John Sides, Michael Tesler, Lynn Vavreck. Journal of Democracy. Volume 28, Issue 2. April 2017.
Around 7:22 p.m. on November 8, the night of the 2016 presidential election, a member of Republican candidate Donald Trump’s campaign team told CNN reporter Jim Acosta: “It will take a miracle for us to win.” The Trump campaign was not alone in this view. Most political observers also expected a win by Trump’s Democratic rival, former secretary of state Hillary Clinton. The Clinton campaign staff certainly did: They were “all smiles” at 5 p.m., when a Boston Globe reporter arrived at the scene of their anticipated victory party.
As the night went on, all of this would change. Trump’s success in key battleground states-Florida, Michigan, Ohio, Pennsylvania, and Wisconsin-gave him the presidency, with an expected 306 Electoral College votes against Clinton’s 232. But Clinton led in the popular vote and, as the days after the election passed, her vote share continued to grow. Ultimately, she won the popular vote by 2.1 points (48.0 to 45.9 percent), or almost 2.9 million people. The divergence between Clinton’s popular-vote victory and her Electoral College defeat was extraordinary. The last time popular and electoral outcomes had differed, in the 2000 election, the margins of victory according to both counts had been considerably narrower: Then-Vice-President Al Gore won the popular vote by about half a point, while losing the Electoral College to George W. Bush by only five votes.
There is, then, no simple way to interpret the election outcome. Of course, Trump was the clear victor given the rules of U.S. presidential elections. But he also received many fewer popular votes than Clinton. Any explanation must be able to account for both of these facts. Our analysis highlights three key factors-two that involved the broader social, economic, and political conditions that the candidates faced, and one that resulted from choices made during the campaign itself. The first factor consists of underlying political and economic fundamentals, which include Democratic president Barack Obama’s 2016 approval ratings and the economy’s growth rate. These were quite consistent with Clinton’s popular-vote victory.
To understand how Trump won the Electoral College, we must consider two other interrelated factors. The second factor is increasing racial and ethnic polarization in the Democratic and Republican party coalitions. Particularly during Obama’s presidency, voters increasingly sorted themselves by party on the basis of identities and views related to race and ethnicity. This shift transformed both parties well before Trump’s 16 June 2015 ride down the escalator in New York’s Trump Tower to announce his candidacy. The shift also set the stage for a third and final factor that helped Trump to win: the centering of both campaigns on issues that tapped into Americans’ racial, ethnic, and social identities and attitudes. An identity-focused framing of the election heightened Trump’s appeal to white voters, and particularly those without a college education-demographics with a strong presence in key swing states.
The presidential election thus also became a referendum on who Americans believed they were, and how they felt about those who were different from them. Ultimately, the election was a manifestation of the country’s broader identity crisis. As the United States changes demographically, socially, and culturally, Americans’ political identities are increasingly driven by competing understandings of what their country is and ought to be-a multicultural society that welcomes newcomers and embraces its growing diversity, or a more provincial place that recalls an earlier era of traditional gender roles and white Christian dominance in economic and cultural life.
The dominant narrative of 2016 centered on allegedly “angry” U.S. voters who were dissatisfied with their government, the economy, and the direction of the country. Indeed, during the campaign, most Americans expressed distrust in government and said that the country was “on the wrong track.” Yet this narrative falls short in several regards. For one thing, levels of dissatisfaction and distrust in 2016 were no higher than they had been in recent years. There was no increase in “anger” among Americans heading into 2016. Moreover, the anger that did exist had no clear electoral implication. After all, President Obama had been comfortably reelected in 2012, when a similar percentage of Americans had been saying that the country was on the wrong track.
In addition, talk of “angry voters” missed a countervailing trend: Americans’ increasing economic optimism. The longest-standing measure of how Americans view their country’s economy, the University of Michigan’s Index of Consumer Sentiment, generally showed rising confidence during Obama’s presidency. Over the course of his time in office, more Americans were telling interviewers that their family was better off financially than it had been a year earlier. More expected that both their family’s financial situation and general business conditions would be good over the next year. More felt that it was a good time to make big purchases such as furniture or a refrigerator. Over the first three quarters of 2016, consumers were only slightly less optimistic than they had been in 1984-when Ronald Reagan won reelection to the presidency with a campaign based on the country’s economic recovery, touted as “morning again in America.” Obama’s approval rating, though lower than the increasingly positive mood among consumers might suggest, also increased in 2016-by roughly eight points between January and election day. In the months leading up to the election, then, “angry Americans” somehow took increasingly favorable views both of the economy and of their president.
