Polarization of the Rich: The New Democratic Allegiance of Affluent Americans and the Politics of Redistribution

Affluent Americans used to vote for Republican politicians. Now they vote for Democrats. In this paper, I show detailed evidence for this decades-in-the-making trend and argue that it has important consequences for the U.S. politics of economic inequality and redistribution. Beginning in the 1990s, the Democratic Party started winning increasing shares of rich, upper-middle income, high-income occupation, and stock-owning voters. This appears true across voters of all races and ethnicities, is concentrated among (but not exclusive to) college-educated voters, and is only true among voters living in larger metropolitan areas. In the 2010s, Democratic candidates’ electoral appeal among affluent voters reached above-majority levels. I echo other scholars in maintaining that this trend is partially driven by the increasingly “culturally liberal” views of educated voters and party elite polarization on those issues, but I additionally argue that the evolution and stasis of the parties’ respective economic policy agendas has also been a necessary condition for the changing behavior of affluent voters. This reversal of an American politics truism means that the Democratic Party’s attempts to cohere around an economically redistributive policy agenda in an era of rising inequality face real barriers.

impact the state of U.S. inequality?The central battle of politics is over using the power of the state to redirect economic resources (Hacker and Pierson 2014), and while education and geographic polarization seem descriptively true, scholars have overlooked other characteristics of evolving voter blocs, which may, in fact, have more obvious implications for party competition over the distribution and redistribution of economic resources. 4In this paper, I show that affluent and even very affluent voters-measured by income, but also stock ownership and occupation-have left the Republican Party to join the Democrats in a decadeslong process, and I argue that we should pay renewed attention since this trend has meaningful implications for the politics of economic (re)distribution.
Empirically, I show that the voter coalitions of the U.S. Democratic and Republican parties have changed significantly in the past couple of decades.The three prominent publications from the early-mid 2000s that aimed to assess the degree that "class politics are alive and well" (Bartels 2008, 96) in America (which all used income as the primary marker of class) include Stonecash (2000), McCarty, Poole, and Rosenthal (2006), and Bartels (2008).All three of those publications concluded that partisan conflict in the United States was increasingly class-based: from the 1950s to the 1990s, Republicans gained more and more support from the rich (and higher-educated), and Democrats won votes from lower-income (and less-educated) voters.In this paper, I first replicate and extend all three of their analyses through the present moment.While those authors were largely correct about the 1950s through the 1990s, their conclusionstraditionally believed to be conventional wisdom-no longer accurately depict contemporary politics.In today's America, both political parties largely split the support of higherincome voters.
Second, in this paper, I expand the measurements of economic interests (to characterize "affluent") that are traditionally used in American politics research.I use income (as others have), but I also rely on occupation and stock ownership, two additional dimensions that characterize distinct, important forms of a voter's economic position and relationship to redistributive economic policies.I analyze data from the American National Election Studies (ANES), 5 Cooperative Election Study (CES), 6 and General Social Survey (GSS). 7This original, holistic assessment reveals that in today's politics, Democrats and Republicans roughly split support among rich, upper-middle income, high-income occupation, and stock-owning voters.In further nuancing the analysis, while the voters in more advantaged economic classes who have increased their support for the Democratic Party are, on average, more educated and reside in larger metropolitan areas, this polarization of the affluent is spread across all races and ethnicities and is not only relegated to college-educated voters.Because the Democratic Party has increased its support from voters in more advantaged economic classes over the past few decades, the voters in each party's coalition have started to look more similar to one another: this is driven by the polarization of the affluent in contemporary American partisan competition.Now, the Democratic and Republican voter coalitions look far more similar to one another by income than they ever have (see, e.g., figures 1 and 2).
Strikingly, in multiple elections since the 2010s, the data actually show a form of "backwards" polarization: 8 majorities of affluent voters voted for Democratic candidates.Specifically, some evidence shows that Democratic candidates actually beat Republicans in attaining support from the top 5% (by income), the highest income stockowning voters, and even the top 1% of voters (by income) over the past decade.Later in the paper, I discuss the significance of this, and the degree to which it may have been caused by the uniqueness of Donald Trump as the Republican presidential nominee in 2016 and 2020.
This paper fills multiple gaps in the American politics literature on affluence and political behavior.Various scholars have studied the allegiance of working-class voters to the Republican Party (e.g., Bartels 2006;Cramer 2016;Hochschild 2018;Carnes and Lupu 2021), but most research has largely left untouched the shift of voters from more advantaged economic classes toward the Democratic Party. 9(Scholars of European politics have led American politics researchers in documenting this trend in their countries-see, e.g., Evans and Tilley 2012;Gingrich and Hauserman 2015;Oesch and Rennwald 2018.)Those that have studied this shift in American politics have made important contributions, but they have been limited.Gelman et al. (2008) showed that in blue states, higher-income voters split their votes between the parties.Nowadays, however, this polarization has happened nationally and in more specific ways than they catalogued then.Kitschelt and Rehm (2019) showed that the more educated top third of the income distribution now supports Democratic candidates.Gethin, Martínez-Toledano, and Piketty (2021) showed that a majority of the top 10% (by income) voted for the Democratic presidential candidate in the 2010s.This is all illuminating work, though it has either focused on a rather large top income group (e.g., top 33%) or has been disconnected from meaningful debates on effects on the substance of politics or the causes of this trend.
While this existing literature has begun to become aware of the changing nature of the U.S. partisan conflict, it still faces significant limitations: this prior work is either largely impressionistic, relies on usually just one dataset, distills economic interest into just income, or only assesses higher-income categories that are as large as the top 33% or top 10%-not any smaller or finer-grained.In this paper's empirical analyses, I transcend each of those limitations-in the American political context-by considering multiple plausible measurements of affluence across various forms of data, reaching as high as the top 5% and top 1% income categories (in addition to occupation and stock ownership categories), and connecting this trend to theoretical (and suggestively empirical) effects on the politics of redistribution and to likely causes of this trend.As a whole, the polarization of the affluent that I document in this paper is a stark change from the U.S. politics of the twentieth century.
