States of Change

For previous States of Change reports, please click here.

Demographics are not destiny, but steady and predictable changes to the electorate play an important role in defining the landscape of American politics. Most demographic groups have a political lean, so a group increasing or decreasing in size over time will tend to benefit one party or type of politics over another. The most well-known example is the growth of the nonwhite population in the United States, which—since nonwhites tend to lean heavily Democratic—is typically viewed as tilting the electoral terrain somewhat toward the Democrats over time as well as increasing the weight of nonwhite voters within the Democratic Party over time. But other changes are important, such as the decline of noncollege educated voters, particularly whites; the aging of the adult population; and the rise of new generations to replace older ones.

In this report, we will explore the effect of these changes on the demographic composition of the electorate and, especially, on the composition of the two major political parties. Reflecting the latter focus, this analysis will not focus on how many individuals from a given demographic group voted or will likely vote for the Democratic or Republican candidate in a particular election. Rather, it focuses on how many people who voted or are likely to vote for the Democratic or Republican candidate in a particular election belong to a given demographic group.

While the former tells us about the political leanings of a given group, the latter answers different questions. While size is not the sole determinant of a group’s influence within a party, it is a significant input and affects how parties formulate positions and present themselves to the electorate. As we head into the 2020 presidential primaries, we are bound to observe the effects of party composition on how candidates for the presidential nomination—especially Democratic candidates, due to their intense competition for voters—position themselves to garner primary votes from different demographic groups within their party.

Our investigation turns up a number of key findings that illuminate how significantly the compositions of the Democratic and Republican parties have changed over the years and are likely to change in the future. We show that the 2016 election was the most demographically divisive election in the past 36 years. The parties were more divided by age, race, and education than in any prior election in modern political history.

Reflecting these intensifying divisions, the parties were more compositionally different in 2016 than at any point in the prior 36 years. This election was the first presidential election white noncollege voters did not make up a plurality of both parties’ coalitions, with white college voters exceeding the share of white noncollege voters in the Democratic coalition.

Nonwhites will continue to grow as a share of both parties’ coalitions, especially Hispanics. We find that, by 2032, Hispanic voters will surpass black voters as the largest overall nonwhite voting group. And, by 2036, black voters will make up a larger share of the Democratic coalition than white noncollege voters.

On the other hand, we find that white voters will continue to decline through 2036 as a share of both the Republican and Democratic party coalitions, though this decline with be considerably quicker in fast-growing states such as Arizona and Texas that are already less white. White noncollege voters, in particular, are projected to decline rapidly as a share of both parties’ coalitions across all states through 2036, although the sharpest declines will, again, be in fast-growing states.

Generational changes will also be substantial. By 2036, Millennial and Generation Z voters—the two youngest generations—will be heavily represented in both the Democratic Party and Republican Party coalitions, while the influence of Baby Boomer and the Silent Generation voters—the two oldest generations—will radically decline. White Millennial and Generation Z voters, in particular, will develop a large presence in the Republican coalition and, combined with nonwhites, will give the GOP a new look in all states—even slow-growing ones such as Wisconsin and Ohio.

Finally, our data indicate that, while shifting turnout and support rates can be pivotal for winning elections, these changes are likely to have a relatively small impact on the overall makeup of the electorate and party coalitions in the future. Thus, most of the effect of demographic change on future party coalitions is already baked in and will reshape party coalitions—in a sense, whether these parties like it or not.

The Democratic Party and the Republican Party are driven by different ideologies, they propose different kinds of policies, and they run different types of candidates for office. They also win different types of voters—voters with disparate demographic profiles. For example, as a group, those who vote for Democrats today are more racially diverse than those who vote for Republicans.

But how different are the parties demographically, and how have those differences evolved over time?

For this report, we developed two compositional gaps to answer those questions. They are metrics that allow us to look for compositional differences between the parties on certain demographic dimensions and observe how those differences have changed over time. Specifically, we developed race-education gaps to study how different the parties’ coalitions are along racial and education lines and age gaps to study how different the parties’ coalitions are along age lines.

As an example, let’s walk through generating a race-education gap for the 1980 presidential election. We would start by finding out what percent of those voting for the Democratic and Republican candidates fell into each relevant demographic group. For example, 76 percent of those who voted for the Republican candidate in that year were white noncollege compared to just 60 percent of those voting for the Democratic candidate. The next step would be finding the absolute difference between the composition of the parties for each demographic group we are concerned with and adding up all those differences. If one were to do that for 1980 presidential race with those five race-education groups mentioned above, the gap between the parties would be 42 points. (See Table 1 for details)

Generating these gaps for every presidential election between 1980 and today makes one thing clear: 2016 was the most demographically divisive presidential election in modern American history. The parties’ coalitions were more dissimilar in terms of their racial, educational, and age composition in 2016 than at any point in the previous 36 years.

Along racial and education lines, the parties’ coalitions have always been different. Between 1980 and 2004, the race-education gap was essentially stable in the high 30s and low 40s. However, in 2008, this gap grew to 62 points, dropped slightly in 2012 to 59 points, and then peaked at 63 points in 2016—the highest recorded gap over this period. This post-2004 jump was driven by changes in the voting behavior of nonwhite voters; the shift toward the Democratic Party among all groups between 2004 and 2008; and a notable increase in black turnout that rapidly diversified the Democratic coalition.

In fact, in the absence of those shifts, the race-education gap between the coalitions actually would have shrunk by about 6 points. Because the majority of white Americans vote Republican, their shift toward the Democratic candidate between 2004 and 2008 was actually working to reduce the race-education gap between the parties’ coalitions.

By contrast, the parties’ coalitions were once extremely similar in terms of age composition and have grown more divided over time. In 1980, the age gap between the parties was just 3 points. While this gap has been steadily increasing since the late 1980s, the post-2004 period is once again a time of change. The three largest increases in the age gap over this 36-year period all occurred after 2004.

Assuming that turnout and support rates stay the same, the demographic shifts occurring nationally will do little to close the gaps between the parties. As a result of these changes, we would expect the race-education gap and age gap to shrink about 2 points by 2036 (61 points and 28 points, respectively).