Economic growth and presidential approval are two of the key fundamentals that affect presidential-election outcomes. In 2016, these two factors suggested a narrow Democratic popular-vote victory, which is precisely what happened in November. Using the president’s approval rating in June of the election year and the growth in U.S. GDP from the first to third quarters of that year, we estimated a statistical model of presidential-election outcomes from 1948 to 2012. We then calculated a forecast based on Obama’s June 2016 approval rating of 50 percent and the economy’s approximately 1.1 percentage points of nonannualized growth from the first to third quarters of 2016.
The forecast gave the Democratic candidate 51.8 percent of the major-party vote (that is, of the total number of votes received by either the Democratic or the Republican nominee). This is almost exactly Clinton’s actual share of the major-party vote (51.1 percent). The popular vote totals defied the historical tendency for incumbent parties to lose after two terms in the White House. If we incorporate into our model the penalty that incumbent parties usually incur after two terms, the forecast is for a narrow Republican win. Clinton beat that forecast by more than two points.
To illustrate how economic fundamentals help us to predict the 2016 popular vote, Figure 1 tracks the relationship between growth in GDP and the incumbent party’s vote share in presidential elections since 1948. The 2016 election falls right on top of the line that captures this relationship, which shows just how closely Clinton’s vote margin corresponded to the historical pattern.
Of course, election outcomes depend on more than a few fundamentals. But these fundamentals tell a key piece of the story. They help us to understand why, contrary to a simplistic narrative of “angry” Americans, a plurality of voters supported a third term for the Democratic Party and the Clintons’ return to the White House-hardly a rebellion against the status quo or the “establishment.”
The Changing Party Coalitions
While the economic and political fundamentals help to explain Clinton’s two-point victory in the popular vote, this margin obviously did not put Clinton in the White House. To understand Trump’s victory in the Electoral College, one must look at developments that occurred well before 2016. More precisely, we should look back about eight years, to the initial election of Barack Obama.
Many hoped that Obama’s election would inaugurate a “post-racial” America. Instead, Obama’s candidacy and then presidency made race matter more. As Obama said in his 10 January 2017 farewell address: “After my election, there was talk of a post-racial America. Such a vision, however well-intended, was never realistic. For race remains a potent and often divisive force in our society.” Divisions over race were palpable throughout Obama’s presidency. Surveys revealed that opinions on virtually every issue connected to Obama became associated with people’s race and, among white voters, with attitudes toward African Americans. One study even showed a correlation between whites’ racial attitudes and whether they liked Obama’s dog.
This (further) racialization of politics profoundly affected the Democratic and Republican coalitions. By 2016, nonwhites had become increasingly Democratic. Pew Research Center data show that among Latinos, self-identified Democrats outnumbered Republicans by 36 points in 2016, as compared to 23 points in 2002. In exit polls, Asian Americans’ support for Democratic presidential candidates increased from 31 percent in 1992 to 73 percent in 2012. Blacks also came to identify even more strongly with the Democratic Party. In the American National Election Study, the percentage of African Americans who said that they were “strong” Democrats increased from 31 percent in 2004 to 55 percent in 2012.
During Obama’s presidency, whites shifted notably toward the Republican Party. In Pew surveys from 2007, whites were just as likely to call themselves Democrats as Republicans (44 percent to 44 percent). By 2010, whites were twelve points more likely to be Republicans than Democrats (51 percent to 39 percent). By 2016, that gap had widened to fifteen points (54 percent to 39 percent). This shift stemmed from changed preferences among whites without a college degree. Whites who did not attend college were evenly split between the two parties from 1992 to 2008. But by 2015, white voters with a high-school degree or less broke for the Republican Party by 24 percentage points (57 percent to 33 percent). Meanwhile, college-educated whites moved toward the Democratic Party. The evidence also suggests greater movement among white men than among white women during Obama’s presidency, accelerating a long-term trend of white-male flight from the Democratic Party.
Racial attitudes were a key factor in these trends. Whether one looks at whites overall or at whites without a college degree, shifts toward the Republican Party occurred mainly among respondents with less favorable attitudes toward African Americans. No other factor predicted changes in white partisanship during Obama’s presidency as powerfully and as consistently as racial attitudes.