Beyond this paper's empirical analysis, I rely on political economy theory and suggestive policy cases to argue that this shift in compositions of the parties may make it more difficult for the Democratic Party to execute an economically redistributive agenda-in an era of rising inequality-since it would have to redistribute away from voters in its own coalition.In the following sections, I replicate and extend the aforementioned income-based partisan analyses by prominent political scientists, showing that the twentieth-century characterizations are no longer accurate.To more fully describe the contemporary party voting coalitions by affluence, I display original analyses of ANES, CES, and GSS data that measure income, occupation, and stock ownership.Finally, I delineate discrete causal factors that may be driving the polarization of affluent American voters and conclude by articulating how an expanded research agenda can clarify our understanding of this new reality of U.S. partisan competition.

Income and the Reversal of Twentieth-Century Voting Trends
After the turn of the millennium, prominent scholars of American politics consistently showed that income polarization had been steadily increasing since the 1950s: moreaffluent Americans were increasingly voting Republican and less-affluent Americans were increasingly voting Democratic (Stonecash 2000;McCarty, Poole, and Rosenthal 2006;Bartels 2008).Those previously accurate findings are no longer true-and they have reversed.My replication and extension of their findings largely shows thatwith the exception of the 2008 Obama-McCain election -from the 1990s through 2020, higher-income voters have consistently drifted from supporting Republican to Democratic presidential candidates (figure 1).The Republican Party used to consistently win support from broadly higher-income voters by 10-20 percentage points, but in the most recent two or three elections, the Democratic Party has won a majority of this group.
Figure 1 shows that the relative advantage of Republican presidential candidates among higher-income (i.e., the top 33%) voters has declined since the 1990s. 10Democratic candidates have increased their support among this affluent upper third to the point of winning a majority of these voters in 2016 and 2020.The left portion of figure 1 shows a replication and extension of Stonecash (2000), who compared the partisan voting behavior of the bottom 33% (by income) to the top 33% (a higher y-value depicts a lower-income voter advantage for Democratic candidates).The right portion shows a similar replication and extension of Bartels (2008), focusing only on the voting behavior of the top 33% (a higher y-value depicts a higherincome advantage for Democratic candidates).Dashed vertical lines show the end of the initial analyses that once showed income polarization.These two authors' primary analyses focused only on white voters-as they were responding to claims about political behavior of the white working class-so I display the results among voters of all races and ethnicities along those of only white voters.Both kinds of analyses show the same result: while the trend is more pronounced among white voters, 11 it is still clearly true that among all voters, the most-affluent 33% have become increasingly Democratic.This is a stark reversal of previous trends in American politics. 12 However, these kinds of analyses face multiple limitations when trying to understand the precise nature of the polarization among rich voters in recent political history.These types of analysis rely on just one dataset (the American National Election Study, ANES), they distill economic position into just income, and they only assess more-affluent categories that are as large as the top 33% or top 20% (i.e., McCarty, Poole, and Rosenthal 2006)not any smaller or finer-grained.In the analyses that follow in this paper, I transcend each of these limitations, showing a fuller and more nuanced picture: High-income voters (even as high as the top 5% and 1%), high-income (and lower-"risk") occupation voters, and stock-owning voters have all increasingly supported the Democratic Party over recent decades and years.

The New Reality: Affluent Voters Have Swung Democratic
For decades, scholars have noted segments of collegeeducated, professional, or middle-class voters trending toward the Democratic Party.Ladd and Hadley (1975) began describing an "inversion of the New Deal class order" (233) as certain educated voters in professional jobs shifted their allegiance toward the Democratic Party.Edsall and Edsall (1992) made a similar argument, describing the polarization between Republican and Democratic agendas on "race, rights, and taxes," drawing some educated, middle-and upper-middle income white voters toward the Democratic Party.But through the latter decades of the twentieth century, Republicans were still most often winning higher-educated and higher-income voters, on average.Gelman et al. (2008) were some of the first authors to show how in richer (often Democratic) American states, higher-income voters split their votes between the parties in the 1990s and 2000s.More recently, Gethin, Martínez-Toledano, and Piketty (2022), as part of a comparative analysis of western democracies, showed the United States is the only country where the decade of the 2010s saw the top 10% of society (by income) roughly split their support between left and right parties (i.e., Democrats and Republicans).Clearly, a restructuring of voters by economic class has been taking hold.Kitschelt and Rehm (2019) have provided the most comprehensive picture and understanding of the restructuring of American voter groups: while income (solely) used to predict partisan support, in today's electorate, college-educated, lower-income (i.e., bottom 66% by income) voters reliably vote Democratic, but collegeeducated, higher-income (i.e., top 33%) voters have shifted toward the Democrats while lower-income, noncollege-educated voters have shifted away from Democratic candidates.Crucially, Kitschelt and Rehm were studying white voters only, for whom these categorizations of voter behavior by income and education are starkest.
Kitschelt and Rehm's thorough characterization is very helpful on a broad (white) electorate level.While they do show an increase in Democratic allegiance of higherincome voters who are college-educated, their paper does not make any arguments about the voting behavior of all affluent voters overall (e.g., the degree to which the affluent are now Democrats, due to education polarization and an increasingly educated population). 13In this paper, my concern is the behavior of the affluent as particularly unique voters, so in what follows, I extend the findings of Kitschelt and Rehm (2019) and Gethin, Martínez-Toledano, and Piketty (2022) to hone in on the specific categories of income (at very high levels), occupation, and stock ownership that characterize the increasingly Democratic and now-majority Democratic affluent voters -in addition to describing these affluent groups by education, race and ethnicity, and geography.
