **p < .05, ***p <.01.Source: Authors’ data.


Table C3. Fixed Effects Regression for Program Replacement Rates (Lagged Model).

Unemployment InsuranceReplacement Rate

(UIRR) t−1Social Assistance

Benefit Rate (SARR) t−1

Top 1 t−1 −.351*** −.179* −1.344*** −.355***(.0755) (.0953) (.110) (.0997)


−3.175*** −3.470*** (.494) (.488)

Ideology t−1 −.0369 .0616** (.0410) (.0280)

State GOP Control t−1 −1.884** .621 (.959) (.455)

Gubernatorial Election t−1 .127 .177** (.102) (.0726)

Per Capita Income (log) t−1 −5.549* −23.85*** (3.089) (3.350)

% Nonwhite t−1 .0109 −.0544 (.0767) (.0725)

(continued)



Unemployment InsuranceReplacement Rate

(UIRR) t−1Social Assistance

Benefit Rate (SARR) t−1

Population over 65 t−1 .0452 −.154 (.377) (.800)

Unionization Rate t−1 −.105 −.0474 (.174) (.126)

Fiscal Pressure t−1 −.955*** −1.783 (.265) (1.782)Welfare Reform (1996) −.782*** −.763**

(.232) (.310)Constant 60.77*** 118.1*** 52.77*** 277.0***

(.996) (30.22) (1.465) (34.08)Observations 1700 1519 1700 1519

Note: Standard errors in parentheses; *p < .10, **p < .05, ***p < .01.Source: Authors’ data.

Table C4. Fixed Effects Regression for Program Replacement Rates (Including Neighbor State Average Inequality).



∆ Top 1 t−1 −.103** −.138* −.212*** −.0710*** (.0466) (.0711) (.0263) (.0225)Unemployment Benefits

First Taxed (1987)−6.250*** −6.152***

(.413) (.452) ∆ Neighbor State Average

Top 1 t-1 .0155 −.0304***

(.0171) (.0101)∆ Ideology t−1 .00784 −.000422 (.0139) (.00580)∆ State GOP Control t−1 −.0373 −.148 (.198) (.106)∆ Gubernatorial Election t−2 −.338** −.300*** (.149) (.0763)∆ Per Capita Income

(log) t−1 −.452 3.333***(2.878) (.985)

∆ % Nonwhite t−1 .128* −.129*** (.0664) (.0377)

(continued)





∆ Population over 65 t−1 −.903* .412* (.461) (.209)∆ Unionization Rate t−1 .00349 −.0173 (.0391) (.0282)Welfare Reform (1996) −.285 −1.366** (.227) (.516)∆ Fiscal Pressure t−1 −.838*** –.761** (.107) (.0615)Constant .0913*** .185* −.367*** −.479*** (.0170) (.106) (.00617) (.0439)Observations 1650 1410 1650 1410

Note: Standard errors in parentheses; *p < .10, **p < .05, ***p < .01.Source: Authors’ data.

Table C5. Fixed Effects Regression for Take Up Rate and Replacement Rate * Take Up Rate (Receipt of Benefits Only from Regular State UI Programs).

% Receipt State Program Only RR* % Receipt State Program

∆ Top 1 t−1 −.112 −.140* −.192** −.101** −.116** −.143**(.0672) (.0792) (.0871) (.0411) (.0517) (.0566)


−1.541*** −1.671*** −1.820*** −2.740*** −2.841*** −2.863***(.349) (.434) (.431) (.262) (.302) (.312)

∆ Ideology t−1 −.0415 −.0445 −.0202 −.0259* (.0273) (.0279) (.0154) (.0153)∆ State GOP

Control t−1 .583 .526 .367 .328

(.453) (.459) (.286) (.290)∆ Gubernatorial

Election t−2 −.0237 −.0407 −.135 −.155(.202) (.208) (.142) (.148)

∆ Per Capita Income (log) t−1

24.38*** 21.56*** 13.52*** 13.10***(4.965) (4.080) (2.514) (2.613)

∆ % Nonwhite t−1 .191** .187** .114** .131** (.0749) (.0743) (.0472) (.0512)∆ Population over