Obama thus bequeathed to Clinton not only an approval rating conducive to her election, but a changed party coalition. Clinton’s campaign knew this, of course, and committed early to reconstituting a coalition similar to the one that had elected and reelected Obama, but perhaps even more tightly concentrated on already Democratic-leaning voters, locales, and demographics. Many analysts believed that the country’s growing ethnic diversity made this a winning strategy.
Obama bequeathed a different coalition to the Republicans as well. Most of the pro-Republican shifts among whites, and especially whites without a college education, occurred before 2016. In the eyes of many analysts-and quite a few Republicans-the growing ethnic homogeneity of their supporters left Republicans at a disadvantage in a diversifying country. After the 2012 election, the Republican National Committee’s postmortem report emphasized the need to reach out to Latinos and Asian Americans.
But a handful of analysts and pundits voiced a contrary view: The Republican Party could win by strengthening its appeal to white voters. Doing so arguably meant speaking to the racial attitudes of whites, and especially whites without a college education. These voters, in Public Religion Research Institute (PRRI) polls between 2012 and 2015, were more likely than other racial groups to say that that American culture had “changed for the worse” since the 1950s and that the country’s “best days are behind us.” Such sentiments appeared to reflect racial concerns as much as or more than economic concerns. In the 2015 PRRI survey, whites without a college education were not much more likely than other groups to say that jobs and unemployment were a “critical issue,” or that the country’s economic system benefited the wealthy. But both this group and whites in general were much more likely to say, as they had in earlier surveys, that discrimination against whites was as big a problem as discrimination against blacks. And when asked whether immigrants “strengthen the country” or “are a burden,” 65 percent of whites without any college education and 45 percent of college-educated whites chose “burden.”
When Hillary Clinton announced her candidacy on 12 April 2015, she talked about “everyday Americans” forging beginnings in new jobs and moving to new houses. Then she said that she was hoping to get a new job and a new house, too.
Two months later, Donald Trump announced his candidacy by declaring that Mexico was sending rapists to the United States and the country needed to build a wall to keep them out. His slogan-“Make America Great Again”-reminded voters of what once was, suggesting that a past “greatness” had been lost and that he alone could restore it. His campaign emphasized plans to keep out people whom it depicted as undeserving of America’s spoils, be they Mexican or Muslim. Trump frequently divided the world into “us” and “them.”
Over the summer of 2015, the Clinton campaign pivoted to address the same theme, but from the completely opposite position. Clinton’s message, that we are “stronger together,” reminded voters that our differences can be what make us great. Clinton’s ads were filled with the faces of Americans from all racial, ethnic, and religious backgrounds. Whereas Trump wanted to keep out certain groups who differed from the country’s white majority, Clinton wanted to welcome and embrace them.
It is not surprising, then, that we can see strong correlations between voter preferences in the 2016 contest and attitudes related to race, ethnicity, and religion:
- During the Republican primary, whites’ support for Trump was related to the importance they placed on whiteness as a part of their identity, as well as to how much they believed that whites suffered from discrimination.
- Attitudes toward African Americans, immigrants, and Muslims were more strongly related to support for Trump in the 2016 primary race than they had been to support for John McCain and Mitt Romney, the eventual victors in the 2008 and 2012 Republican races;
- Both white ethnocentrism and whites’ attitudes toward African Americans were more strongly related to voter preferences in the Clinton-Trump race than they were to preferences in matchups between Clinton and Marco Rubio or Ted Cruz, two of Trump’s leading Republican competitors;
- White voters’ attitudes toward African Americans were also more strongly related to their preferences in the Clinton-Trump contest than they had been to preferences in the general elections pitting Obama against McCain in 2008, and against Romney in 2012;
- The educational divide in whites’ support for Clinton against Trump disappeared after racial attitudes were taken into account-suggesting that differing attitudes toward ethnic minorities among more and less educated white voters were a key reason for the educational split in voting.
In short, views on race and ethnicity played a heightened role in the Trump-Clinton matchup as compared to past electoral contests.
In December 2016, we re-interviewed eight-thousand respondents whom we had originally interviewed in 2011-12 as part of a study of the 2012 presidential election. Thus we can determine how respondents voted in each election, whether they shifted between the parties, and what factors were associated with any shifts. As one would expect in an era of strong partisanship, the vast majority of Obama and Romney voters were loyal to their parties in 2016. About 86 percent of Obama voters voted for Hillary Clinton, and 88 percent of Romney voters voted for Donald Trump. But that still leaves a substantial proportion of respondents who did not vote for the same party in both years. About 9 percent of Obama voters supported Trump, while 4 percent supported a candidate who was not a major-party nominee. About 5 percent of Romney voters supported Clinton, while 6 percent supported a third-party or independent candidate. Consistent with exit-poll data, shifts from Obama to Trump were more prevalent among whites who did not attend college. Among this group, 22 percent of Obama voters voted for Trump.