Original analysis of ANES and Cooperative Election Study (CES, formerly CCES) data yields evidence that the Democratic voting coalition of the 2010s has taken the form of a "U-shape" by income, which is a departure from the past. 14Figure 2 shows simple breakdowns of marginal vote choices for the Democratic presidential candidate (i.e., Democratic minus Republican vote percentages) by voter income group.The left portion of figure 2 shows ANES data, whose time-series data file gives dissimilarly sized (but finer-grained at the top) income groups (i.e., 0%-33%, 34%-66%, 67%-95%, and 96%-100%).The right portion shows CES data, which only goes back to 2008 (for presidential election years). 15Both charts show that it is increasingly the case that the income groups that most prefer Democratic candidates are the lowest and highest income categories-hence, a "Ushape."For example, in 2016 and 2020, CES data shows that the top two income quintiles (i.e., 80%-100% and 60%-80%) preferred the Democrat (i.e., Hillary Clinton or Joe Biden) over the Republican (i.e., Donald Trump) more than the twentieth through sixtieth percentiles did.ANES data shows that in the past, more-affluent voters preferred Republican presidential candidates more than any other income groups-until 1992 (which figure 1 showed was a moment of particular increase among affluent voters drifting to Democrat Bill Clinton's coalition) as a sign of what would come later: 2012 then looks similar to 1992, and 2020 (as well as 2016, although not shown) brings on the U-shape.CES data shows how in recent Figure 1 Replicating and extending Stonecash (2000, left) and Bartels (2008, right) years, the U-shape has become increasingly more pronounced, as Obama, Clinton, and Biden won increasing shares of affluent voters.Online appendix figures 2 and 3 show that similar U-shapes characterize voting for candidates for U.S. House and governor-this is not unique to the presidential level.(Refer to online appendix figure 4 for a similar analysis of GSS data by self-reported wealth categories.) How affluent, exactly, are these new Democratic voters?CES uniquely began collecting data on family income categories up to $500,000 and above in 2011 (including $350,000-500,000, $250,000-350,000, etc.), which

Figure 2
The contemporary Democratic voting coalition's U-shape (ANES, CES) appears to be the highest income sub-categories to exist in any over-time political survey data.For reference, $500,000 and higher (by family income) is roughly the top 1% of society; $200,000 and higher is roughly the top 5% (in 2012) to 10% (in 2020).Figure 3 (left portion) shows that $200,000þ category clearly preferred the Democrat in 2012, 2016, and 2020-and even the $500,000þ category reported voting for the Democratic candidate more often than the Republican in 2012 and 2016, dropping to just below 50% in 2020.
The same chart from figure 3 also shows all stockowning voters (just over half of American society) 16 have increased their Democratic allegiance consistently since 2008-particularly the stock owners in the top 10%-15% of the income distribution (i.e., $150,000þ family income in 2008, 2012, 2016, and 2020).Stock ownership is not often used as a relevant political variable for individuals.However, two reasons make a voter's household stake in the stock market plausibly politically meaningful.First, the financialization of the American (and global) economy theoretically makes the ownership of stocks and other financial assets a more significant indicator of economic position (e.g., Davis and Kim 2015).Second, financial market investments are a source of government tax revenue via capital gains taxes and (less directly) corporate taxes.Therefore, whether a voter owns stock or not is a potentially important component of their relationship to economic policies.Taken all together, this is some evidence of polarization among the very high-income and most-affluent stock owners, as they have come to support Democratic candidates even more often than Republican candidates.
The right portion of figure 3 shows an increase in Democratic allegiance across the highest income-earners (i.e., top 25% in any given year) across multiple-nearly all-occupation categories (GSS data).In particular, the highest income-earners in "human services/arts" and "professional/scientific" occupations (i.e., most likely to fall into Kitschelt and Rehm's (2019) higher-education, higher-income typology) have become the most Democratic in recent decades. 17While occupation type is generally correlated with income, it is not perfectly correlated, and there are theoretical (and empirical) reasons to think that occupation is a unique economic interest that drives preferences and behavior.Iversen and Soskice (2001) famously argued that occupation-and the economic risks of investing in different skills associated with different jobs-is a uniquely good predictor of social insurance policy preferences, above and beyond income.Other work has shown that people at higher risk of unemployment are particularly likely to support welfare state policies (e.g., Rehm, Hacker, and Schlesinger 2012;Trubowitz 2022).Studying changes in the class basis of British political parties, Evans and Tilley (2012) use income and occupation to characterize economic class.The trend in affluent U.S. voters swinging left (i.e., Democratic) by voting behavior has been happening among multiple professional and managerial occupations, in addition to merely having higher incomes and owning stock. 18  The Increasingly Democratic Affluent Are Higher-Educated, Multiracial, and Reside in Big Metro Areas American politics literature has established trends such as geographic, education, and racial and ethnic polarization between the parties: urban voters, more-educated voters, and voters of color have become (on average) more likely to support the Democratic Party, while rural voters, lesseducated voters, and white voters have shifted toward the Republican Party.Thus far, I have shown that higherincome, stock-owning, and high-income occupation voters have shifted from the Republican to Democratic coalition over recent decades.However, this may not be similarly true across all other categories of voters, and the nuances within this overall trend will likely give us clues about potential effects and causes of the new Democratic affluent voter bloc.Further, the existing studies documenting partisan voting behavior changes of the affluent have completely ignored race/ethnicity and geographic location of residence as purely descriptive quantities of interest (to be sure, Kitschelt and Rehm 2019 and Gethin, Martínez-Toledano, and Piketty 2022 did center educational attainment).Therefore, in this section, I analyze the income trends by education, race and ethnicity, and geography.
The CES data provides the best opportunity for analysis of affluent voting behavior by sub-group.Looking at presidential vote choice of the $150,000þ family income group (i.e., the top 10%-15%) by educational attainment and race and ethnicity in figure 4 show the patterns that American politics literature has described more broadlythat higher-educated voters (i.e., those with college and post-graduate degrees) and voters of color are much more likely to support the Democratic Party than white voters.However, the data show this income polarization among the affluent happening to somewhat similar degrees between voters of different education levels and of different races and ethnicities.Since 2008, higher-income voters with post-graduate and college degrees have increased their Democratic presidential vote share by roughly fifteen percentage points.Perhaps surprisingly, voters with less than a college degree have also increased their Democratic support by about ten percentage points.Similarly, high-income respondents who select CES raceethnicity options Black, Hispanic, or Asian have all increased their Democratic support since 2008 (although Hispanic support dropped from 2008 to 2012 before increasing from 2012 to 2020), while high-income white respondents have increased their Democratic support as well.The Democratic Party's increasing support by these higher-income voters is not limited to voters of any particular race or ethnicity.It is concentrated among more-educated voters but is not exclusive to them.