65 t−1 −5.541*** −5.659*** −3.329*** −3.352***

(.853) (.899) (.473) (.502)∆ Unionization

Rate t−1 .173** .152** .101** .101**

(.0693) (.0753) (.0465) (.0485)∆ Fiscal

Pressure t−1 4.256*** 4.369*** 2.409*** 2.395***(.352) (.362) (.214) (.230)

(continued)



% Receipt State Program Only RR* % Receipt State Program

∆ Neighbor State Average Top 1 t−1

.0659** .0385***(.0265) (.0108)

Constant −.0160 −.0396 .0251 .0329*** .0680 .0594 (.0199) (.174) (.152) (.0117) (.0903) (.0924)Observations 1650 1470 1410 1650 1470 1410

Note: Standard errors in parentheses; * p < .10, ** p < .05, *** p < .01.Source: Authors’ data.

Table C6. Fixed Effects Regression for Take Up and Replacement Rate * Take Up Rate (Receipt of Benefits from Regular State and Federally Financed Benefit Extension UI Programs).

% Receipt All Program RR* % Receipt All Program

∆ Top 1 t−1 −1.356*** −.676*** −.644*** −.799*** −.418*** −.397*** (.102) (.0885) (.102) (.0690) (.0610) (.0704)Unemployment

Benefits First Taxed (1987)

−1.970*** −1.230*** −1.225*** −2.976*** −2.600*** −2.536***(.403) (.437) (.459) (.295) (.313) (.327)

∆ Ideology t−1 .0688* .0571 .0471** .0364 (.0366) (.0376) (.0222) (.0222)∆ State GOP

Control t−1 1.856* 1.964** 1.188* 1.218*(.970) (.983) (.618) (.629)

∆ Gubernatorial Election t−2

−.272 −.217 −.314* −.311*(.244) (.240) (.176) (.179)

∆ Per Capita Income (log) t−1

11.07* 6.615 5.862* 4.742(6.349) (5.087) (3.506) (3.564)

∆ % Nonwhite t−1 .270*** .283*** .162*** .189*** (.0972) (.103) (.0585) (.0670)∆ Population

over 65 t−1 −7.237*** −7.105*** −4.376*** −4.253***(1.106) (1.163) (.633) (.667)

∆ Unionization Rate t−1

.346*** .348*** .196*** .212***(.112) (.123) (.0709) (.0769)

∆ Fiscal Pressure t−1

3.408*** 3.266*** 1.914*** 1.759***(.422) (.447) (.283) (.288)

∆ Neighbor State Average Top 1 t−1

−.0235 −.0159(.0555) (.0401)

Constant 1.043*** .429** .500*** .615*** .346*** .340***(.0198) (.191) (.183) (.0147) (.116) (.122)

Observations 1650 1470 1410 1650 1470 1410

Note: Standard errors in parentheses; * p < .10, ** p < .05, *** p < .01.Source: Author’s data.


Acknowledgments

The authors would like to thank participants at the UConn Political Economy Workshop, Christopher Dennis, Patrick Flavin, and the editors at Politics & Society for helpful comments on earlier versions of the article.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

1. Angus Maddison, Growth and Interaction in the World Economy: The Roots of Modernity (Washington, DC: AEI Press, 2005).

2. Thomas Piketty and Emmanuel Saez, “Income and Wage Inequality in the United States 1913–2002,” In A.B. Atkinson and T. Piketty, eds., Top Incomes over the Twentieth Century. A Contrast between Continental European and English-Speaking Countries (New York: Oxford University Press, 2007).

3. L.M. Bartels, Unequal Democracy: The Political Economy of the New Gilded Age (Princeton, NJ: Princeton University Press, 2008); Christopher Ellis, “Understanding Economic Biases in Representation Income, Resources, and Policy Representation in the 110th House,” Political Research Quarterly 65, no. 4 (2012): 938–51; Patrick Flavin, “Direct Democracy and Political Equality in the American States,” Social Science Quarterly 96 (2015): 119–32; Martin Gilens, “Inequality and Democratic Responsiveness,” Public Opinion Quarterly 69, no. 5 (2005): 778–96; Martin Gilens, Affluence and Influence: Economic Inequality and Political Power in America (Princeton, NJ: Princeton University Press, 2012); Thomas J. Hayes, “Responsiveness in an Era of Inequality: The Case of the U.S. Senate,” Political Research Quarterly 66 (2013): 585–99.

