Elites moved toward democrats more than nonelites moved away: Income, education, and occupational class in US presidential elections, 1980–2020

Significance

Recent work on partisan realignment has often highlighted shifts among the White working class. Examining elections from 1980 to 2020, we show that the partisan realignment of White voters has been primarily driven by higher status Whites—those with the highest decile household incomes, college degrees, and white-collar jobs—making large moves toward the Democratic Party. Working-class Whites, however, show no clear long-term trend. They have indeed become less Democratic since 2012, but these levels remain within historically observed bounds.

Abstract

Recent discussion of voting in US elections claims a strong movement of White working-class voters away from voting for Democrats, with much discussion focusing only on elections between 2012 and the present. We examine longer-term trends from 1980 to 2020 in how more and less privileged White voters—measured by household income, education, and occupational class—moved toward or away from voting Democratic. We also explore how these movements changed the shape of the relationships between these three socioeconomic indicators and voting Democratic. We find little evidence of a long-term movement away from Democrats among voters with lower income, less education, or working-class jobs, although there is some evidence of this after 2012. The clearest long-term trend is that voters in the highest decile of income, college graduates, and white-collar workers moved steadily toward voting Democratic across the 40 y. Thus, the change from negative to flat for income’s relationship to voting Democratic, and from negative to positive for education’s relationship to voting Democratic comes less from a movement of less privileged voters away from Democratic voting and more from a long-term movement of those in the top decile of income, college graduates, and white-collar workers toward voting Democratic. Whether the post-2012 movement away from voting Democratic among voters without a high school degree and in working-class jobs becomes an enduring trend or is idiosyncratic to Trump’s candidacy is an important question for future research.

In the aftermath of Donald Trump’s 2016 presidential victory, many researchers and commentators claimed that the White working class had moved away from Democrats and thus played a pivotal role in his victory (1–4). The claimed exodus of the White working class from the Democratic Party was explained in a variety of ways. By some accounts, the Democratic Party had failed to offer meaningful economic solutions for the working class in the face of deindustrialization and offshoring (1, 4, 5). Others claimed, more generally, that Democrats lost working-class voters because they had turned away from the redistributive policies that appealed to them (6–8). But the non-White working class, who were also facing manufacturing job losses and erosion of the minimum wage, had not turned away from the Democratic Party (9, 10). Thus, other commentators argued that the discontent unique to White working-class voters had its roots in the election of Barack Obama, who, as the first Black president, was argued to have presented a racial status threat to many White working-class voters (11–13).

Trump came to political prominence in the midst of a long-term partisan realignment of White voters (10, 14). In the decades since World War II, income has become less predictive of vote choice while education has switched from being negatively to being positively associated with Democratic support (15–17). The changing relationship between income, education, and vote choice could come about, as much recent literature has suggested, from a shift of low status voters from the Democratic Party to the Republican Party. It is also true, however, that voters with college degrees and those in professional occupations have moved to the left (14, 15, 18). Approaches that examine trends in the gaps between groups often do not make clear whether the changes in these gaps are the result of lower status Whites moving to the right, higher status Whites moving to the left or some combination of the two (8, 10, 15).

This paper contributes to our understanding of ongoing partisan realignment along the dimensions of income, education, and occupational class by documenting which groups of voters have moved, to what extent, and when. Unlike previous analyses which have often treated income and/or education as a binary variable (e.g., refs. 15 and 18) or assumed its effects are linear (e.g., ref. 8), we allow for nonlinearities across multiple levels of income and education. We also examine broad occupational groupings that can be interpreted roughly as “class.” Unlike many analyses that have focused just on elections where Trump was on the ballot, we examine 40 y of elections, from 1980 to 2020.

This timeline covers many important political, economic, and social developments believed to influence voting preferences. These include the Republican Party’s appeals to Southern White voters in the decades following the Civil Rights Movement (19), growth in the social and economic returns to a college degree (17), and increases in economic inequality that have been particularly damaging to rural and working-class Whites (20).

Materials and Methods

Considering presidential elections from 1980 to 2020, we used data from the cumulative 1982–2024 General Social Survey (GSS) which is a nationally representative survey of English- and Spanish-speaking households (21). We limited our analyses to White individuals who reported having voted in the most recent presidential election, resulting in an analytic sample of 27,795.^* We limit to White voters as this is the group among whom partisan realignment has taken place (10). The dependent variable was a respondent’s vote choice, operationalized as either voting for the Democratic candidate or another candidate (Republican or third party).