Beyond the standard y-axis values (share of sub-group voting for the Democratic presidential candidate), figure 4's charts display an additional piece of information: the share of the entire affluent (i.e., $150,000þ family income) group made up by the particular sub-group identifier.Specifically, the upper-left portion of figure 4 shows that in 2008, just 27.8% of this most-affluent voter group had post-graduate degrees, while 41.4% of the group had less than a college degree-but by 2020, 35.8% of this affluent group had post-graduate degrees (an 8.0 percentage point increase), and just 31.1% had less than a college degree (a 10.3 percentage point decrease).In other words, this most-affluent group has gotten increasingly educated (by college and post-graduate degree attainment) over time-and similarly, this most affluent group has become less white and increasingly Black, Asian, and Hispanic (upper-right portion of figure 4).
Analogous analysis by voters' geographic residency (i.e., how urban or rural) in figure 4 is unsurprising, given what the existing literature has shown.CES collects data on the respondent county of residence, so it is possible to analyze partisan support by population density. 19Analyzing data from 2008-2020, an extremely stark pattern emerges: Among the top 10%-15% by income, the shift toward the Democratic Party has occurred particularly among voters who live in "central" and "fringe" areas of metropolitan areas with one million or more people.As figure 4 shows, voters who live in rural areas ("Non-core") or small towns ("Micropolitan") have stayed or become increasingly supportive of the Republican Party (by presidential voting).It is the higher-income voters who live in big metropolitan areas who have become Democrats. 20 The shift of affluent voters toward the Democratic Party appears true across voters of all races and ethnicities, is concentrated among (but not exclusive to) collegeeducated voters, and is only true among voters living in larger metropolitan areas.

Regression Analysis
Plotting the trends visually on over-time charts is useful, but that kind of analysis does not allow the detection of precise statistical relationships.Therefore, in figure 5 I present plots of linear regression coefficients (for income) to show that high-income status as a binary indicator (measured in five ways across two charts) has become uncorrelated with presidential vote choice, by presidential election year, in both ANES and CES data.The dependent variable is Democratic presidential candidate choice (0/1). 21Crucially, control variables are included in these regressions: education, race/ethnicity, and gender (all binary).For many years in the ANES data, earning a high income (either top 33% or top 5%) was correlated with choosing the Republican candidate (i.e., earning a higher income meant a negative likelihood of choosing the Democratic candidate, as the negative, statistically significant coefficients in figure 5 show).
However, starting in 2012 for the top 5% (and 2016 for the top 33%), earning a high income is no longer correlated with presidential vote choice.The statistical significance has disappeared, even independent from education, race-ethnicity, and gender-meaning that affluent voters started splitting their votes between the parties.In the CES data, earning a top 10%-15% income (i.e., $150,000þ family income) or 20%-30% (i.e., $100,000þ) has not correlated with vote choice in any election from 2008-2020, and for the top 10%-15%, 2016 and 2020 show a nearly (or actual) statistically significant positive chance of voting for the Democratic candidate.Income polarization among the rich has taken hold.
In summary, Democratic politicians have increasingly attracted support from the top 33% (by income), top 20%, top 10%-15%, top 5%-10%, and top 1%, in addition to stock-owning and high-income occupation voters in recent decades.Some of these affluent groups have swung majority-Democratic in recent elections (e.g., the top 5%-10% per figure 3 and top 10%-15% per figure 5).The new affluent Democratic voters generally do live in big or medium metro areas, but this group is very multiracial (including a slim majority of affluent white voters and clear majorities of affluent voters of color), and while affluent Democrats are more likely to be college-educated, the non-college-educated affluent have also become more Democratic.Why does all this matter?I now turn to the potential consequences of this new political reality-and the political-economic forces driving this polarization of affluent American voters.

Effects of the Increasingly Democratic Affluent on the Politics of Redistribution
Winning support from the most affluent American voters is quite likely to enhance constraints on the economically redistributive nature of a Democratic Party policy agenda.Long literatures in political science and political economy have studied the ways that the economic class bases of political party coalitions may structure policy agendas and outcomes (the macro).A few contemporary moments also shed some light on how the contemporary Democratic Party attempts to do economic policy, which is occasionally redistributive but often is not (while reacting to an increasingly affluent voter base).Further, recent scholarship has interrogated the ways that economic interests influence the policy preferences of individual voters (the micro), which is relevant to understanding the degree to which affluence implies opposition redistribution (and what forms)-in general, the literature shows that economic interests do condition policy preferences, even if other factors may also structure what voters demand from public policy.
Comparative political economists have, for decades, empirically shown that the working classes-a concept not always measured in the same way, e.g., sometimes by occupation type, income, education, or otherwise 22 have historically been the voting bases of socialdemocratic, center-left, and socialist political parties (e.g., Lipset 1960;Przeworski and Sprague 1986;Kitschelt 1994).These parties are far more likely than conservative or center-right parties to use the power of the state to enact economically redistributive policies.Gunderson (2022) recently showed that the degree to which political parties (in Europe) polarize on economic issues is conditioned by how polarized the respective party voter bases are by income.A few scholars of U.S. politics have mirrored these types of inquiries, showing some evidence that the parties that gained higher shares of poor, working-and middle-class voters also enacted more robust welfare state policies across the American states (Jennings 1979;Brown 1995).Class bias in voter turnout (independent of who votes for which party) has been correlated with-and plausibly a causal driver of-economic policy choices in American states (Franko, Kelly, and Witko 2016).However, not much explicit, contemporary attention has been paid to the policy consequences of the economic class basis of partisan conflict in America.
Social-scientific measurement innovations have shown that economic inequality (whether by income, wealth, upward mobility, or other plausible measures) has increased over several decades in America (e.g., Chetty et al. 2014;Saez and Zucman 2016).At the same time, both major political parties have found themselves with complete control of the federal executive and legislative branches since 1980 (Republicans from 2003(Republicans from -2007(Republicans from and 2017(Republicans from -2019;;Democrats from 1993Democrats from -1995Democrats from , 2009Democrats from -2011Democrats from , and now 2021Democrats from -2023)).It is not necessarily surprising that the economically conservative Republican Party governing agenda did not cause inequality to decrease (and in fact oversaw an increase in inequality, per Bartels' 2008 analysis).It is more puzzling that Democratic control of the federal government has not meaningfully impacted the state of American inequality-and the class basis of Democratic voters may be related to this governing record.