White voters

In 1980, almost 9 in 10 (87 percent) voters were white, including about 7 in 10 (69 percent) with no college degree and 18 percent with a college degree. Even at this time, the parties already were quite different. While whites made up the overwhelming majority (95 percent) of Republican voters, they were less than three-quarters (74 percent) of Democratic voters. Compared to Republican voters, a lower share of Democratic voters was white noncollege (76 percent versus 60 percent) and white college (19 percent versus 14 percent) in 1980.

In the intervening 36 years, two important trends reshaped the American electorate. First, immigration from Asia and Central and South America slowly diversified the voting population. Second, the educational attainment rates of young Americans increased, resulting in a marked rise in the number of Americans with a college education.

As a result of that first trend, about three-quarters (74 percent) of voters in 2016 were white. Almost 9 in 10 (88 percent) Republican voters in that election were white—a 7-point decline that mirrored the changes occurring in the overall voting population. By contrast, just 6 in 10 (60 percent) Democratic voters were white, a shift twice as large as the one observed among Republican voters and voters overall.

As a result of both trends, the percent of voters who were white noncollege dropped a dramatic 25 points over 36 years. By 2016, this group made up just 44 percent of voters.

Throughout this period, white noncollege voters have consistently made up a larger share of Republican voters. The gap between Republican and Democratic voters for this group was smallest in 1992 (64 percent versus 54 percent) and 1996 (62 percent versus 51 percent) but never dipped below a 10-point difference during this period.

(The solid lines are observed figures for the composition of the overall voting electorate, of Republican presidential voters, and of Democratic presidential voters through 2016; the dotted lines are projections through 2036 using our baseline model.)

By 2016, white noncollege voters made up 6 in 10 (60 percent) Republican voters but just 29 percent of Democratic voters. While the compositional shift for both parties has been dramatic, it has not been equal. The 36-year decline among Democrats (60 percent versus 29 percent) is almost twice as large as the decline among Republicans (76 percent versus 60 percent).

The story for white college voters is quite different. Despite an overall decline in the number of voters whom were white, rising educational attainment and higher turnout rates caused the share of white college voters to rise by 12 points over this period. White college voters made up 30 percent of voters in 2016. Compared to 1980, the share of white college voters in 2016 was 17-points higher among Democrats (14 percent versus 31 percent) and 9-points higher among Republicans (19 percent versus 28 percent).

This makes 2016 unique in a number of ways. First, the 2016 presidential election represents one of the largest compositional gaps between Republican and Democratic voters over this time period. In 2016, the number of voters who were white noncollege was 31-points higher among Republicans voters than Democratic voters. This compositional gap between Republicans and Democrats was part of a trend over the last two elections, with the next two largest gaps occurring in 2008 (62 percent versus 40 percent) and 2012 (60 percent versus 33 percent).

Second, it is the first election where white noncollege voters did not make up a plurality of both parties’ coalitions. Among those who voted Democratic, the share of white college voters (31 percent) was slightly larger than the share of white noncollege voters (29 percent). Furthermore, it is the first election where white college voters made up a plurality of Democratic voters. These are both significant milestones in the long-term trajectory of white voters.

Third, while white college voters consistently made up a larger share of Republican voters than they did Democratic voters between 1980 and 2012, this flipped in 2016. White college voters made up a slightly larger share of Democratic voters (31 percent) than Republican voters (28 percent) in that election.

Going forward into the future, these trends will continue. Holding turnout rates and voting preferences constant, we expect white voters to make up 72 percent of the electorate by 2020 and just 64 percent by 2036. Giving current voting patterns, we would expect more than 8 in 10 Republican voters to be white in 2020 (86 percent) and 2036 (80 percent). Those same patterns would suggest that white voters will make up about 6 in 10 (58 percent) Democratic voters in 2020 and drop to 50 percent by 2036.

Once again, this drop is driven almost entirely by the decline of white noncollege voters. If current demographic trends persist, white noncollege voters should be 42 percent of voters in 2020 and 35 percent of voters by 2036. We expect white noncollege voters to be 58 percent of Republican voters and just over one-quarter (27 percent) of Democratic voters in 2020. Despite their dramatic decline, white noncollege voters will still make up the majority (51 percent) of Republican voters by 2036 but just about 1 in 5 (22 percent) of Democratic voters.

We do not expect the share of white college voters to change dramatically over the next five elections. They should continue to be 30 percent of voters in 2020 and shrink slightly to 29 percent of voters by 2036. Similarly, by 2020 and 2036, the share of Democratic voters (31 percent versus 28 percent) and Republican voters (28 percent versus 29 percent) that are white college graduates will change only slightly. Notably, white college voters will make up the plurality of Democratic voters over this entire 16-year period.

Black voters

In 1980, about 1 in 10 (9 percent) voters were black. The relatively lopsided nature of black support for Democratic candidates created a large compositional gap between the parties. While black voters made up 1 in 5 (20 percent) Democratic voters in 1980, they were just 1 percent of the Republican coalition.

Black voters are, far and away, the most stable coalitional element in our study. Due to relatively low immigration and modest birth rates, black eligible voters have only grown at about the same pace as all voters. As a result, the rough size of this group has not changed significantly over time. Black voters have grown just slightly as a share of all voters (12 percent), Democratic voters (22 percent), and Republican voters (2 percent).

Current demographic trends suggest that this group will only grow marginally going into the future. Absent any shifts in turnout, black voters should make up an identical portion of voters, Democratic voters, and Republican voters in 2020. However, by 2036, black voters will make up 13 percent of all voters, 23 percent of Democratic voters, and 3 percent of Republican voters.

These are small changes relative to 2016, but this stability will result in black voters surpassing white noncollege voters by 2036 as the second largest voting block within the Democratic Party (23 percent versus 22 percent). This would be another milestone event in the demographic evolution of electorate.

Hispanic voters

In 1980, Hispanic voters made up just 3 percent of all voters. While the compositional differences between Democratic (4 percent) and Republican (2 percent) voters were not large in absolute terms, the gap was representative of the divide observed in later elections.

As a result of significant immigration in the intervening years and a younger age distribution, Hispanic voters have grown slowly and steadily as a share of all voters. By 2016, they made up almost 1 in 10 (9 percent) of all voters.