4. Martin Gilens and Benjamin I. Page, “Testing Theories of American Politics: Elites, Interest Groups, and Average Citizens,” Perspective on Politics 12, no. 3 (2014): 546.

5. Kevis Roberts, “Voting over Income Tax Schedules,” Journal of Public Economics 8 (1977): 329–40; Thomas Romer, “Individual Welfare, Majority Voting, and the Properties of a Lienar Income Tax,” Journal of Public Economics 14 (1975): 163–85; Allan Meltzer and Scott F. Richard, “A Rational Theory of the Size of Government,” Journal of Political Economy 89 (1981): 914–27.

6. Ira Katznelson, Fear Itself: The New Deal and the Origins of Our Time (New York: Liveright, 2013); L. Garfinkel, I. Fox, N. Kaushal, J. Waldfogel, and C. Wimer, “Waging War on Poverty: Historical Trends in Poverty Using the Supplemental Poverty Measure” (working paper, Columbia University of Social Work, 2013); Christopher Wimer, Liana Fox, Irv Garfinkel, Neeraj Kaushal, and Jane Waldfogel, “Trends in Poverty with an Anchored Supplemental Poverty Measure,” CPRC Working Paper no. 13-01 (New York: Columbia Population Research Center, 2013).

7. Karin Martinson and Pamela A. Holcomb, “Reforming Welfare: Institutional Change and Challenges,” Occasional Paper no. 6 (Washington, DC: Urban Institute, July 31, 2002).


8. Christopher Howard, “The American Welfare State, or States?” Political Research Quarterly 52 (1999): 421–42; Urban Institute, State and Local Finance Initiative, “State and Local Expenditure”; online at http://www.urban.org/policy-centers/cross-center-initiatives/state-local-finance-initiative/projects/state-and-local-backgrounders/state-and-local-expenditures.

9. Alan Weil and Kenneth Feingold, Welfare Reform: The Next Act (Washington, DC: Urban Institute, 2002); Matthew C. Fellowes and Gretchen Rowe, “Politics and the New American Welfare States,” American Journal of Political Science 48 no. 2 (2004): 362–73.

10. Gilens, Affluence and Influence.11. Bartels, Unequal Democracy; Gilens, Affluence and Influence.12. Elizabeth Rigby and Gerald C. Wright, “Political Parties and Representation of the Poor

in the American States,” American Journal of Political Science 57, no. 3 (2013): 552–65; Patrick Flavin, “Direct Democracy and Political Equality in the American States,” Social Science Quarterly 96 (2015): 119–32; Patrick Flavin, “Income Inequality and Policy Representation in the American States,” American Politics Research 40, no. 1 (2012): 29–59; Nathan J. Kelly and Christopher Witko, “Federalism and American Inequality,” Journal of Politics 74 (2012): 414–26.

13. Ping Xu and James C. Garand, “Economic Context and Americans’ Perceptions of Income Inequality,” Social Science Quarterly 91, no. 5 (2010): 1220–41; Benjamin J. Newman, Christopher D. Johnston, and Patrick L. Lown, “False Consciousness or Class Awareness? Local Income Inequality, Personal Economic Position, and Belief in American Meritocracy,” American Journal of Political Science 60, no. 3 (2014): 805–6.

14. Estelle Sommeiller and Mark Price, “The Increasingly Unequal States of America: Income Inequality by State, 1917 to 2011,” Economic Analysis and Research Network Report (Economic Policy Institute, February 19, 2014); online at http://www.epi.org/publication/unequal-states/.

15. Kim QuaileHill and Jan E. Leighley, “The Policy Consequences of Class Bias in State Electorates,” American Journal of Political Science 36 (1992): 351–65; Evan Ringquist, Kim Quaile Hill, Jan E. Leighley, and Angela Hinton-Andersson, “Lower-Class Mobilization and Policy Linkage in the U.S. States: A Correction,” American Journal of Political Science 41, no. 1 (1997): 339–44.