These defections were especially pronounced among the sizable share of white Obama voters with conservative positions on immigration. In December 2011, we asked our survey respondents three questions about immigration: 1) whether they supported a path to citizenship for undocumented immigrants; 2) whether undocumented immigrants were mostly a benefit to or a drain on society; 3) and whether it should be easier or harder to immigrate to the United States. When we used voters’ responses to place them on a scale ranging from positive to negative views of immigration, we found that these views were strongly associated with vote changes from Obama to Trump, and from Romney to Clinton.
The left-hand panel of Figure 2 shows that Clinton retained almost all of Obama’s white voters with positive views of immigration. But she lost about a third of white Obama voters whose attitudes on immigration were at the negative end of our scale. This trend had significant electoral implications due to the substantial number of white Obama voters who tended to hold anti-immigrant attitudes in 2011: For example, 33 percent of white Obama voters said that “illegal immigrants” were “mostly a drain,” compared to 40 percent who said that they “mostly make a contribution.” Similarly, 34 percent said that it should be harder “for foreigners to immigrate to the United States,” compared to just 33 percent who said it should be easier and 21 percent who said there should be no change. In other words, lurking within the Obama coalition were many white voters with unfavorable views of immigration and immigrants. Many of these individuals voted for Trump in 2016.
Views of immigration were also associated with defections in the other direction-from Romney to Clinton. The left-hand panel of Figure 2 shows that Clinton won about 28 percent of Romney voters with the most favorable attitudes toward immigration. The problem for Clinton, however, was that there were far fewer white Romney voters in this category: Only 14 percent of white Romney voters had attitudes that placed them on the positive side of the scale. For example, only 6 percent thought that “illegal immigrants” made a contribution to American society; the vast majority, nearly 80 percent, thought that these immigrants were a drain.
Other evidence suggests that views on immigration were especially relevant to voter choice in 2016. We estimated statistical models of white voters’ candidate preference in both 2012 and 2016, including attitudes in 2011 toward not only immigration but also economic issues, social issues, health care, trade, and African Americans as a social group, as well as party identification and gender. Many of these variables, including the widely discussed factor of gender, were correlated with Americans’ presidential preferences in both 2012 and 2016. But the factors that were more correlated with voting behavior in 2016 than they had been in 2012 were attitudes toward immigration and attitudes toward African Americans.
Donald Trump’s signature issue of immigration thus appeared wellpositioned to reinforce the white flight from the Democratic Party that had taken place during Obama’s presidency. The consequence was a historically large education divide among white voters that came down in large part to attitudes about race and ethnicity. The education divide among whites provided Trump with a narrow path to victory. With black Democratic support and turnout returning to their pre-Obama levels, the large share of whites without a college degree in the key states of Michigan, Pennsylvania, and Wisconsin helped to put Trump over the top in the Electoral College.
This does not mean, of course, that the election was only about race and ethnicity. But demographic shifts in the respective major-party coalitions during the eight-year tenure of the country’s first black president, combined with a presidential campaign that presented diametrically opposed visions of what America should stand for, set the stage for the election outcome to turn largely on issues related to racial, ethnic, and social identities.
In a narrowly decided election like that of 2016, many factors could conceivably have shifted the outcome, including some that we lack space to discuss here. One example is the impact of attention to Clinton’s e-mail server on her levels of support. Another is the question of whether her campaign could have done more in the Rust Belt to mobilize Democratic voters, particularly black voters. For this reason, the list of factors that we have examined is far from exhaustive.
Nevertheless, these factors are key building blocks in explaining how it was that Clinton won the popular vote and Trump the Electoral College. Several of these factors-economic growth, evaluations of Obama, the changing party coalitions-were outside the candidates’ control and even pre-dated the campaign. They are yet another reminder that presidential elections hinge on fundamentals far removed from the day-today hurly-burly of retail politicking, television ads, and tweets. Yet the strategic choices of the candidates also helped to determine the impact of racial and ethnic identities on the election outcome. In 2016, Donald Trump and Hillary Clinton took sharply divergent positions on issues emanating directly from the country’s growing ethnic and religious diversity. Attitudes about those issues then became central to how Americans voted.