Recent American politics literature has shown that Republican presidents tended to exacerbate economic inequality, while Democrats, after winning office, mitigated inequality (Bartels 2008;Kelly 2009).However, the effects of Democratic presidents on inequality tended to disappear beginning in the 1990s-which is when the Democratic Party began attracting increasing shares of affluent voters (e.g., figure 2).Some scholars have argued that the Democratic and Republican Party agendas converged, more or less, on multiple areas of regulation of the economy (e.g., financial regulation) beginning in the 1990s, and this is one way that politics has perpetuated inequality in society (Keller and Kelly 2015;Kelly 2020).Historical work has shown how the Democratic Party agenda moved in an anti-welfare, anti-organized labor, and ideationally pro-free market direction in the 1990s as the Democratic Leadership Council and Bill Clinton rose to prominence within the party (Geismer 2022).
During Barack Obama's tenure as president, the Democratic Party did pass the Affordable Care Act, which raised taxes on high incomes and expanded health insurance coverage to lower-income Americans.This was undoubtedly economically redistributive.Further, Republican opposition in Congress and institutional barriers to policy change (e.g., the filibuster in the Senate, Enns et al. 2014) also mitigate Democratic attempts to pass some economic policies. 23During Joe Biden's tenure thus far, the Democratic Party expanded the child tax credit for the 2021 calendar year (disproportionately helping lowerincome parents, although without raising anyone else's taxes) and-at the time of this writing-has passed a bill through the U.S. House that would raise some taxes on the super-rich (and corporations) in order to fund various welfare policies. 24It is clear that the Democratic Partythe center-left United States political party-does enact some forms of a redistributive economic policy agenda.
However, the ability to tax higher incomes-both by campaigning perception and governing execution-has seemingly waned over time for the Democrats.Obama campaigned on the promise to not raise taxes on any families making less than $250,000. 25At the time, that was roughly the bottom 95% (by income)-promising to leave the ninetieth to ninety-fifth percentiles of Americans completely untouched.In 2020, Biden campaigned on not raising taxes on anyone making $400,000 or more. 26 This was roughly the entire bottom 98% in 2020, up even from Obama's promise to the entire bottom 95%.Further, as referenced earlier, the recently-passed U.S. House bill's second most costly provision actually cut taxes for the highest-earning Americans (via raising the state and local tax deduction cap). 27Earlier in time, during the Obama presidency, some Democrats floated reforming (in a redistributive direction) two other types of tax breaks that benefit wealthier Americans-the mortgage interest tax deduction 28 and 529 college savings plan tax break 29but those efforts were squashed, after opposition from some other Democrats in Congress.
Regarding government spending policies, some recent comparative political economy work, American politics research, and Democratic Party policy agenda priorities have some clues about what effects this shift in voting bases portends.Multiple recent studies of European politics show some suggestive evidence that center-left partieswho have been gaining increasing support from highereducated, middle-and high-income professionals-have shifted their economic policy agendas from mitigating "old" labor market risks (e.g., unemployment) to "new" risks (e.g., education, childcare spending), therefore overall still supporting some forms of welfare state policies (Gingrich and Hausermann 2015;Abou-Chadi and Emmergut 2019).One American study of the Democratic Party platform found that since the 1990s, it has increased its rhetorical mentions of welfare policies, although it has not much increased its mentions of redistributive taxation (Malpas and Hilton 2021).On the federal level in the contemporary political moment, the Democratic presidential and congressional policy strategy has included the American Rescue Act, Infrastructure Investment and Jobs Act, and the Build Back Better Act (which only passed through the U.S. House), displaying a party agenda that prioritizes spending on infrastructure (both physical and "human" infrastructure) and will do short-term welfare spending when necessary, but largely leaves increased progressive taxation-or, more broadly, imposing any costs on more-affluent voters-untouched.This is suggestive evidence that a Democratic Party that wins significant support from affluent voters may find itself with increased opposition within its own party to a more significantly redistributive policy agenda.At a basic theoretical level, is it in a political party's interest to redistribute away from some members within its own voting coalition?Taking widely employed political economy assumptions seriously, the answer is clearly no (all else equal).Further, there is evidence that affluent voters' preferences may have outsized influence on federal policy (e.g., Gilens and Page 2014), which may mean that these newer, richer voters in the Democratic coalition have disproportionate power over the party's agenda.To be sure, the national Democratic Party's economic policy agenda has moved "left" in some ways since the 1990s and even since the 2000s and early 2010s (e.g., on levels of public spending, labor rights regulation, climate policy), but these are arguably due to the rise of intraparty challengers like Bernie Sanders (in 2016 and 2020) and aligned political forces, in spite of increasing party allegiance from affluent voters, who stand to lose from various forms of redistributive policies.The increasing share of voters from more advantaged economic classes in the Democratic coalition may be one force harming America's center-left party's ability to meaningfully mitigate economic inequality.
The evidence displayed in this paper is all effectively about economic class interests, as distinct from policy preferences, and preferences are what we assume to be the demands that drive political parties to choose certain policy agendas.Basic political-economic theory-that material interests are significant drivers of preferences and behavior-indicates that having richer voters in the Democratic Party coalition would make it more difficult for Democratic politicians to raise some forms of taxes on this higher-income, professional or managerial occupation, stock-owning voters and even to respond to different priorities that these voters have about government spending or regulatory policy.Further, some evidence about the campaign promises of Presidents Obama and Biden and legislation passed in 2021 by congressional Democrats under Biden suggest that this may be the case.But how significant might this effect be?Some evidence from survey research on policy preferences suggests that some affluent Democratic voters do express support for enhanced welfare policy spending (Gilens and Thal 2018) and tax-and-spend redistribution (Franko, Tolbert, and Witko 2013), Democratic donors do express economically redistributive tax-and-spend preferences (Broockman and Malhotra 2020), and high-income voters who work in (big metropolitan area) knowledgeeconomy sectors 30 -who lean Democratic, as this paper showed-do support various forms of government social insurance spending (Ansell and Gingrich 2021).Overall, this evidence suggests that more-affluent Democratic voters may support some forms of economic redistribution.To be sure, factors such as educational attainment and partisanship likely influence policy preferences, independent from economic interests, as various scholarship suggests (e.g., Franko, Tolbert, and Witko 2013;Kitschelt and Rehm 2019)-but that does not mean that material interests are not always operating in some (at least) partial form.