16. Charles Beard, An Economic Interpretation of the Constitution of the United States (New York: Macmillan, 1923); Gabriel Almond, The American People and Foreign Policy (New York: Harcourt, Brace, 1950; C. Wright Mills, The Power Elite (New York: Oxford University Press, 1956); E.E. Schattschneider, The Semi-Sovereign People (New York: Harcourt Brace Jovanovich, 1960); Suzanne Keller, Beyond the Ruling Class: Strategic Elites in Modern Society (New York: Random House, 1963); William G. Domhoff, The Higher Circles (New York: Vintage, 1971); Tom B. Bottomore, Classes in Modern Society (New York: Pantheon. 1966).

17. Meltzer and Richard, “A Rational Theory of the Size of Government.”18. Examples of such policies might include lobbying to support public policies that shelter

certain sources of (high) income from taxation, limit vertical redistribution (from rich indi-viduals to poor individuals) via transfers or nonexcludable public goods, restrict voting rights for the poor, expand opportunities for capital mobility.

19. The term originated, not from Peter H. Lindert, Social Spending and Economic Growth since the Eighteenth Century (Cambridge: Cambridge University Press, 2004), 15, but from Walter Korpi and Joakim Palme, “The Paradox of Redistribution and Strategies of Equality: Welfare State Institutions, Inequality, and Poverty in the Western Countries,

http://www.urban.org/policy-centers/cross-center-initiatives/state-local-finance-initiative/projects/state-and-local-backgrounders/state-and-local-expenditures



http://www.epi.org/publication/unequal-states/

http://www.epi.org/publication/unequal-states/


“American Sociological Review 63, no. 5 (1998): 661–87, which observed that countries (like the United States) relying more on targeting those in need and taxing the rich (which they call a Robin Hood strategy) achieve less poverty relief than states (like Sweden) pro-viding more generous universal benefits funded by broad-based taxation. Korpi and Palme explained the “paradox” by a median voter logic: the middle classes do not benefit from (and might, like the rich, pay for) means-tested programs, but do benefit from universal programs.

20. D. Acemoglu, S. Naidu, P. Restrepo, and J.A. Robinson, “Democracy, Redistribution and Inequality,” NBER Working Paper no. 19746 (Cambridge, MA: National Bureau of Economic Research, 2013).

21. Noam Lupu and Jonas Pontusson, “The Structure of Inequality and the Politics of Redistribution,” American Political Science Review 105, no. 2 (2011): 316–36.

22. T. Iversen and D. Soskice, “Electoral Institutions and the Politics of Coalitions: Why Some Democracies Redistribute more than Others.” American Political Science Review 100, no. 2 (2006).

23. Their explanation is that the “middle class” is more likely to side with the rich under majoritarian institutions, because of the risk that they and the rich are taxed to pay for transfers to the poor. In our setup the electoral systems are fixed, but the relative distance between poor, middle, and rich varies over time and space.

24. “Distribution of Household Income and Federal Taxes, 2011” (Washington, DC: Congressional Budget Office, November 2014); online at https://www.cbo.gov/publication/49440.

25. We see a similar, but more subtle trend if we look at the median compared to the a quintile just above and just below: the median has not grown closer to the quintile below, but has grown further from the one above, so the ratio differences in 2011 has fallen 10 to 15 per-cent since the late 1970s and early 1980s.

26. Emmanuel Saez, “Striking It Richer: The Evolution of Top Incomes in the United States” (unpublished manuscript, University of California, Berkeley, June. 2015); online at http://eml.berkeley.edu/~saez/saez-UStopincomes-2014.pdf. Saez’s estimates show even larger income concentrations than the CBO data, because the latter add estimates of the value of employer- or government-provided health insurance to incomes, increasing the measured pre- and posttax income of the lower quintiles considerably.