Throughout this 36-year period, Hispanic voters have consistently been a larger share of Democratic voters than Republican voters. While Democratic and Republican voters were about equally Hispanic in 2000 (6 percent versus 5 percent) and 2004 (7 percent versus 5 percent), this compositional gap grew after 2008. By 2016, the size of the Hispanic group within the Democratic Party was twice the size of the group within the Republican Party (12 percent versus 6 percent).

Going forward into the future, the size of the Hispanic population will continue to grow. With the majority of the growth among Hispanics coming from native births rather than immigration, Hispanics will continue to make up a larger and larger share of both eligible voters and voters overall.

By 2020, Hispanic voters will make up a slightly larger share of all voters (10 percent) and Democratic voters (14 percent) and a roughly equal share of Republican voters (6 percent). Even if we assume that this group’s turnout does not increase, Hispanic voters will surpass black voters as the largest nonwhite voting group by 2032. By 2036, Hispanic voters should make up about 14 percent of all voters, almost 2 in 10 (18 percent) Democratic voters, and 1 in 10 (10 percent) Republican voters.

Asian and other race voters

In 1980, just 1 percent of voters were Asians or belonged to another racial group. These voters made up an equal share of Democratic (1 percent) and Republican (1 percent) voters.

Due to rising immigration from Asian countries, Asian Americans have become the fastest growing racial group. As a result, Asian voters and those belonging to other racial groups made up 6 percent of all voters in 2016—including 7 percent of Democratic voters and 4 percent of Republican voters.

Going forward, current demographic trends predict slow but steady growth. By 2020, Asian voters and those belonging to another racial groups will together make up a very similar number of all voters (6 percent), Democratic voters (7 percent), and Republicans voters (5 percent). These figures will rise slightly by 2036 (8 percent, 9 percent, and 7 percent, respectively).

Voters by age

In 1980, just more than half of voters were under the age of 45, including 23 percent who were ages 18 to 29 and 28 percent who were ages 30 to 44. About one-third (32 percent) of voters were ages 45 to 64 and just fewer than 1 in 5 (17 percent) voters were ages 65 and older. At this time, the Republican and Democratic parties were very similar in terms of their age composition, with roughly equal numbers of 18- to 29-year-old voters (22 percent versus 21 percent), 30- to 44-year-old voters (28 percent versus 27 percent), 45- to 64-year-old voters (32 percent versus 33 percent), and voters 65 and older (18 percent versus 18 percent).

Over the next 36 years, two significant changes occurred. First, the number of voters younger than 45 shrank while the number of voters 45 and over grew. Second, the parties’ once similar age compositions grew apart.

By 2016, the number of voters ages 18 to 29 dropped 7 points (16 percent). While the number of young voters among Democratic and Republican voters was relatively similar prior to 2004, that election marked the first significant divide between the parties (19 percent versus 13 percent). By 2016, about 2 in 10 (19 percent) Democratic voters and about 1 in 10 (11 percent) Republican voters were ages 18 to 29.

While the number of 30- to 44-year-old voters was relatively stable between 1980 and 2004, it dropped about 6 points by 2016 (22 percent). While the number of voters ages 30 to 44 among Democratic and Republican voters was relatively similar prior to 2000, there was a larger difference that year (28 percent versus 33 percent). This Republican advantage held in 2004 (25 percent versus 30 percent) but flipped in 2008 (26 percent versus 23 percent) and then expanded in 2012 (26 percent versus 20 percent). By 2016, 30- to 44-year-old voters made up one-quarter (25 percent) of Democratic voters but just fewer than 1 in 5 (19 percent) Republican voters.

(The solid lines are observed figures for the composition of the overall voting electorate, of Republican presidential voters, and of Democratic presidential voters through 2016; the dotted lines are projections through 2036 using our baseline model.)

While the number of 45- to 64-year-old voters was relatively stable between 1980 and 1992, this group grew in size afterward—peaking in size in 2012 when 4 in 10 (39 percent) voters fell into this age range. In 2016, the percentage of voters in this group dropped slightly (37 percent), but still represented the largest age group in the electorate. While 45- to 64-year-old voters made up roughly equal shares of Democratic and Republican voters between 1980 and 2004, the parties began to separate in 2008 (37 percent versus 40 percent). By 2016, 45- to 64-year-old voters made up more than 4 in 10 (41 percent) Republican voters but just 35 percent of Democratic voters.

By 2016, the share of voters 65 and older grew by 7 points (24 percent). While the number of seniors in the party coalitions was relatively similar prior to 2008, they made up a notably larger share of Republican voters that year (22 percent versus 17 percent). By 2016, the gap between the Republicans and Democrats had grown even larger (30 percent versus 21 percent).

Overall, the 36-year period between 1980 and 2016 saw those younger than 45 shrink as a share of the electorate and begin to make up a larger share of Democratic voters. Consequently, those 45 and older made up a larger share of voters and became a disproportionate percentage of Republican voters.

By 2020 and 2036, the number of 18- to 29-year-old voters will shrink slightly (15 percent and 14 percent, respectively); 30- to 44-year-old voters will remain stable (23 percent and 23 percent, respectively); 45- to 64-year-old voters will shrink (35 percent and 32 percent, respectively); and those 65 and older will grow (26 percent and 32 percent, respectively). Notably, while 45- to 64-year-old voters will have constituted the plurality of voters between 1980 and 2032, growth among voters age 65 and older will finally make these two groups the same size in 2036.

The compositional differences between the parties will remain roughly proportional during this time period. While 18- to 29-year-old voters will make up fewer than 1 in 5 Democratic voters in 2020 and 2036 (19 percent and 17 percent), they will be just about 1 in 10 Republican voters (11 percent versus 9 percent). Similarly, 30- to 44-year-old voters will make up about one-quarter of Democrats in these two elections (26 percent and 25 percent) and just fewer than 2 in 10 Republican voters (19 percent and 19 percent).

By contrast, older Americans will continue to make up a disproportionate share of Republicans voters. While 45- to 64-year-old voters will make up almost 4 in 10 (39 percent) Republican voters in 2020 and more than one-third (34 percent) in 2036, they will be just 33 percent and 30 percent of Democratic voters in those same years.