When trying to understand voters' policy preferences using survey data-as one relevant kind of evidence to evaluate the possible effects of the polarization of the affluent-there are two issues with taking stated survey preferences completely seriously.First, they are "stated" preferences (as distinct from "revealed" preferences), and for various cognitive reasons, they may not equate to how affluent voters behave politically, when voting for candidates or ballot questions, giving campaign donations, and more.Second, they do not capture the relative intensities of the preferences that affluent voters hold on various policies: It may be the case that more-affluent Democratic voters do favor some forms of economic redistribution beyond the status quo (i.e., we can believe that their stated preferences on surveys are their true policy preferences), but they hold more intensely policy preferences on less redistributive social/cultural issues that drive their behavior.
Other forms of evidence show that affluent liberals and Democratic voters may not favor redistributive economic policies as much as they say in abstractly worded statedpreference surveys-evidence that their economic interests do cause them to prefer less redistributive economic policies than other voters. 31For example, liberal homeowners oppose redistribution in the form of expanding affordable housing access in their neighborhoods (Marble and Nall 2021).Historical evidence shows that selfidentified liberal and Democratic voters in the Boston suburbs were partially motivated by low property taxes and tough-on-crime policies (Geismer 2014).One analysis of dark money donations (i.e., non-survey quantitative evidence)-a rare form of evidence to come to light-showed that liberal donors donated significant amounts in opposition to redistributive ballot initiative campaigns in California in 2012 (Oklobdzija 2019).On surveys that measure preferences on concrete progressive taxation policies-that are explicit about certain affluent voters paying more in taxes-affluent Democrats express less favorable preferences than all other Democrats, and about half the time do not express majority support at all (Zacher 2022).Since the policy preferences of the more-advantaged classes of Democratic voters are one key mechanism through which the party's redistributive policy agenda may change (as a result of polarization among affluent voters), better understanding the economically redistributive policy preferences of voters on various policies will be key to adjudicating any effects on the policy agenda.For these reasons, evidence described earlier in this paper about Democratic elites' inability to raise income taxes on upper-middle-and high-income voters-and even recent passage of a tax cut for rich voters in a Democratically controlled U.S. House bill-may lend some credence to the "materialist" hypotheses, that affluent voters in the coalition may mitigate the economically redistributive potential of the Democratic Party.
My analysis ultimately aims to contribute to the broader understanding of why we observe inequality in democracies-and more specifically, the rather significant and distinct form of economic inequality in the contemporary United States. 32Numerous scholars have put forth unique reasons why economic inequality persists-and even expands-in democratic countries, primarily highlighting divisions and resentments between racial, ethnic, and religious groups and "capture" of the political system by privileged economic interests (for recent reviews, see Bonica et al. 2013;Scheve and Stasavage 2017).(To be sure, while I focus only on the changing political behavior and economic interests of voters, I do not dispute that the battle of organized interests may significantly drive the politics of policymaking, e.g., Hacker and Pierson 2014.)This paper underscores how the class basis of party competition may be an additional, distinct-and interactive-factor that drives inequality in America and other contemporary democracies.In the United States, a Democratic Party that wins increased support from affluent voters may diminish its ability to enact redistributive policies that would keep rising inequality in check.

Why Have Affluent Americans Swung Democratic?
Political scientists tend to agree that political competition generally happens on two dimensions of issue bundles, the social/cultural and the economic.Debates surrounding the shifting of American party voter coalitions and party policy agendas have often centered on social/cultural issues.I propose that there are four possible proximate causes to the changing voting behavior of affluent Americans: a) social/cultural issue shifts among i) voter preferences or ii) party policy agendas or b) economic issue shifts among i) voter preferences or ii) party policy agendas.I will describe each in turn, highlighting the under-recognized nature of the economic dimension for both party policy agendas and for voter preferences.
As Kitschelt and Rehm (2019) make clear, the rights revolution of the 1960s and 1970s, followed by the transition from an industrial to a knowledge economywhere the economic gains to educational attainment have vastly increased-has created a new class of voters with "libertarian" (a.k.a.socially/culturally liberal) views on issues such as abortion, LGBTQþ rights, drug laws, racial inclusion, and other issues that involve extending civil rights to subgroups of Americans.Dovetailing with a literature that argues that voters with more education and more income are more likely to hold more socially/ culturally liberal views than other voters (e.g., Inglehart 1981;Broockman, Ferenstein, and Malhotra 2019) which may even be more likely than their economic views to drive their voting behavior (Enke, Polborn, and Wu 2022), many scholars agree that many more Americans today simply hold more socially/culturally liberal views than ever before.One likely crucial element to the increase in socially/culturally liberal voters is the increasing (although still minority) shares of Black, Latino, and other voters of color earning relatively high incomes (e.g., figure 4), in addition to the decrease in white Americans expressing "racially resentful" views-which are symbolic and do not necessarily imply concrete economic policy preferences-over time (Clemons 2022).These various interrelated forces shaping the social/cultural views of many affluent voters, in particular, very likely drive more of them to feel more allegiance to the Democratic over Republican Party, as there is clear polarization between the parties on these issues.This explanation certainly carries some weight in explaining the Democratic shift among affluent voters.
A large chorus of scholars and pundits have described how the U.S. parties' policy agendas (and messaging) have polarized on the social/cultural dimension over many decades.The Democratic Party's relative warmth toward the movements for Civil Rights and women's liberation of the 1960s and 1970s created the impression that Democrats care more than Republicans about the well-being of Americans who are not white and not men (e.g., Ladd and Hadley 1975;Edsall and Edsall 1992;Kitschelt and Rehm 2019).Followed quickly by the rise of the Christian Right's ascendance within the Republican Party, by the 2000s, Democrats and Republicans had clearly opposing policy agendas on abortion and LGBTQ+ rights, and Republicans had begun winning majorities of white Americans (e.g., Miller and Schofield 2008;Abrajano and Hajnal 2015;Hacker and Pierson 2020).The Democrats had become the party of social/cultural liberalism, through and through (Brownstein 2021).The pulling apart of the party agendas on this range of issues likely attracted some affluent (more educated) Americans, particularly in stark contrast to the increasingly conservative Republican agenda-which, interacted with an increasingly social/ cultural base of affluent voters, has clearly been a main part of the story of the new rich Democratic voting base.