27. Gilens “Inequality and Democratic Responsiveness”; Gilens, Affluence and Influence.28. Gilens and Page “Testing Theories of American Politics”; Evelyne Huber and John Stevens,

Development and Crisis of the Welfare State (Chicago: University of Chicago Press, 2001); Jeffrey Winters, Oligarchy (New York: Cambridge University Press, 2011).

29. Adam Bonica, Nolan McCarty, Keith T. Poole, and Howard Rosenthal, “Why Hasn’t Democracy Slowed Rising Inequality?, “Journal of Economic Perspectives 27 (2013): 103–23.

30. This doubling of contributions was more than matched by the growth in the share of income accruing to the top .01 percent in the period. That figure increased from around 1 percent to 4 or 5 percent (depending on whether or not capital gains are included).

31. Bonica et al., “Why Hasn’t Democracy Slowed Rising Inequality?,” 10632. N. McCarty, K.T. Poole, and H. Rosenthal, Polarized America: The Dance of Ideology and

Unequal Riches (Cambridge, MA: MIT Press, 2006).33. Bonica, H. Rosenthal, and D.J. Rothman, “The Political Polarization of Physicians in the

United States: An Analysis of Campaign Contributions to Federal Elections, 1991 through 2012,” JAMA Internal Medicine 174, no. 8 (2014): 1308–17.



http://eml.berkeley.edu/~saez/saez-UStopincomes-2014.pdf

http://eml.berkeley.edu/~saez/saez-UStopincomes-2014.pdf


34. Benjamin I. Page, Larry M. Bartels, and Jason Seawright, “Democracy and Policy Preferences of Wealthy Americans,” Perspectives on Politics 11 (2013): 51–73.

35. D. Kuo and N. McCarty, “Democracy in America,” Global Policy 6 (2015): 49–55.36. Martin Gilens, “Preference Gaps and Inequality in Representation,” PS: Political Science

and Politics 42 (2009): 335–41.37. S. Verba, K.L. Schlozman, and H.E. Brady, Voice and Equality (Cambridge, MA: Harvard

University Press, 1995).38. P.K. Piff, M.W. Kraus, S. Côté, B.H. Cheng, and D. Keltner, “Having Less, Giving More:

The Influence of Social Class on Prosocial Behavior,” Journal of Personality & Social Psychology 99 (2010): 771–84.

39. M.W. Kraus, P.K. Piff, R. Mendoza-Denton, M.L. Rheinschmidt, and D. Keltner, “Social Class, Solipsism, and Contextualism: How the Rich Are Different from the Poor,” Psychological Review 119 (2012): 546–72; Paul K. Piff, “Wealth and the Inflated Self: Class, Entitlement, and Narcissism,” Personality and Social Psychology Bulletin 40 (2014): 34–43.

40. Michael W. Kraus, Stéphane Côté, and Dacher Keltner, “Social Class, Contextualism, and Empathic Accuracy,” Psychological Science 21 (2010): 1716–23.

41. Incomes for the GSS item are in constant dollars.42. The income distribution of donors and fundraising is, of course much more skewed toward

those with incomes over $150,000. M.J. Barber, “Representing the Preferences of Donors, Partisans, and Voters in the US Senate,” Public Opinion Quarterly 80, no. S1 (2016): 225–49, finds that 30 percent of donors to a recent Senate campaigns had family incomes over $350,000.

43. About 2 percent of responses in both surveys (twenty-one of 1140 in the June survey and seventeen of 931 in December) indicated incomes above $250,000, which is consis-tent with estimates from the Current Population Survey. Respondents with income over $150,000 were somewhat fewer in both surveys (6.5 percent versus 8.3 percent). It is important to point out that as these benefits extensions were federally financed, they had limited direct impact on budgeting.

44. ABC News/Washington Post Poll: 2014 Congressional Election/Congress/Affordable Care Act (New York: Langer Research Associates/Capital Insight/Abt SRBI, January 2014); ABC News/Washington Post Poll: July Monthly—Gulf Oil Spill/Elections/Economy/War in Iraq and Afghanistan (Horsham, PA: TNS Intersearch, July 2010); RoperExpress (Ithaca, NY: Roper Center for Public Opinion Research, Cornell University); accessed November 2106.