Voters age 65 and older will continue to make up a disproportionate share of Republican voters but grow as a share of both parties. By 2020, they will make up about one-third (32 percent) of Republican voters. Astoundingly, seniors (36 percent) will overtake 45- to 64-year-old voters (34 percent) and become the plurality of the Republican coalition by 2028. By 2036, they will constitute 38 percent of the Republican voting coalition. Among Democratic voters, those age 65 and older will make up almost one-quarter of the coalition by 2020 (23 percent) and nearly 3 in 10 (28 percent) by 2036.

As noted earlier, we examined projected changes in party composition not just using our baseline model but also with significant changes in turnout and party support rates among different groups. What these data show is that, while shifting turnout and support rates can be pivotal for winning elections, these changes should have a relatively small impact on the overall makeup of the electorate and the parties’ coalitions. This indicates that most of the effect of demographic change on future party coalitions is already baked in and will reshape party coalitions—in a sense, whether these parties like it or not.

Equalized turnout across racial groups

In the modern political era, the turnout rates of blacks, Hispanics, Asians, and those belonging to other racial groups has typically lagged behind that of whites. What would the electorate and the parties look like if those gaps were closed?

In order to determine this, we set the turnout rates of various subgroups equal to the highest turnout observed across racial groups. So, for example, suppose that among Californians ages 45 to 64 with a college degree the highest turnout observed was among whites. For this simulation, we would set the turnout rate of everyone who was Californian ages 45 to 64 with a college degree equal to that of whites. We believe this to be a more realistic manner of equalizing turnout compared to naively setting turnout rates between racial groups equal to each other. It allows important demographic features, such as the age and educational composition of racial groups, to play a role rather than be ignored.

Under this scenario, we would expect the share of nonblack voters of color to increase in 2020 while the share of white voters declined. Relative to our baseline estimates—which were derived using 2016 turnout rates—the percent of voters who would be Hispanic (10 percent baseline scenario versus 12 percent equalized turnout scenario) and Asian or belonging to another racial group (6 percent versus 8 percent) in 2020 would increase. As a result, the share of voters who are white college (30 percent versus 28 percent) and white noncollege (42 versus 40 percent) would shrink by the same amount. The percent of voters who are black under this scenario would not be different than what we would expect under normal turnout rates (12 percent). Because this simulation does not alter the rates at which these groups vote for the Democratic Party or Republican Party, these overall changes are roughly translated for the coalitions, with similar increases and declines among bother parties for each group.

A similar effect would occur for every election between 2020 and 2036. While the electorate would continue to diversify as a result of demographic change, equalizing turnout would increase the share of voters who were Hispanic, Asian, and those belonging to another nonblack racial group by about 4 points while shrinking the share of voters who are white by the same amount.

Shifting support rates

Beyond turnout, we can also simulate what would happen if various groups started voting for Democrats and Republicans at different rates. How much would shifts in support rates alter the composition of the parties?

It turns out that even shifts that we imagine to be relatively large—say the shift among white noncollege voters toward the Republican Party between 2012 and 2016—does little to affect the overall composition of the parties’ coalitions.

Suppose that Hispanic, Asian, and other racial groups were to swing about 15 margin points toward the Democratic Party in 2020.¹ The share of Democrats that belong to these groups would only increase by 2 points (21 percent versus 23 percent). A similarly sized shift toward the Republican Party would only increase their share by 3 points in the Republican coalition (11 percent versus 14 percent).

Alternatively, suppose white noncollege voters shifted toward the Republican Party by about 10 points—about twice the size of the shift we saw for this group between 2012 and 2016. The percent of Republican voters who would be white noncollege in 2020 would only increase by 2 points (58 percent versus 60 percent). A similarly sized shift toward the Democratic Party would result in a similarly sized 3-point increase in the share among Democratic voters (27 percent versus 30 percent).

Thus, while there will be some modest variability in the magnitude of demographic change in the Democrats’ and Republicans’ party coalitions, the basic contours of these changes are already set by the likely demographic evolution of the country.

The national demographic coalitions of Republican and Democratic voters take somewhat different forms across the 50 states. This is because states differ in their underlying demographic profiles and because voters in specific demographic groups, within a state, do not exactly follow national Republican and Democratic voter turnout and candidate preference patterns.²

In this section, we take a deep dive into how Republican and Democratic coalitions differ across key swing states with respect to various attributes: race, age, education, and generation. We look at these coalitions in 2016 and sketch how they are likely to change in the future through the 2036 election. Understanding these recent and projected coalitions at the state level are particularly important because of the critical role that states play in presidential elections—not to mention other aspects of the nation’s politics.

Of course, these swing states are not necessarily representative of all parts of the country. Moreover, the demographic compositions of nonswing states do not necessarily reveal the state’s voting preferences. Many of the Republican-leaning Great Plains states, such as North and South Dakota, Nebraska, and Missouri, are heavily white. Yet Vermont, Maine, and Minnesota are also heavily white but voted Democratic in recent presidential elections because their white voters tend to be more educated and have different attitudinal profiles. While the discussion below focuses on states that are likely to be swing states in 2020 and several future elections, the 2020 and projected 2036 demographic profiles for Republican and Democratic voters for all 50 states are presented in Appendix 2.

Swing state party coalitions by race

As we noted earlier, for the nation as a whole, the racial breakdowns of the Republican and Democratic coalitions for the United States show that the Republicans had a far higher share of white voters than the Democrats in 2016. This broad pattern of higher white representation among Republicans occurred for all of the major swing states, though with quite a bit of variation. Figure 4 depicts the white shares of 2016 Republican and Democratic voters in 14 swing states—six in the Northeast and Midwest (New Hampshire, Iowa, Michigan, Ohio, Pennsylvania, and Wisconsin) and eight in the nation’s South and West (Florida, Georgia, North Carolina, Texas, Virginia, Arizona, Colorado, and Nevada). The states are arranged in descending order by the shares of whites in their Republican coalitions—beginning with Iowa, where 96 percent of Republican voters were white, and ending with Nevada, where the Republican white share was 75 percent.

The disparities between Republican and Democratic white shares are evident but vary sharply across states and not always in proportion to the whiteness of the Republican constituencies. In the white states of Iowa and New Hampshire, Democratic coalitions are almost as white as those of Republicans. At the other extreme is Georgia, where there is a 54 percent gap between the white share of the state’s Republican coalition (at 87 percent) and its Democratic coalition (33 percent).