We use three variables to identify dimensions along which realignment has been thought to take place: income, education, and occupational class. Each of these variables measures a distinct type of socioeconomic status, but each is positively correlated with the other two. For income, we use household income, adjusted for inflation and household size, expressed in deciles. Specifically, we follow the methods described in Hout (22) to assign dollar values to the categorical income variable available in the GSS, adjust those numbers for inflation, and then divide those numbers by the square root of household size (23). Finally, we divide these numbers into deciles for each year. We calculate deciles using the full sample, but the analyses use only White respondents who reported voting. We split education into five levels: less than a high school diploma, high school diploma, some college, college (baccalaureate) degree, and advanced degree. We follow Morgan and Lee’s (24) system of class categorization which uses occupation to assign individuals to the working class, an intermediate class, and a white-collar class.^† The working-class category includes manual workers (e.g. roofers, dishwashers) and service workers of a lower grade (e.g. cashiers, bus drivers, mail clerks). The intermediate class includes nonprofessional self-employed workers, higher-grade technicians (e.g. clinical laboratory technicians), and supervisors of manual workers (e.g. construction supervisors). The white-collar class includes professionals and managers (e.g. lawyers, computer programmers, registered nurses, elementary and middle school teachers) and routine nonmanual and service employees of a higher grade (e.g. flight attendants, tax preparers, secretaries). To see the survey questions underlying these variables, refer to SI Appendix, Tables S1–S4. With these variables, we created descriptive plots showing vote outcomes separately for each income group, educational group, and occupational class across elections from 1980 to 2020. These show the trend for each income, education, or occupational class group. Then, arraying the same data another way, we created descriptive plots showing vote outcomes across levels of each independent variable for each election. These show the shape of the relationship between voting and income, education, or occupational class. Lowess smoothing is used to give a sense of overall trends, but our graphs also show the observed percent voting Democratic for each group at each election.

Results

Fig. 1 shows the trend in voting Democratic over time for each decile of income, separately. Fig. 1 provides little support for the idea that the Democratic Party is hemorrhaging low-income Whites. Deciles 2 and 5 show drops in Democratic support for both Trump elections, but the CIs are quite large (in lower income deciles, White respondents are underrepresented, and voter turnout is lower). Furthermore, deciles 1, 3, and 4 fail to exhibit any clear Trump effect. The lowest decile remains fairly evenly split in its vote choice, as has been true since 2004. The third decile saw a drop in Democratic support in the 2020 election, but not for the first Trump election. The fourth decile shows a large drop in Democratic support for the 2016 election, but a complete recovery of Democratic support for the 2020 election. Also, between 1984 and 1992, Democratic support rose in most of the lower income deciles. Looking at the full time period, there is no clear long-term trend of voters in the bottom five deciles moving away from voting for Democratic presidents.

Fig. 1.

Trend in percent of White voters voting for the Democratic presidential candidate for each income decile.

What is clear, however, is the long-term shift of the top decile toward the Democratic Party. In 1980, only 27% of White voters in the top decile supported the Democratic candidate, but in the 2020 election, this number was 61%. The smoothed trend shows this as a linear increase, but it seems to be taking place in spurts. Between 1988 and 2004, Democratic support in the top decile hovered around 35%. In 2008, this support jumped to 50% and stayed near that for the next two elections. The 2020 election saw another jump, with the top decile supporting the Democratic candidate 61% of the time. It also appears that since the 1980s, deciles 8 and 9 have also moved toward the Democratic Party, albeit much less dramatically than the top decile.