45. Gilens, Affluence and Influence ; Bartels, Unequal Democracy.46. Elizabeth Rigby and Gerald C. Wright, “Political Parties and Representation of the Poor in

the American States,” American Journal of Political Science 57, no. 3 (2013): 552–65.47. Gilens, Affluence and Influence.48. Barber, “Representing the Preferences of Donors, Partisans, and Voters.”49. Gilens and Page, “Testing Theories of American Politics,” 577.50. E.E. Schattschneider, The Semi-Sovereign People (New York: Harcourt Brace Jovanovich,

1960).51. Objections to this account are that it neglects the importance of other issues to voters and

that existing intraparty class divides are due to more important issues. Poor Republicans, e.g., prefer conservative social policy over a strong safety net; rich Democrats prefer lib-eral social policies more than they dislike paying for social transfers. Of course, each of these subgroups prefers changing policy to suit their preferred outcome on both dimen-sions if they can do it. Some commentators have noted that the 2016 presidential election


may increase the class divides inside both major political parties by driving more pro-business Republicans to support Democratic candidates. See, e.g., Thomas Edsall, “The Great Trump Reshuffle,” New York Times (May 4, 2016).

52. R. Goodin and J. Dryzek, “Rational Participation: The Politics of Relative Power,” British Journal of Political Science 10, no. 3 (1980): 273–92; V.A. Mahler, “Exploring the Subnational Dimension of Income Inequality: An Analysis of the Relationship between Inequality and Electoral Turnout in the Developed Countries,” International Studies Quarterly 46, no. 1 (2002): 117–42; F. Solt, “Economic Inequality and Democratic Political Engagement, “American Journal of Political Science 52, no. 1 (2008): 48–60. C.J. Anderson, and P. Beramendi, “Left Parties, Poor Voters, and Electoral Participation in Advanced Industrial Societies,” Comparative Political Studies 45, no. 6 (2012): 714–46. However, some studies find little inequality effect internationally. Daniel Stockemer and Lyle Scruggs, “Income Inequality, Development and Electoral Turnout: New Evidence on a Burgeoning Debate,” Electoral Studies 31 no. 4 (2012): 764–73.

53. F. Solt, “Does Economic Inequality Depress Electoral Participation? Testing the Schattschneider Hypothesis, Political Behavior 32, no. 2 (2010): 285–301. L.E. Leighley and J. Nagler, Who Votes Now?: Demographics, Issues, Inequality, ad Turnout in the United States (Princeton, NJ: Princeton University Press, 2013).

54. The Department of Labor publishes “average replacement rate” statistics for states, but these divide the average benefit paid by the average insured wage, which yields an indi-cator more of UI spending than of the program’s wage replacement rate. See also Robert Moffitt, David C. Ribar, and Mark Wilhelm, “The Decline of Welfare Benefits in the U.S.: The Role of Wage Inequality,” Journal of Public Economics 68, no. 3 (1998): 421–52.

55. J.P. Allan, and L. Scruggs, “Political Partisanship and Welfare State Reform in Advanced Industrial Societies,” American Journal of Political Science 48 (2004): 496–512; W. Korpi and J. Palme, “New Politics and Class Politics in the Context of Austerity and Globalization: Welfare State Regress in 18 Countries 1975–95,” American Political Science Review 97 (2003): 425–46; Moffitt, Ribar, and Wilhelm, “The Decline of Welfare Benefits”; G. Esping-Andersen, The Three Worlds of Welfare Capitalism. (New York: John Wiley & Sons, 1990).

56. “Fully qualified” means that they have worked at least long enough to receive the standard duration of benefits under state unemployment insurance law. In most states this requires about fifteen months of employment.

57. “Significant Provisions of State Unemployment Insurance Laws Insurance Laws Effective January 2011”; online at http://www.ows.doleta.gov/unemploy/content/sigpros/2010-2019/January2011.pdf.

58. Unemployment Insurance Financial Data Handbook (Washington, DC: Department of Labor, 2015); online at http://workforcesecurity.doleta.gov/unemploy/hb394.asp.