Just as noteworthy is the fact that these diversity disparities continue as the nation as a whole becomes more diverse and, as the projections assume, a dispersal of Hispanics and Asians into the middle of the country continues, along with the migration of blacks into the South.³ The projected 2036 state Republican and Democratic constituencies—shown in Figure 4—reveal the same patterns of Republican-Democratic white share disparities, though with uniformly lower white shares than in 2016. An extreme example is Nevada, where the white share of Republican voters drops from 75 percent in 2016 to 61 percent in 2036. At the same time, the white share of Democratic voters falls from 56 percent to 41 percent, thus keeping the white disparity between the two party’s coalitions almost the same.

It is significant to note that some of the larger white share disparities between Republican and Democratic voters occur in states with substantial or dominant black populations. Georgia and North Carolina display the largest such disparities among Southern states, as is the case for Ohio, Pennsylvania, and Michigan in the North. Even Texas, with a larger Hispanic than black population, exhibits a substantial Republican-Democratic white disparity because its black population is sizeable.

This is because the 2016 African American Democratic-Republican vote margins, which are projected ahead to 2036, are far larger than for any other racial group—both nationally and for these states.⁴ When nearly 9 of 10 blacks vote Democratic, this heavily skews the demographic makeup of the Democratic and Republican voting coalitions of a state.

This is illustrated in Figure 5, which depicts Georgia’s 2016 and projected 2036 Democratic and Republican coalitions. Note that among 2016 voters in Georgia, Democratic candidate Hillary Clinton received 86 percent of the votes of its substantial black voting population but only 24 percent among whites. Unsurprisingly, blacks comprised around 60 percent of Georgia’s Democratic voters but just 6 percent of Republican voters. Those unbalanced shares of black voters between the parties are projected to continue through 2036.

A different scenario occurs in Texas, where Hispanics are the dominant nonwhite group. In the 2016 election, 60 percent of Hispanics voted Democratic compared with 88 percent of blacks and just 28 percent of whites. When applied to the diverse demography of the Lone Star state, these voting preferences produced a 2016 Democratic coalition that was 29 percent Hispanic and 27 percent black. In contrast, Hispanics make up only 14 percent of Republican voters, with blacks making up just 2 percent.

By 2036, if 2016 voter preferences and turnout rates are projected ahead, each coalition is less white, but the racial profiles of the Democratic and Republican coalitions are still dissimilar. The Hispanic share of Republican voters increases from 14 percent to 22 percent, but the Latino share of Democratic voters rises by 10 points, from 29 percent to 39 percent. The black share remains extremely unbalanced: 3 percent of Republican voters compared with 26 percent of Democrats. However, the small Asian and other race population rises slightly more in the Republican coalition than for Democrats.

The 2036 projections shown above assume that 2016 voting preferences of different racial groups will continue. What would happen to Texas’s party coalitions if it were assumed that Hispanic, Asian, and those of other races voting preferences swung significantly to the Republicans? Or what if those same preferences swung the same amount toward the Democrats? Figure 6 shows the outcomes of these different scenarios as projection Alternative 1 (assuming Hispanic and Asian/other race voting preferences swing 7.5 points toward the Republicans and 7.5 points away from the Democrats, for a total margin gain of 15 points for the GOP) and projection Alternative 2 (assuming the reverse pattern for a 15-point margin gain for the Democrats among these groups).

The results show that, under these two scenarios, there are noticeable shifts in each coalition. In particular, the combined nonwhite share of the 2036 Republican coalition increases to 36 percent under Alternative 1 but drops to 28 percent under Alternative 2, with greater and lesser contributions by Hispanics and those of Asian and other races. Likewise, the white share of the Democratic coalition varies between 31 percent in Alternative 1 and 28 percent in Alternative 2. Yet, even when assuming these fairly sizeable voting preference shifts among Hispanic and Asian and other race voters, there remain sharp diversity differences in the Republican and Democratic coalitions in the projections for 2036. This is also the case when similar alternative projections were conducted for other states with large Hispanic populations.

Swing state party coalitions by age

As illustrated earlier for the nation as a whole, party coalitions by age have shown sharper variations in recent elections with older age groups trending toward Republican candidates and younger age groups trending toward Democrats. We also showed that, while the senior shares of both party coalitions increase through 2036 as the country continues to age, the age divide between the parties continues and becomes quite distinct among seniors, who would comprise fully 38 percent of Republican voters, compared with 28 percent of Democratic voters.

This senior voter disparity between Republican and Democratic voter coalitions was evident for each key swing state in 2016. Figure 7 displays senior share comparisons for these states sorted in descending order by the senior share sizes of each state’s Republican voters. Florida and Arizona, two long-standing retiree magnet states, lead all others—both with the sizes of their senior shares of Republican voters and in terms of their Republican-Democratic disparities in senior shares. In 2016, the senior shares of Republicans in both states were more than 10-points higher than the senior shares of Democrats. Michigan and New Hampshire exhibit the smallest senior share disparities. This is partly because Michigan’s youthful voters voted less Democratic than national voters and New Hampshire’s older voters voted less Republican than national voters. These state variations in senior party disparities are evident in 2036 projections, which assume the same age-related voting and turnout patterns, though the coalitions of both parties show higher shares of seniors than in 2016.

One reason why some states hold greater age disparities between Republican and Democratic coalitions is due to the racial makeup of younger and older voters. For several states, such as Arizona, Florida, and Texas, the younger voting population is decidedly more racially diverse than those of older ages. Because Hispanics, blacks, and other nonwhites are more likely to vote Democratic, their strong presence among younger voters makes the Democratic coalition more youthful in highly diverse states.