Fig. 2 shows trends by education level. White voters with less than a high school diploma certainly did move away from the Democratic Party in the Trump elections. Between 2012 and 2020, White voters with less than a high school diploma decreased from 49% voting Democratic to only 26%. That being said, this is a shrinking portion of the electorate; although those with less than a high school diploma made up 18% of all White voters in the 1980 election, they represented only 4% of all White voters in the 2012 election and 3% in the 2020 election. The decrease in Democratic support among White voters with a high school diploma is less dramatic. Democratic support among White voters with a high school diploma decreased eight percentage points between the 2012 and 2016 presidential elections (43% vs. 35%) before increasing one percentage point (to 36%) in 2020. But their level of Democratic support in the 2020 election is higher than in the 1984, 1988, 1992, and 2000 elections, meaning that the level of Democratic support among Whites with a high school diploma in the 2020 presidential election is well within historically observed bounds. White voters with some college have been on a downward trend since 2008, but their Democratic support in 2020 was also not dramatically low compared to their historical average. In 2008, 47% of White voters with some college supported the Democratic candidate, and although this number dropped to 37% by the 2020 election, it was still higher than several previous elections including 1984, 1988, 1992, 2000, and 2004.

Fig. 2.

Trend in percent of White voters voting for the Democratic presidential candidate for each education level.

When we look at the top two educational levels, Fig. 2 shows that both bachelor’s and advanced degree holders exhibit a clear, long-term shift toward the Democratic Party. This is quite extreme for White voters holding advanced degrees. In 1980, 34% of voters in this group supported the Democratic candidate. By 2020, this number had grown to 68%. Democrats also saw gains among White voters with a college degree, a group whose Democratic support rose from 29% to 53% between the 1980 and 2020 presidential elections. White voters with a bachelor’s degree saw a particularly dramatic increase, 17 percentage points, from 36% to 53%, in their Democratic support between 2004, when Republican George W. Bush defeated John Kerry, and 2008, when Democrat Barack Obama beat John McCain. Since the 2008 election, Democratic support among White voters with a bachelor’s degree has remained relatively stable at these new, higher levels.

Fig. 3 shows these same trends by occupational class. The working and intermediate classes follow similar trends. Each shows a decrease in Democratic support in the elections with Trump on the ballot. Among working-class White voters, Democratic support decreased from 47% in 2012 to 37% in Trump’s first election and 36% in the 2020 election. Among White voters in the intermediate class, Democratic support dropped from 43% in the 2012 presidential election to 32% and 33% in the 2016 and 2020 elections, respectively. This recent low for both the working and intermediate classes is similar to previous low points of Democratic support in the 1984 and 2000 elections. However, while Fig. 3 does show the working class turning toward Trump between 2012 and 2020, it also makes clear that the working class has not been moving away from Democrats consistently, but fluctuated up and down between 1980 and 2012 without a clear pattern. In contrast to this lack of a long-term trend for the working class, white-collar White voters exhibit a large increase in their Democratic support for the entire 1980 to 2020 period. In the 1980 presidential election, only 32% of white-collar voters supported the Democratic nominee, but this number grew to 54% by the 2020 election. Similar to the 10th income decile and voters with a bachelor’s degree, this increase seems to happen at a somewhat uneven pace. Democratic support among white-collar voters saw a particularly large increase of 11 percentage points between the 2004 and 2008 elections (from 41% to 52%). This new level of Democratic support has been sustained in subsequent elections.

Fig. 3.

Trend in percent of White voters voting for the Democratic presidential candidate for each occupational class.

These descriptive statistics offer some support for the idea that lower status Whites have moved away from the Democratic Party since Trump entered politics. Often, however, the level of Democratic support among these lower status groups is comparable to previous elections. Moreover, it is important to recognize that there is not a long-term pattern of the White working class moving away from voting for Democrats. Time will tell if the changes in the Trump elections are the beginning of a new, long-term downward trend or a temporary movement in response to a particularly popular candidate.

A clearer takeaway from these descriptive plots, however, is that, in the long term, the partisan realignment of White voters has been primarily driven by higher status Whites—that is those with the highest decile household incomes, college degrees, and white-collar jobs—moving toward the Democratic Party. Often, although not always, these increases in Democratic support are particularly pronounced in 2008 with relatively steady levels of Democratic support in the following elections.

How have these shifts influenced the relationship between vote choice and measures of socioeconomic status? Fig. 4 shows, descriptively, the shape of the relationship between income and Democratic support among White voters, separately for each election in the dataset. Income often has a negative relationship with Democratic voting for White voters. But, since 2008, income has failed to have the clearly negative association with Democratic support that it has generally been thought to have. By 2020, except for the 10th decile voting for Democrats more than any other decile, the relationship is quite flat, with income simply not associated with vote choice across the first 9 deciles.