59. UI was not subject to state and federal income tax or FICA before 1986. Subsequently, it is taxable income (not earned income), but is not subject to FICA. Daniel Feenberg and Elisabeth Coutts, “An Introduction to the TAXSIM Model,” Journal of Policy Analysis & Management 12, no. 1 (1993).

60. We also examine the effect of using recipiency rates and replacement rates* recipi-ency as dependent variables in supplemental analyses (Tables C5 and C6 of Appendix C). Those results suggest that larger increases in income concentration are also asso-ciated with less generosity on these indicators, too. Wayne Vroman, “Low Benefit Recipiency in State Unemployment Insurance Programs,” Research Report (Washington,

http://www.ows.doleta.gov/unemploy/content/sigpros/2010-2019/January2011.pdf

http://www.ows.doleta.gov/unemploy/content/sigpros/2010-2019/January2011.pdf

http://workforcesecurity.doleta.gov/unemploy/hb394.asp


DC: Urban Institute, 2001); online at http://www.urban.org/research/publication/low-benefit-recipiency-state-unemployment-insurance-programs.

61. The unemployment benefit paid to an average earner is above the threshold for SNAP benefits.

62. Social assistance benefits are untaxed, unearned income in all the years in our data set, so net and gross benefits are the same.

63. The Welfare Rules Databook is online at http://wrd.urban.org/wrd/databook.cfm. The SNAP program operates with a national benefit schedule, so benefits are essentially iden-tical across the states. Hawaii and Alaska have higher Food Stamp allotments than are included in our estimates for those states. However, because SNAP benefit amounts count AFDC/TANF benefits as income, state SNAP entitlements differ across states and years depending on the state AFDC/TANF benefits.

64. Jeffrey Winters and Benjamin I. Page, “Oligarchy in the United States?,” Perspectives on Politics 7 (2009): 731–51.

65. Also see Christopher Dennis, Marshall H. Medoff, and Kerri Stephens, “The Impact of Tax Rates, Partisanship and Economic Variables on the Distribution of State and Local Tax Burdens,” Social Science Journal 48, no. 2 (2011): 305–13; Sommeiller and Price, “The Increasingly Unequal States of America.”

66. Thomas Piketty and Emmanuel Saez, “Income and Wage Inequality in the United States 1913–2002,” in A.B. Atkinson and T. Piketty, eds., Top Incomes over the Twentieth Century: A Contrast between Continental European and English-Speaking Countries (New York: Oxford University Press, 2007).

67. Ibid.68. Unemployment Insurance Financial Data Handbook.69. Carl E. Klarner, Justin H. Phillips, and Matt Muckler, “Overcoming Fiscal Gridlock:

Institutions and Budget Bargaining,” Journal of Politics 74 (2012): 992–1009; Carl Klarner, “State Economic and Government Finance Data,” 2015, Harvard Dataverse, V1 [UNF:6:wM3f6VAc2dlP8m+CmrXzUw==]: online at https://dataverse.harvard.edu/data-set.xhtml?persistentId=doi:10.7910/DVN/CJBTGD.

70. William D. Berry, Evan J. Ringquist, Richard C. Fording, and Russell L. Hansonn “Measuring Citizen and Government Ideology in the American States, 1960–93,” American Journal of Political Science 42, no. 1 (1998): 327–48; F.S. Berry and W.D. Berry, “Tax Innovation in the States: Capitalizing on Political Opportunity,” American Journal of Political Science (1992): 715–42.

71. National Conference of State Legislators, “State Partisan Composition” (September 26, 2016); online at http://www.ncsl.org/research/about-state-legislatures/partisan-composi-tion.aspx.

72. Alberto Alesina, Edward Glaeser, and Bruce Sacerdote, “Why Doesn’t the United States Have a European-Style Welfare State?,” Brookings Paper on Economic Activity no. 2 (Washington, DC: Brookings Institution, 2001),187–278.

73. Nathan J. Kelly and Christopher Witko, “Replication Data for: Federalism and American Inequality,” hdl:1902.1/16842, Harvard Dataverse, V3 (2011). Nathan J. Kelly and Christopher Witko, “Federalism and American Inequality,” Journal of Politics 74 (2012): 414–26.