In Arizona, for example, the nonwhite shares of voters in 2016 varied by age: 43 percent nonwhite among voters aged 18 to 29; 34 percent among voters aged 30 to 44; 25 percent among those aged 45 to 64; and just 14 percent among senior voters. The different racial profiles for these age groups tend to bolster the youthfulness of Arizona’s Democratic coalition and the large senior share of its Republican coalition. When age- and race-related voter preferences are projected to continue, Arizona’s year 2036 coalitions show an even sharper old-young disparity. (See Figure 8)

Swing state party coalitions by education

We have already reviewed the projected change in the racial composition of the Democratic and Republican parties in various states as demographic change continues to unfold, both if 2016 voting and turnout patterns continue in the future and if they change in subsequent elections. The white share of voters in each party’s coalition is projected to decline in the future under all scenarios, with declines sharpest in less white, faster-growing states such as Arizona, Florida, Nevada, and Texas—all of which have grown more than 12 percent in population this decade—and more moderate in whiter, slower-growing states such as Michigan, Ohio, Pennsylvania, and Wisconsin, all of which have grown under 3 percent in the decade. This means, for example, that the overwhelmingly white GOP party coalitions in these slow-growing states are likely to remain largely so, with some minor variations, through a number of future election cycles.

But, of course, all white voters are not the same, particularly when it comes to having and not having a four-year degree. This is currently a central divide in U.S. politics and has only been strengthening over time. Therefore, we need to look not only at shifts in the white share of voters in each party’s state coalition but also at shifts in the shares of both white college and white noncollege voters. Here we see some quite large and significant differences.

First, the Republican and Democratic coalitions start from very different places. As noted earlier, white noncollege voters dominate the Republican coalition nationally. In the 2016 election, three-fifths of Republican voters nationally were white noncollege, compared to just 28 percent who were white college. Put another way, 68 percent of the GOP’s white voters were noncollege. In contrast, 29 percent of Democratic voters in 2016 were white noncollege, actually slightly less than the number of white college voters (31 percent).

We see similar relative patterns on a state-by-state basis, though levels, particularly of white noncollege, vary considerably by whether the state is less white/fast growing versus whiter/slow growing. Pennsylvania’s Republican voters in 2016 were 67 percent white noncollege and 28 percent white college. In Wisconsin, the analogous figures were 70 percent white noncollege and 26 percent white college, while in Ohio, it was 68 percent white noncollege and 27 percent white college.

But in less white and faster-growing states, we see somewhat different results: Arizona GOP voters in 2016 were 54 percent white noncollege and 29 percent white college; Nevada Republican voters were 53 percent white noncollege and 22 percent white college. And in Texas, the white noncollege GOP voter share was just 50 percent compared to 29 percent white college.

Among Democratic voters on a state by state basis, we see white noncollege and college shares that are much closer to one another than among Republican voters. But there are some other important variations by state. In less white, fast-growing states, the white noncollege share of Democrats tends to be relatively low—from 32 percent in Arizona and 30 percent in Florida down to just 16 percent in Texas and 14 percent in Georgia, reflecting both populations that are less white noncollege and low Democratic support rates. But in the whiter, slow-growing states, where white noncollege voters dominate, we see white noncollege shares of Democratic voters as high as 52 percent in Iowa, 47 percent in Wisconsin, and 42 percent in Michigan. In these and similar states, white noncollege voters in the Democratic coalition tend to outnumber their white college counterparts by significant margins.

These are the 2016 numbers. But, of course, they will not remain stable over time—even if voting and turnout patterns should somehow remain the same—due to continuing demographic shifts in the underlying electorates of the states. We have already seen that white voter shares in party coalitions will decline between 2016 and 2036 across all states, though much less in some states (4 points in GOP coalitions in Iowa, Michigan, Pennsylvania, and Wisconsin as well as 3 points in Ohio) than in others (14 points in Nevada’s Republican coalition, 12 points in Texas, and 10 points in Arizona).

But when we look at white noncollege shares among Republican voters, we see large declines across all states. While white noncollege shares are projected to fall the most in less white, fast-growing states such as Nevada (-12 points) and Texas (-11), there should also be sharp declines in whiter, slow-growing states such as Pennsylvania and Wisconsin (-9 in each). The reason for this is that the general pattern in these slow-growing states is for the decline in the white noncollege share to be partially balanced by an increase in the white college share. For example, in both Pennsylvania and Wisconsin, the projected drop of 9 points in the GOP coalition white noncollege share is balanced by a projected increase of 5 points in the white college share. The fast-growing states, in contrast, tend to see very little change over the 2016–2036 period in their projected white college share of Republican voters.

Among Democratic voters, we see a similar, though more muted, pattern. White noncollege declines are the sharpest in fast-growing states, but there are substantial declines in slow-growing states as well, partially balanced by upticks in white college shares.

The bottom line here is that white noncollege shares will decline significantly over time in both the Republican and Democratic coalitions and in both fast-growing and slow-growing states. Combined with the tendency for white college shares to rise slightly over time or, at worst, decline only slightly, this will bring shares of white noncollege and college closer together in both parties’ coalitions. Among Democratic voters, by 2036, white college voters will be roughly equal to white noncollege voters in even slow-growing states.

Among Republican voters, these trends will mean that, by 2036, white college voters and nonwhites will outnumber white noncollege voters in many states: 56 percent to 44 percent in Arizona; 53 percent to 47 percent in Florida; 59 percent to 41 percent in Nevada; and 61 percent to 39 percent in Texas. And even in slow-growing states, roughly 4 in 10 GOP voters are projected to be white college graduates or nonwhites, considerably reducing the overwhelming dominance of the white noncollege group.

The projections just discussed simply combine anticipated demographic change with the voting preferences and turnout patterns from 2016. Scenarios that assume different voting and turnout behaviors produce somewhat different results for the respective party coalitions, though the basic contours of change remain roughly similar.

Looking at Republican voters, several scenarios could further depress the white noncollege share. One is turnout equalization: If groups move toward equal turnout within state and age groups, our estimates indicate that a number of fast-growing states would see an additional several points of decline in the white noncollege share. For example, in Texas, where Republican white noncollege voters were already projected to decline from 50 percent to 39 percent of the GOP total from 2016 to 2036, turnout equalization would push that number down further to 35 percent. The analogous figures for Arizona are 54 percent down to 44 percent and then further to 41 percent. Slow-growing states are less affected by this projected change.