When it comes to possible voter preference changes on economic issues, the transition from an industrial to a knowledge economy is once again relevant.Economic activity in the twenty-first century is now far more likely to be generated in cities and metropolitan areas and primarily powered by people with college and postgraduate degrees, and these voters do seem to favor increased government spending on infrastructure, research and development, transportation, and what might be called "new labor market risks" such as education and childcare spending-essentially, policy demand created by the economic activity of a more educated workforce living in dense metropolitan areas in both the United States and Europe (Gingrich and Hausermann 2015;Abou-Chadi and Emmergut 2019;Ansell and Gingrich 2021;Hacker, Pierson, and Zacher 2021).In this a certain class of upper-middle income (educated) voters may have preferences for new kinds of government spending (which the Republican Party has generally opposed on ideological grounds for decades, e.g., Hacker and Pierson 2020), potentially pulling some affluent voters toward the Democratic coalition.There is some merit in this explanation, although likely not as much as the social/cultural issue dimension, given the inconsistent salience of issues like infrastructure, education, and government-funded research in national political competition over decades (especially compared to the salience of social/cultural issues).
Finally, the economic policy platforms of the Democratic and Republican parties have changed over decades, and I argue this is a more important part of the story of affluent voters swinging Democratic than previously imagined.In the 1990s and 2000s, Democratic elites converged with Republicans on immigration and trade (i.e., becoming more open to them, e.g., Greenberg 2017; Geismer 2022), financial deregulation (Keller and Kelly 2015;Kelly 2020;Geismer 2022), and somewhat on welfare (Geismer 2022).Democratic politicians like Bill Clinton were also more hostile toward organized labor than any Democrats for decades (Geismer 2022; in part a reflection of the decline of the power of organized labor, e.g., Rosenfeld 2014), and while the Democratic Party platform has increased its rhetorical focus on means-tested and public goods spending since the 1990s, it has not much increased its focus on increasingly progressive taxation (Malpas and Hilton 2021), even after Democrats played a key role, alongside Ronald Reagan, in massively cutting the top income tax rates in the 1980s (Prasad 2019).To be sure, the Republican Party's economic agenda has remained largely in favor of minimal government taxation, spending, and regulation for decades (Hacker and Pierson 2020).While Obama and congressional Democrats did pass the Affordable Care Act and Biden and congressional Democrats did pass large spending on COVID-19 stimulus and physical infrastructure in 2021, the Democrats at the head of the party have not meaningfully altered the policy agenda in ways that would threaten the interests of affluent Americans.Therefore, it is quite likely that the Democratic economic policy agenda's relative friendliness toward affluent income earners, homeowners, and stock owners (even as high as the top 5%) is a necessary condition for keeping and increasing the share of affluent voters from the 1990s through 2020.
In summary, the divide between the parties on social/ cultural issues interacted with the increasing share of socially/culturally liberal voters-who prefer relatively economically costless forms of extensions of civil rights to more subgroups of Americans-is clearly a major driving force in bringing more-affluent Americans to the Democratic Party voter base.And it may be the case that educated, somewhat affluent voters who live and work in large metropolitan areas have new demands for economic investments that only the Democratic Party has been willing to consider.I also propose that a necessary condition to the increasingly-and now-majority Democratic allegiance-of affluent voters is the moderate nature of the Democrats' economic policy agenda: Obama, Biden, and other Democratic leaders do want to use government to increase spending and some kinds of regulation, but they do not want to impose direct economic costs on any segments of affluent voters to execute a redistributive agenda.

Did Donald Trump Cause Polarization among Affluent Voters?
Readers may wonder about how these trends interact with another major new force in American politics: Donald Trump.He was clearly a unique Republican presidential candidate in 2016 and 2020.Trump campaigned on a platform that was more anti-immigration, anti-trade, and pro-social safety net than nearly all Republicans in recent memory.He also broke from the mold via his rhetoric, framing other political actors more starkly as winners or losers and employing insults, among other ways (e.g., Ross and Rivers 2020). of his uniqueness-and because he provoked seemingly stronger reactions, among both supporters and opponents, than prior presidential candidates-it is worth inquiring, how much of the swing of affluent voters to the Democratic Party may be caused by Trump's prominence and two-time candidacy?
A few clues show that this trend is not primarily related to Trump.First, the ANES and CES over-time data (e.g., figures 1 and 2) show that the shift of higher-income voters toward the Democratic Party started decades before Trump's rise.Second, the data on voting for U.S. House and gubernatorial candidates (online appendix figures 2 and 3) show that mostly similar kinds of trends have been underway in the decisions for non-presidential (i.e., non-Trump) candidates.Third, European studies of shifts in the voter coalitions of different parties show that the trend of more-educated, urban, middle-and higher-income voters toward center-left parties (and less-educated voters toward right populist parties) is happening in those countries, too (e.g., Gingrich and Hauserman 2015;Oesch and Rennwald 2018).
At the same time, nearly all the figures in this paper show that 2016 and 2020 were the elections when moreaffluent voters most preferred Democratic candidates in recent memory, even at above-majority levels (a sort of "backwards"-i.e., severely unexpected-polarization).Those two elections were the first time that majorities of the top 5% (by income) voted for the Democratic candidate.There are a few reasons why this recent political behavior by the rich may not be unexpected.Survey evidence has consistently shown that the wealthiest Americans are the most liberal (or moderate) on issues like trade and immigration (e.g., Kitschelt and Rehm 2019; Broockman and Malhotra 2020) that Trump campaigned against, contrary to other Republicans in recent memory.Relatedly, an impression of Trump was that he was more antielite in his rhetoric, in addition to being less predictable in general-two things that higher-income Americans may be more likely to oppose than other voters, as they are a certain kind of "elites," and they may relatively prefer political-economic stability.Finally, as establishment Democratic politicians, nominees Hillary Clinton and Joe Biden (in 2016 and 2020, respectively) may have seemed relatively safe choices for the richest Americans.For all these reasons, Trump's candidacy in 2016 and 2020 may have made those two recent elections outliers, even in comparison of this trend through the 2000s and early 2010s.Therefore, while Trump clearly did not spark the trend of affluent voters choosing Democrats over Republicans-which really began in the 1990s and has been happening in other advanced democracies-his candidacy may very well have accentuated the trend in recent elections.

