Recent presidential elections in the United States have been characterized by sharp divisions between Democrats and Republicans on a wide range of issues along with high levels of party loyalty and straight-ticket voting. Voting patterns in these elections have been very stable — the same voter groups and the same geographic areas have consistently supported either the Democratic candidate or the Republican candidate. Democratic candidates have depended heavily on younger voters, nonwhites, and single women while Republican candidates have depended heavily on white voters — especially southern whites — as well as married men, those over the age of 60, and those without college degrees. Overall, 40 out of 50 states voted for the same party in all four elections between 2000 and 2012.
But none of these elections involved a major party nominee like Donald Trump — an insult-hurling businessman who has never run for elected office, has no longstanding ties to the Republican Party, and who has regularly attacked the party’s established leadership including its most recent president and its two most recent presidential nominees. Since Trump’s emergence as the favorite to win the Republican nomination, there has been considerable speculation about whether his candidacy could result in major shifts in the voting patterns that have characterized recent elections.
The ultimate answer to this question will have to await the November election. However, data from a survey conducted earlier this year allow us to draw some tentative conclusions about the likelihood of major shifts in voting patterns. In this article I analyze data from the 2016 American National Election Study Pilot Survey, an online survey of 1,200 voting-age Americans conducted between Jan. 22-28, 2016. Results presented in this article are based on approximately 750 respondents who identified with or leaned toward one of the two major parties and who indicated that they were registered to vote.
Although the ANES pilot survey was conducted in January, before the Iowa caucuses, Trump had already emerged as the clear frontrunner for the GOP nomination — 39% of registered Republican identifiers and leaners favored Trump in this survey. Ted Cruz was a distant second with 17% support and Marco Rubio was third with 12% support. On the Democratic side, Bernie Sanders had already established himself as a serious challenger to frontrunner Hillary Clinton: Clinton led Sanders by 52% to 37% among registered Democratic identifiers and leaners.
The first question I will address is whether Democratic and Republican voters remain sharply divided on major issues. I will also examine whether there were substantial divisions on major issues between supporters of different candidates within each party and whether these intra-party divisions were larger or smaller than the inter-party divisions. I examined voters’ opinions on three types of issues — economic issues, cultural issues, and racial issues. The economic liberalism-conservatism scale was measured by three questions — one on spending for child care, one on spending for health care, and one on raising taxes on upper-income households. The cultural liberalism-conservatism scale was measured by two questions — one on whether religious organizations should be allowed to opt out of providing coverage for contraceptives for their employees and one on whether businesses providing wedding services should be able to refuse service to same-sex couples. Finally, the racial liberalism-conservatism scale was based on four questions measuring symbolic racism or racial resentment that have regularly been asked in ANES surveys.
Table 1 displays average liberalism scores for Democratic and Republican voters on economic, cultural, and racial issues. The table also displays average liberalism scores for Democrats supporting Clinton and Sanders and for Republicans supporting Trump and other Republican candidates. Liberalism scores are based on the percentage of voters in each subgroup who received scores above the mean liberalism score for all registered voters on each issue.
Table 1: Policy polarization in 2016 — percentage liberal by party and candidate preference
Source: 2016 ANES Pilot Study
The results in Table 1 show that on all three issues there was a deep divide between Democratic and Republican voters. Moreover, on all three issues, the differences between Democrats and Republicans were much larger than the differences between supporters of different candidates within each party. In each issue area, 80%-84% of all Democratic identifiers and leaners scored above the overall mean for liberalism among registered voters compared with only 22%-26% of all Republican identifiers and leaners. In contrast, the differences between supporters of Clinton and Sanders and between supporters of Trump and other Republican candidates were quite modest in size.
The largest difference between Clinton and Sanders supporters was not on economic issues, as one might have expected given Sanders’ focus on income inequality, but on cultural issues. This may reflect the fact that Sanders did best in this survey, as in just about every other national survey as well as exit polls of Democratic primary voters, among Democrats under the age of 30 — a group that tends to be somewhat more liberal than older Democrats on cultural issues. However, Clinton supporters were far more liberal on cultural issues than supporters of any Republican candidate.
Not surprisingly, given Trump’s frequent attacks on immigrants and his questioning of Barack Obama’s citizenship and patriotism, the largest difference between Trump supporters and supporters of other GOP candidates was on racial issues. Trump supporters scored significantly higher than other Republicans on our racial resentment scale. However, even supporters of other Republican candidates scored far higher on racial resentment than supporters of either Democratic candidate.
Opinions on the economic, cultural, and racial issues scales were strongly related — the correlation between economic and cultural attitudes was .53, the correlation between economic and racial attitudes was .57, and the correlation between cultural and racial attitudes was .49. Therefore, in order to measure the overall level of partisan polarization in the 2016 electorate, I combined the three issue scales into a single measure of liberal-conservative policy orientations. This liberal-conservative issues scale runs from 0 to 10, with a score of 0 representing consistently liberal attitudes and a score of 10 representing consistently conservative attitudes. Figure 1 displays the distributions of Democratic and Republican voters on this scale.
Figure 1: Policy polarization in 2016 — distributions of Democratic and Republican voters on liberal-conservative policy scale
Source: 2016 ANES Pilot Study
The results in Figure 1 show that Democrats and Republicans were sharply divided in their ideological orientations — 79% of Democratic identifiers and leaners were located to the left of center (0-4) on this scale while 78% of Republican identifiers and leaners were located to the right of center (6-10). The average score on the scale was 3.0 for Democrats compared with 7.1 for Republicans.