A similar pattern of additional white noncollege decline among Republican voters can be observed with two other scenarios: where Hispanics, Asians, and those of other races swing toward the GOP compared to their 2016 preferences and where white noncollege voters swing toward the Democrats. On the other hand, when these two scenarios are reversed—Hispanics, Asians, and those of other races swing toward the Democrats or white noncollege voters swing toward the GOP—the decline in white noncollege voters in the GOP coalition is mitigated. In Florida, white noncollege GOP voters decline to a 49 percent share instead of a 47 percent share, and in Michigan, the white noncollege share decreases to 58 percent or 59 percent instead of 57 percent.

Among Democratic voters, we see the same pattern of results for turnout equalization: The white noncollege share of Democrats declines a bit more under this scenario than in a standard scenario that assumes no change in voting preferences or turnout behaviors. But for the other scenarios, the results are exactly reversed. Where white noncollege GOP voters were projected to decline more under a scenario (Hispanics, Asians, and those of other races swing toward the GOP or white noncollege voters swing toward the Democrats), white noncollege Democratic voters decline less. And where white noncollege GOP voters were projected to decline less under a scenario (Hispanics, Asians, and those of other races swing toward the Democrats or white noncollege voters swing toward the GOP), white noncollege Democratic voters decline more.

Swing state party coalitions by generation

The changes we have described over the 2016–2036 time period for the Republican and Democratic coalitions are substantial. But so far, we have not looked at one factor that will produce even larger changes in party coalitions: the generational evolution of the electorate. This will dramatically change the composition of the parties over time, even if voting patterns by age do not change over time. Of course, that is unlikely to be the case, and in the next year, we will investigate how the electorate and electoral outcomes may be affected by the tendency of generational cohorts to hold their political and policy preferences as they age.

But for now, we will simply look at the level of change we might expect in parties’ coalitions—even if voting patterns by age remain stable—as Generation Z fully enters the electorate, Millennials enter their high turnout years, Boomers dominate the senior population, and the Silent Generation passes from the political scene. These changes are very large in a relatively short period of time and are potentially of great political significance, since Millennials and Generation Z are by far the most liberal generations, and the Boomers and Silents are by far the most conservative.

Generation X is intermediate between these two generational clusters, both in terms of timing and political leanings. We do not discuss them here (though the data are displayed in the relevant charts) since we wanted to focus on the replacement of the oldest, most conservative generations with the youngest, most liberal generations. It is worth noting, however, that a very significant chunk of Generation X tends to be quite liberal, which may further enhance the effects of the generational replacements discussed here. This possibility will be investigated in next year’s report on the projected political effects of generational replacement.

Looking briefly at the national level, in 2016, the GOP coalition was 19 percent Millennials and Generation Z and 56 percent Boomers and Silents. Flash forward to 2036, holding voting and turnout patterns constant, we would expect the Republican coalition to be 47 percent Millennials and Generation Z and just 22 percent Boomers and Silents. For the Democrats, the analogous figures are 30 percent to 44 percent in 2016 and 59 percent to 15 percent in 2036. These are massive changes, especially given the significantly more liberal cast of the Millennial and Generation Z generations when compared to the oldest cohorts.

Naturally, these trends will affect the composition of white voters in each party, as older whites, who tend to be much more conservative, are replaced by more liberal Millennial and Generation Z whites. This may be of particular significance for the GOP, where conservative older whites have tended to set the tone for the party. In 2016, the Republican coalition was 16 percent Millennial/Generation Z whites and 50 percent Boomer/Silent whites. By 2036, we project the Republican coalition will be 36 percent Millennial/Generation Z whites and 18 percent Boomer/Silent whites, an exact 2 to 1 ratio from around 1 to 3 today. Interestingly, since we expect that the GOP coalition will be about 20 percent nonwhite at that point, these voters as well as the Millennial/Generation Z whites should constitute an absolute (56 percent) majority of the party.

For Democrats, the analogous figures are 16 percent Millennial/Generation Z whites and 28 percent Boomer/Silent whites in 2016 versus 27 percent Millennial/Generation Z whites and 9 percent Boomer/Silent whites in 2036, so the trends are similar, though the levels are far lower due to the lower percentage of whites in the Democratic coalition.

Turning to swing states, we see the same large effects from generational replacement, whether we look at the less white, faster-growing states or the whiter, slower-growing states. Consider Arizona: In 2016, the GOP coalition was 17 percent Millennials/Generation Z and 60 percent Boomers/Silent. By 2036, the figures should be 42 percent Millennials/Generation Z versus 25 percent Boomers/Silent. For Democrats, the analogous figures are 29 percent Millennial/Generation Z and 45 percent Boomers/Silent in 2016 versus 57 percent Millennial/Generation Z and 16 percent Boomers/Silent in 2036.

The composition of white voters shows the same trends. Arizona Republican voters in 2016 were 12 percent Millennial/Generation Z whites and 52 percent Boomer/Silent whites, which moves to 27 percent and 20 percent, respectively, by 2036. Combined with the expected 27 percent nonwhite voters, that suggests a total of 54 percent of Arizona GOP voters in that year being nonwhites or Millennial/Generation Z whites.

For Democrats, the corresponding figures are 15 percent Millennial/Generation Z whites and 33 percent Boomer/Silent whites in 2016 versus 24 percent Millennial/Generation Z whites and 11 percent Boomer/Silent whites in 2036, with fully half of Democratic voters being nonwhites.

Looking at the classic Rustbelt state of Michigan, 18 percent of GOP voters in 2016 were Millennials/Generation Z compared to 57 percent who were Boomers/Silents. By 2036, Republican voters in that state should be 48 percent Millennials/Generation Z and just 21 percent Boomers/Silents. The analogous figures for Democrats are 28 percent Millennial/Generation Z and 46 percent Boomers/Silent in 2016 versus 56 percent Millennial/Generation Z and 17 percent Boomers/Silent in 2036.

Looking at white voters, especially salient in a white state such as Michigan, we find GOP voters in 2016 were 17 percent Millennial/Generation Z whites in 2016 and 54 percent Boomer/Silent whites. By 2036, Boomer/Silent whites should fade to just 19 percent of Republican voters, while 41 percent will be Millennial/Generation Z whites. Combined with the projected 11 percent nonwhite voters in the GOP in that year, that should make 52 percent of Republican voters either nonwhites or Millennial/Generation Z whites.