Fig. 4.

Percent of White voters voting for the Democratic presidential candidate at each income decile, by election.

Past work has argued that education used to be negatively associated with Democratic support (15). Fig. 5—which descriptively plots the relationship between level of education and Democratic support among White voters in each election—does show this for 1980, but more often shows a slightly U-shaped relationship between education and vote choice whereby White voters with the highest and lowest levels of education were the most likely to support Democratic candidates in presidential elections. But starting around the 2008 election and continuing into the present, this U-shape has become a positive and fairly linear relationship. This is particularly striking in the 2020 election where White voters with an advanced degree were 42 percentage points more likely to vote Democratic than those with less than a high school diploma (26% vs. 68%).

Fig. 5.

Percent of White voters voting for the Democratic presidential candidate at each education level, by election.

Finally, Fig. 6 shows the relationship between occupational class and Democratic support for each election. In earlier elections, the working class was more Democratic than the white-collar class. In 1980 this gap was quite large: The working class was 17 percentage points more Democratic than the white-collar class (49% vs. 32%). In most of the early elections, however, the gap was smaller: five percentage points in 1984, three percentage points in 1988, seven percentage points in 1992, five percentage points in 1996. In the 2000 election, for the first time, the white-collar class was more Democratic than the working class by two percentage points. In the 2004 election, this switched back (with the working class voting more Democratic by two percentage points). But since 2008, the white-collar class has voted Democratic in presidential elections at higher rates than the working class. This difference became particularly pronounced in the elections where Trump was on the ballot. The white-collar class voted more Democratic than the working class by a margin of 13 percentage points in the 2016 election (50% vs. 37%) and a margin of 18 percentage points in 2020 (54% vs. 36%). The relative position of the intermediate class is also worth noting. Other than the 1980 election, the intermediate class has always been the least Democratic of the three occupational classes. This may be because about half of White voters in Morgan and Lee’s (24) intermediate class are self-employed in nonprofessional jobs, many as small business owners, a group past research has found to be quite conservative (25).

Fig. 6.

Percent of White voters voting for the Democratic presidential candidate at each occupational class, by election.

Supplementary Analyses.

In SI Appendix we show that the descriptive bivariate associations between voting choice and each of income, education, and occupational class shown above are almost the same as net associations obtained from regression analyses controlling for region, age, and gender (SI Appendix, Figs. S1–S3). In figures showing associations with income, we control education; in figures showing associations with education, we control income; and in figures showing associations with occupational class, we control education. Associations change little when we add these additional controls.

Analysis in this manuscript focuses on White voters, but results are robust to the inclusion of voters of all races (SI Appendix, Figs. S4–S6). This is in part because non-White voters constitute a small enough portion of all voters that they have a negligible impact on aggregate patterns. We also show figures separately for White, Black, and Hispanic voters (SI Appendix, Figs. S7–S9); levels and patterns are often different by group. While there is a persistent gender gap in voting, the patterns of association between voting and the three measures of socioeconomic status are quite similar for men and women (SI Appendix, Figs. S10–S12).

SI Appendix, Fig. S13 shows that patterns are similar whether we use size-adjusted household income as above (which includes income from all household members from all sources) or personal earnings (which only includes income from the respondent’s occupation, and is missing for those who were not employed). Using a different occupational categorization, dividing occupations into professionals, managers, and others, we find, in SI Appendix, Fig. S14 that both managers and professionals moved toward the Democratic Party steeply and nearly continuously across this time period similar to the trend we observed for the broader category of all white-collar workers. By contrast, those workers outside professional and managerial occupations move up and down without a clear trend across much of the period, but their Democratic support has somewhat decreased since 2008, similar to what we found for the working-class category above (Fig. 3).

Where possible, we replicate our findings using the American National Election Studies (ANES) (SI Appendix, Figs. S15 and S16). For these comparisons of ANES and GSS findings, we consider a longer period going back as far as is possible in the GSS. The ANES only provides household income percentile as a five-category ordinal variable with the following levels: 0 to 16th percentile, 17th to 33rd percentile, 34th to 67th percentile, 68th to 95th percentile, and 96th to 100th percentile. We created these percentile categories with the GSS size-adjusted household income measure. SI Appendix, Fig. S15 reveals that the two datasets show similar trends by income group. Trends by educational group (SI Appendix, Fig. S16) also look markedly similar across the GSS and ANES. The ANES does not include information about respondents’ occupation for years after 2004, so the occupational class analysis we did with GSS cannot be replicated with ANES.