74. Federal extensions of unemployment benefits affect the duration or benefits and are not considered in our calculation of state benefit replacement rates.

75. Empirical literature in this area has increasingly relied on estimation using single equation error-correction models. Recent reviews have suggested that such models are not nearly

http://www.urban.org/research/publication/low-benefit-recipiency-state-unemployment-insurance-programs

http://www.urban.org/research/publication/low-benefit-recipiency-state-unemployment-insurance-programs

http://wrd.urban.org/wrd/databook.cfm

https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/CJBTGD

https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/CJBTGD

http://www.ncsl.org/research/about-state-legislatures/partisan-composition.aspx

http://www.ncsl.org/research/about-state-legislatures/partisan-composition.aspx


as appropriate or flexible as commonly believed. Taylor Grant and Matthew Lebo, “Error Correction Methods with Political Time Series, “Political Analysis 24, no. 3 (2016): 6. There is, in fact, little to suggest that the vast majority of political science papers in the last decade that rely on ECMs meet the appropriate specification conditions or have been cor-rectly used and interpreted by the authors (or reviewers and editors) of those papers. This leads us to rely primarily on more robust, if less complex, methods. A Harris-Tzavalis and a Hadri test for unit roots in panel data and for individual series (augmented Dickey-Fuller test), suggest that both of our dependent variable series, may not be stationary, while the top income share series does appear to be stationary for many states when detrended. These problems suggest that differencing is more appropriate, if somewhat more demanding, estimation method. We also report standard variety ECM results in Appendix C, Tables C1–C6, to indicate that our results are not simply an artifact of our estimation approach.

76. Our approach is consistent with other work estimating relationships between inequality and social benefits (which also use annual data). See, e.g., Matthew C. Fellowes and Gretchen Rowe, “Politics and the New American Welfare States,” American Journal of Political Science 48, no. 2 (2004): 362–73; Nathan J. Kelly and Christopher Witko, “Federalism and American Inequality,” Journal of Politics 74 (2012): 414–26; and for an example at the cross-national level see Karl Ove Moene and Michael Wallerstein, “Inequality, Social Insurance, and Redistribution,” American Political Science Review (2001): 859–74.

77. Kelly and Witko, “Federalism and American Inequality,” find a positive influence of one state level policy (minimum wage) on inequality.

78. We also employed a number of alternative model specifications using a dynamic fixed effects error correction model, difference in difference model, and lagged model. These results (available in Appendix C) are similar to those reported in the next section.

79. The p-value of the estimate in Column 1 (the constrained model) is .031, while the p-value of the estimate in the full model estimated with OLS is .041. An over identification test suggests that pooled OLS is consistent, and thus there is no need to use the more inefficient fixed effects approach.

80. The over identification test suggests that a pooled OLS model is inconsistent, and that fixed effects estimation is the better specification. This is not unexpected; there are clear downward trends in most states in social assistance replacement rates over the years, even before the passage of federal welfare reform.

81. To examine the possibility that extreme inequality or shocks from such inequality from neighboring states might be influential, we estimated a variable that measures the aver-age percentage income going to the top 1 percent of neighboring states (weighted by population). We then estimated our model including this new neighboring states variable (Neighbor State Average Top 1). The results (Appendix C, Table C4) show that our esti-mates do not change in a significant manner, though the standard errors increase slightly.

82. See. e.g., Winters and Page, “Oligarchy in the United States?”

Author Biographies

Lyle Scruggs ([email protected]) is a professor of political science and member of the Human Rights Institute’s Research Program on Social and Economic Rights at the University of Connecticut. His recent research examines the impact of economic conditions on attitudes about climate change and the relationship between social insurance protection and health outcomes.

Thomas J. Hayes ([email protected]) is an assistant professor at the University of Connecticut. His research focuses on American politics and political behavior, with an empha-sis on the politics of economic inequality.



Lyle Scruggs and Thomas J. Hayes...Scruggs and Hayes 37total public spending on welfare and are...

Documents

Transcript of Lyle Scruggs and Thomas J. Hayes...Scruggs and Hayes 37total public spending on welfare and are...