Once again, the distances between supporters of different candidates within each party were much smaller than the distance between supporters of the two parties. On the Democratic side, 89% of Sanders supporters were located to the left of center as were 77% of Clinton supporters. The average score on the scale was 3.0 for Clinton supporters compared with 2.5 for Sanders supporters. On the Republican side, 81% of Trump supporters were located to the right of center as were 78% of supporters of other Republican candidates. The average score on the scale was 7.3 for Trump supporters compared with 7.1 for supporters of other Republican candidates.
Recent studies have shown that in addition to being sharply divided on policy issues, Democratic and Republican voters are sharply divided in their feelings toward the political parties and their leaders. Since the 1990s, affective polarization has increased along with policy polarization. Supporters of each party now have much more negative feelings about the opposing party and its leaders than in the past, which has contributed to increasing party loyalty and straight-ticket voting in elections.
Table 2: Affective polarization in 2016 — average feeling thermometer ratings of Clinton and Trump by party and candidate preference
Note: Thermometer ratings range from 0 to 100
Source: 2016 ANES Pilot Study
We can measure the extent of affective polarization in the 2016 electorate by comparing the ratings of the presidential candidates on the ANES feeling thermometer scale by Democrats and Republicans and by supporters of competing candidates within each party. The more positive a respondent felt about a candidate, the higher the thermometer score. Table 2 displays these ratings. The data in this table show that Democrats and Republicans gave their own party’s frontrunner fairly positive ratings — Democrats gave Clinton an average rating of 71 degrees and Republicans gave Trump an average rating of 65 degrees. But the data show that Democrats and Republicans gave the opposing party’s frontrunner extremely negative ratings — Democrats gave Trump an average rating of only 19 degrees and Republicans gave Clinton an average rating of only 12 degrees. Supporters of both parties, but especially Republicans, disliked the opposing party’s frontrunner more than they liked their own party’s frontrunner — a phenomenon that reflects the prevalence of negative partisanship in the contemporary American electorate.
Not surprisingly, supporters of the leading candidate in each party — Clinton for the Democrats and Trump for the Republicans — had much more positive feelings about that candidate than supporters of competing candidates. Sanders supporters gave Clinton a fairly mediocre average rating of 55 degrees while supporters of Republican candidates other than Trump gave Trump an even lower average rating of 50 degrees. At the same time, however, supporters of these competing candidates expressed strong dislike for the opposing party’s frontrunner. On the Democratic side, Sanders supporters gave Trump an even lower average rating than Clinton supporters — 14 degrees versus 20 degrees. And on the Republican side, supporters of candidates other than Trump gave Clinton an average rating of 13 degrees, almost identical to the average rating of 11 degrees by Trump supporters. This intense dislike for the opposing party’s frontrunner among Sanders Democrats and non-Trump Republicans is potentially very significant because it suggests that even those supporting unsuccessful candidates would be very likely, in the end, to vote for their party’s nominee.
Group voting patterns
If the analysis presented in this article is correct, we would expect group voting patterns in the 2016 presidential election to be very consistent with group voting patterns in the 2012 presidential election. As a preliminary test of this claim, we can examine the relationship between group voting preferences in a recent national poll with those from the 2012 national exit poll. Figure 2 displays a scatterplot of these results. The data for 2016 come from a CNN/ORC national poll from June 16-19.
Figure 2: Group voting patterns in 2012 and 2016
Sources: CNN/ORC, 2012 national exit poll
The results displayed in Figure 2 show an extremely high degree of consistency in group voting patterns between these two elections. There is a correlation of .99 — an almost perfect relationship — between Obama’s margin over Mitt Romney in the 2012 exit poll and Clinton’s margin over Trump in the 2016 CNN/ORC poll across these voter groups. These findings strongly support the belief that group voting patterns in 2016 are likely to closely resemble group voting patterns in 2012. Despite his unusual background and controversial views, Trump is likely to do best with the same voter groups that favored the much more conventional Romney in 2012 and Clinton is likely to do best with the same voter groups that favored Obama in 2012.
The 2016 presidential campaign has already produced many unexpected developments, none more surprising than the fact that Trump was able to overcome a large field of experienced candidates and intense opposition from almost the entire GOP establishment to become the presumptive Republican nominee. Trump’s highly unusual background, personality, and unorthodox views on certain issues have led to considerable speculation that his nomination could upset normal voting patterns by producing high defection rates among some groups of Democratic and Republican identifiers and putting new states in play in November.
No definitive conclusions about how Trump’s candidacy will affect voting patterns in the 2016 presidential election can be reached until after the results are known. However, an examination of data from the 2016 ANES Pilot Study strongly suggests that claims by some pundits and political scientists that Trump’s candidacy could result in very different voting patterns from those seen in recent presidential elections are probably mistaken. These data show that the American electorate remains deeply divided along party lines. Democrats and Republicans, including independents who leaned toward each party, differed sharply on economic, cultural, and racial issues. Moreover, Democrats and Republicans, including Sanders Democrats and non-Trump Republicans, held strongly negative feelings about the opposing party’s likely nominee.
Based on these findings, voting patterns in the 2016 general election should closely resemble those seen in recent presidential elections. Overwhelming majorities of party identifiers and leaning independents are likely to support their party’s nominee. As a result, the large majority of states are likely to support the same party as in 2012: states that were strongly Democratic in 2012 will probably remain strongly Democratic in 2016, states that were strongly Republican in 2012 will probably remain strongly Republican in 2016, and states that were battlegrounds in 2012 will probably remain battlegrounds in 2016. It seems that when it comes to presidential elections, the more things change, they more they remain the same.