This manuscript has focused on voting, which is the outcome that determines the president. But partisan realignment influences other political outcomes as well. In SI Appendix, we examine associations between each of the three socioeconomic status measures (income, education, and occupational class) and party identification using both ANES and GSS data where possible (for income and education but not occupation) (SI Appendix, Figs. S17–S19). The seven-point party identification scale includes strong Democrat, not very strong Democrat, independent that leans Democrat, independent, independent that leans Republican, not very strong Republican, and strong Republican. We dichotomize this seven-category party identification scale into Democrat (the first three categories of the scale) and all other combined. To make the samples comparable, we limited our analysis of party identification to those who reported that they voted in the most recent election. These analyses show White voters at low and middle levels of socioeconomic status moving away from Democratic identification more strongly across the period than we found taking voting as the outcome (see also ref. 26). Among high socioeconomic status White voters, these figures show increased Democratic identification since the 1980s, but this growth is not as dramatic as the growth in Democratic voting among these same groups that our main analysis of voting showed. Future work should further explore the connection between party identification and the party for which individuals vote (including whether an individual votes) as well as which variable—identification or voting—is the leading or lagging indicator of partisan realignment.

Discussion

Like previous researchers, we find that the income gradient on voting has decreased over time such that income no longer has much if any power to predict vote choice. Meanwhile, the gradient on education has reversed and become quite steep with more educated people voting more Democratic.

What we add to the discussion is that these developments are largely driven by the movement of high status White voters toward voting for Democrats rather than from lower status White voters moving away from Democratic presidential candidates. In some, but not all, of our analyses, the movement of groups of higher status White voters toward the Democratic Party was particularly pronounced in the 2008 election, and this new, higher level of Democratic support has been sustained ever since.

Our findings, like those of others, show a movement of less privileged Americans toward voting Republican in presidential races, but it was mostly after 2012, not longer-term. Will data on the 2024 and future elections show that this realignment was short-lived and idiosyncratic to Trump’s appeal, or was it the start of a long-term shift toward working-class voters supporting Republicans? One study suggests the latter; since around 2008, there was an increase in the share of recent voters who call themselves Independents rather than either Republicans or Democrats, and this shift was disproportionate among those in working-class occupations (26).

The proportions of the population who hold college degrees and who work in white-collar occupations have grown, and we showed that the share of both these groups choosing Democratic presidential candidates has increased steadily for four decades, suggesting possible long-term problems for Republicans. On the other hand, lower status groups have historically low turnout rates, which means increased voter participation can increase their share of the electorate even in the face of population-level declines. Indeed, increased voter turnout has given Republican candidates advantages in recent years (27).

Looking at the large and sustained shifts among higher status Whites compared to the smaller, shorter-term shifts among lower status Whites, one might ask why the latter has received so much more attention in recent years than the former. Here, we can only speculate, but we suggest the possibility that the people who conduct research on elections and consume political analysis are disproportionately members of the higher status groups who have moved toward the Democratic Party in recent decades. Choices of groups to which they belong seem unremarkable to them so they do not notice the change; they assume the change must have come from a working class moving right, failing to notice that it is educational, income, and occupational class groups that they belong to that are moving most systematically.

An important and often overlooked first step in work related to partisan realignment is to establish which groups of voters have moved. This work aims to clarify these developments, but leaves the task of explaining these changes to future researchers. Results here offer guidance for those efforts: Future research should not assume there has been a long-term shift of the White working class away from Democrats, but rather interrogate whether the short-term movement since Trump has been on the ballot marks the beginning of a new, long-term trend, and seek to understand why those in higher educational and occupational groups have seen such a large and long-term movement toward Democrats.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All data used here are publicly available at the GSS website: gss.norc.org/ (21). Code for data analysis is archived on Open Science Framework (https://osf.io/f3jck/) (28). Previously published data were used for this work (21).