For Democrats, we expect their coalition to go from 18 percent Millennial/Generation Z whites and 33 percent Boomer/Silent whites in 2016 versus 34 percent Millennial/Generation Z whites and 12 percent Boomer/Silent whites in 2036, with 37 percent nonwhite voters.

We see these patterns again and again across all manner of swing states. The generational replacement dynamic will clearly have tremendous effects on both parties, perhaps even more than the trends by race and educational attainment. Next year, we will investigate this dynamic in detail using updated demographic projections for all 50 states and analysis that take the full range of possible generational effects into account.

We have traced the effect of demographic change on presidential election party coalitions back in time to 1980 and projected forward under several different scenarios to the presidential election of 2036. We found that demographic change has profoundly reshaped the support bases of both the Republican and Democratic parties since 1980—especially in terms of race and educational levels—and should continue to do so in the future, despite possible changes in turnout and voting behavior by demographic group.

Demographic change by age levels has also had important effects on party coalitions and may well have even greater effects in the future, given the ongoing greying of the American population. Perhaps even more profound than that, the process of generation replacement—the rise of Millennials and Generation Z as well as the decline of the Silent and Baby Boomer generations—will radically shift the generational composition of both parties. We will investigate the possible political effects of this change in next year’s report.

We also assessed the possible future effects of demographic change on party coalitions at the state level. We found that, just as states today differ widely in their demographic compositions—both overall and among party supporters in the various states—so too are they likely to differ in how rapidly state party coalitions change as demographic shifts continue. By and large, less white, fast-growing states such as Arizona and Texas will see party coalitions evolve fairly quickly, whereas whiter, slow-growing states such as Wisconsin and Ohio will see a considerably more sedate rate of change. But even these latter states will be significantly affected as both Democratic and Republican party coalitions become less white noncollege, more nonwhite, and more based in the Millennial and Generation Z generations.

Robert Griffin is the research director, and a participating author, for the Democracy Fund Voter Study Group. He also serves on the editorial committee of PS: Political Science and Politics, the journal of record for the American Political Science Association.

Before joining the Voter Study Group, Robert was the associate director of research at the Public Religion Research Institute (PRRI), focusing on demographic change and American political behavior. He also served as a director of quantitative analysis at the Center for American Progress. Robert has taught courses on research methodology, statistics, public opinion, and political advocacy for The George Washington University, Pennsylvania State University, and Loyola University Chicago. He received his Ph.D. in political science and research methodology from The George Washington University.

William H. Frey is an internationally regarded demographer, known for his research on urban populations, migration, immigration, race, aging, and political demographics, as well as his expertise on the U.S. census. He is a senior fellow at the Brookings Institution, a research professor in population studies at the University of Michigan, and a co-director of the States of Change project. His most recent book is Diversity Explosion: How New Racial Demographics are Remaking America, just reissued in a revised and update edition.

Ruy Teixeira specializes in the study of demographic change, voting behavior, and public opinion. He is a senior fellow at the Center for American Progress and a co-director of the States of Change project. His most recent book is The Optimistic Leftist: Why the 21st Century Will Be Better Than You Think. His other books include The Disappearing American Voter; America’s Forgotten Majority: Why the White Working Class Still Matters; The Emerging Democratic Majority; and Red, Blue, and Purple America: The Future of Election Demographics.

We would like to thank the William and Flora Hewlett Foundation and the Democracy Fund for their generous support of this project and our board of advisers for helping shape the direction and output of the project. We would also like to acknowledge some of the numerous people who helped us conduct this research and produce this report. They include: John Fortier, Matthew Weil, Michael Thorning, and Brenna McAfee at the Bipartisan Policy Center and Karlyn Bowman of the American Enterprise Institute for valuable input and guidance; Cathy Sun of the University of Michigan Population Studies Center for programming our original projections; David Lanham of the Brookings Institution and Allison Preiss of the Center for American Progress (CAP) for publicity; Lauren Vicary, Will Beaudouin and Shanée Simhoni of CAP for report editing; Keenan Alexander, Chester Hawkins and Bill Rapp of CAP for report graphics; and, more generally, the entire Editorial and Art team at CAP for their excellent production work.

Estimates of voter and party composition, 1980–2008

For the estimates from 1980 though 2008, the analysis relied on three sources of data: census population data on eligible voters, turnouts rates derived from the November Supplement of the Current Population Survey, and American National Election Study.

Data on the composition of the eligible voter population⁵ was downloaded from the Integrated Public Use Microdata Series for 1980, 1990, 2000, 2004 and 2008. These populations were divided into 32 groups based on four racial groups (white, black, Hispanic, and Asian and other racial groups); four age groups (18 to 29, 30 to 44, 45 to 64, and those 65 and older); and two education groups (those with less than a four-year college degree and those who have a four-year college degree or higher). For election years falling in-between those years—1984, 1988, 1992, and 1996—the size of the group is taken as the average between the two relevant census years. So, for example, suppose a group made up 10 percent of eligible voters in 1980 and 20 percent of eligible voters in 1990. For the purposes of this report, we assume that group changed linearly over those 10 years, making up 14 percent of eligible voters in 1984 and 18 percent in 1988.

From there, turnout rates were derived from the November Supplement of Current Population Survey and applied to those 32 eligible voter groups. Specifically, for each election year we ran cross-nested multilevel models that estimated the turnout rate for each race, age, and education level group represented in the data. Many of these groups can be small, but this approach provides more realistic estimates of turnout for low-sample populations by partially pooling data across individuals’ demographic characteristics. Applying those turnout rates allows us to determine the size of each group among voters.

From there, a similar process is undertaken using data from the American National Election Study. The voting choices for those same 32 groups—that is, the rate at which they voted for the Democratic and Republican presidential candidate—was derived using cross-nested multilevel models that estimated the support rates for each year, race, age, and education level group represented in the data. Applying these support rates allows us to determine the size of each group in the Democratic and Republican coalitions.

Estimates of voter and party composition, 2012–2036

Estimates generated for 2012 onward are produced using the same types of data—eligible voter composition estimates, turnout rates, and support rates for 32 demographic groups. The only difference is that those building blocks are derived from different sources and more complex processes. For more information on those sources and processes, see the methodology section of the 2018 States of Change report.⁶

For full tables, please see PDF.