Discussion and Conclusion
In this paper, I have introduced original analyses (including replications and extensions of prominent American politics research) to show that affluent voters have swung toward the Democratic Party over recent decades, after a twentieth century when Republicans increasingly had a strong grip over rich voters.In recent decades, the top 33% (by income), top 20%, top 10%-15%, top 5%-10%, and top 1%, in addition to stock-owning and high-income occupation voters in recent decades, have all increased their allegiance to the Democratic Party.And in multiple recent elections (i.e., 2016 and 2020, and to a lesser degree, 2012), some most-affluent groups supported Democrats at above-majority levels.This new group of affluent Democratic voters is very multiracial, and while higher-income Democrats are more likely to be collegeeducated, the non-college-educated affluent have also become more Democratic.
I have also made the case that this trend may have implications for the politics of inequality and redistribution via the economic policy agenda of the Democratic Party.While the Republican Party remains committed to an economically conservative agenda of cutting taxes and public spending, in the face of decades-long rising economic inequality the Democratic Party is the only party embroiled in intraparty battles over the redistributive nature of the policy agenda.Affluent voters have some clear interests opposed to many forms of redistributive taxation and spending and regulatory reforms, and they occasionally express preferences to this effect on surveys.The Democratic Party's economic agenda has become friendlier to larger levels of public spending (e.g., American Rescue Plan, Build Back Better proposals) and has prioritized some new problems for government to solve (e.g., childcare, incentives for states and municipalities to reform zoning laws), though these shifts seem to have been driven by the rise of the left flank of the Democratic Party and political competition within party primaries and occasionally within intraparty legislative negotiation within Congress.Democratic forces that fought Obama's efforts to reform some tax deductions (e.g., mortgage interest, 529 college savings account) seem still alive and well in their staunch defense of cutting taxes for higher-income voters via raising the state and local tax deduction cap in 2021 Democratically led legislative dealmaking.A Democratic Party that wins significant support from affluent Americans may find it more difficult cohere around an economically redistributive policy agenda.Effects may take multiple elections to take hold, so we may not yet have seen the extent the consequences of new affluent Democratic voters on the policy agenda.
Future research can do much to tease out effects of this trend.Studying the nature of Democratic economic agendas and legislative victories in finer detail through policy and electoral case studies and over-time quantitative analyses could enhance our understanding of the Party's evolution.This may be particularly fruitful at the state and local levels, since there are many states with complete Democratic control-allowing for variation and possible causal analyses-and because subnational governments have unique powers over many policy areas that are crucial to the contemporary forms of inequality we observe (e.g., housing, college and K-12 education, police and prisons).Causal analyses may be particularly possible by leveraging changes in electoral district boundaries (of voters by differing economic interests), for example.More correlational work may also be useful to shed light on differing policy stances by legislators on various economic issues, based on what kinds of voters they represent.At the micro voter level, more work can be done to understand the revealed policy preferences (e.g., state initiatives and referenda questions) and in primary election settings, where Democratic candidates sometimes distinguish themselves from their competition.
Subsequent scholarship can also better illuminate the causes of the polarization of affluent voters.How much is due to changes in the party agendas on social/cultural issues-and on which specific issues-or to independent changes in voter social/cultural preferences?Whereas how much may be due to changes-or, more subtly, stasis-in economic policy agendas of Democratic versus Republican candidates?Primary election settings may also be useful here, to study how affluent voters (compared to others) respond to certain candidates, when they have a true choice in a primary.It may be possible to study exogenous changes to the saliency of issues that may be differently likely to create interest and preference cleavages within Democratic constituencies on candidate choices in primaries.Moreover, better understanding true policy preferences by voters on a host of issues-transcending the limitations of stated-preference survey data and incorporating the intensities of various preferences-can shed light on the microfoundations of causes of voter coalition evolution, in addition to potential effects.
In this paper, I have centered voters as relevant political actors.Voters surely have political power-even if only as a constraint on candidates in primary and general election settings-but organized groups surely do, too.In understanding the economic policy cleavages within the Democratic Party, we should also look more closely at coherence and disagreement between organized actors such as businesses (particularly those earning profits in large metro areas where affluent, educated Democrats live and work and where Democratic mayors govern), labor (which has been weak in recent decades but is experiencing a recent upsurge since 2020), 33 homeowners and renters (most active in local politics, where housing policy saliency has risen), and left-leaning social movement groups (e.g., Movement for Black Lives) pushing more radical economic demands than the Democratic Party has seen for many decades.Further, individual affluent voters also attempt to influence politics via donating campaign funds, which may be an additional force impacting the Democratic policy agenda-particularly in primary election settings-worth studying.Voters play a vital role in American politics, and the polarization of the affluent may portend other sorts of cleavages in the battle over the politics of redistribution that the Democratic Party wades through, in the face of entrenched inequality and further transition from an industrial to a knowledge economy.labor, i.e., they do not own any of the tools (forms of capital) used in the production process in which they work.These workers "were almost exclusively the manual laborers in mining, manufacturing, construction, and in some countries agriculture and forestry" (Przeworski and Sprague,33).American politics scholars like Bartels (2008) and Stonecash (2000) have sometimes measured "working class" by lacking a college degree and sometimes by earning a relatively low income.Carnes and Lupu (2021) combined income and education in their measure.Carnes (2013) measured working class based on occupation type.23 However, even when using the "budget reconciliation" tactic in the 2021-2022 Congress, for which Democrats need only fifty votes to pass some policies and can get around the sixty-vote filibuster threshold, they still face opposition within their own ranks to many forms of redistributive economic policies.

Figure 3
Figure 3The now-majority Democratic support of the top 5%, top 1%, stock owners, and highest-income occupations