Explaining Electoral Change in the 2018 US Midterm Elections: The Three Components of Electoral Mandates

Introduction

The widely anticipated blue wave hit the shore of American politics on Election Day 2018.^[1] The Republican House majority was swept out to sea and a Democratic majority was left in its place. Democrats gained 41 House seats. This moved them from 194 seats following the 2016 election to 235 seats after the midterm. The 24 seat Republican majority after 2016 was replaced with an 18 seat Democratic majority. On the Senate side, Republicans added two members. This brought their numbers up to 53, a three seat majority.

An Intense Midterm Referendum

In several respects, the 2018 midterm election was unusual. As midterms go, turnout soared. According to the US Election Project, about 50.3% of the voting eligible population voted (McDonald 2018a). This exceeded turnout in the 2014 midterm by more than 13 percentage points. The last midterm election with turnout over 50% was held over 100 years ago (McDonald 2018b). Not coincidentally, campaign spending also soared. The Center for Responsive Politics (2018), as of late November, projected that $5.2 billion was spent by candidates, the parties, and PACs. By these incomplete numbers, campaign spending in 2018 increased by 35% over 2014. The greater intensity behind both the turnout and spending increases reflects the high level of polarization in the public and between the parties as well as the unusually strong feelings about President Trump (Campbell 2016).

Although unusual in its intensity, the 2018 midterm ran true to form in other respects. As usual in midterms, the president’s party lost House seats (Campbell 1997). The presidential party sustained seat losses in the House in all but three of the 38 midterms since 1870.^[2] The three exceptions were cases in which presidents were unusually popular at the time of the midterm (FDR in 1934, Clinton in 1998, and G.W. Bush in 2002).^[3]

Although the fact of midterm seat losses for the president’s party has been nearly a constant, the extent of these losses has varied quite a bit. The size of midterm seat losses depends, in no small part, on how popular the president is during the midterm campaign (Tufte 1975; Lewis-Beck and Rice 1984; Campbell 1997). Table 1 presents a simple sort of midterm election outcomes in 15 midterms since 1954.^[4] Midterms are sorted into two groups-those in which the president was relatively popular and those in which the president was less popular. More popular presidents are defined as those with Gallup job approval ratings in early September (or as near to then as was available) greater that 53%. Less popular presidents are defined as having Gallup approval ratings under 47% (Gallup 2018).^[5]

The importance of presidential approval is clearly evident in this simple division. When presidents were popular, their party’s seat losses were minimal, typically in single digits. On the other hand, the president’s party normally took a beating when the midterm electorate was less pleased with the president. The median seat loss for the party of less popular presidents was 35 seats. There is, of course, variation within each of these two groups, likely owing to what happened in the previous presidential election, but the overall midterm referendum on the president’s job performance shows through quite clearly nonetheless.

The 2018 midterm election was a normal election from this perspective. President Trump’s approval rating in early September of 2018 (and late October) was 40% according to Gallup (2018) and was 45% in the exit polls.^[6] As a relatively less popular president, like his predecessor President Obama, big midterm losses in the House were to be expected. From a different perspective, my “seats-in-trouble” forecasting model of House seat change (essentially forecasting the outcome of a normal election under the circumstances) predicted in mid-August that Democrats would gain 44 seats over their 2016 numbers, only three seats off of election’s outcome (Campbell 2018).^[7]

Based on this, we could conclude that the Republican Party in the 2018 midterm elections lost 41 seats because the public’s approval of the job performance of President Trump was more negative than positive. That conclusion, however, only begs the question. Why was President Trump’s approval rating at the midterm as low as it was? Why did the American electorate elect a solid majority of Republicans to the House of Representatives in 2016 then 2 years later replace it with a solid majority of Democrats? What explains the 2016 election and the electoral turn around in 2018? What changed?

Interpreting Electoral Mandates

These questions raise a larger question of how elections can be generally explained – why did an electorate elect one candidate rather than another? Why did the winner win and the loser lose? What decided the election? No area of American politics has been more thoroughly studied than voting behavior, yet our understanding and interpretation of election results has not received the serious attention it deserves. The determination of why voters decided to vest powers with one leader rather than another is at the core of democratic representation in governing and is critical to future elections (Stokes, Campbell, and Miller 1958, 1966; Kelley 1983; Jones 1991). Leaders or parties who ignore or misconstrue an electorate’s mandate do so at their own political peril both during their tenure and at their next election. They may well have good reasons occasionally not to heed an electorate’s wishes, but they ought to know they are doing so and what risks they are taking with their political futures.

As important as it is, the idea that elections provide mandates by the electorate to those elected has met with a great deal of resistence by scholars (Kelley 1983, p. 129–137; Dahl 1990; Hershey 1994). A major reason for skepticism about mandates is that they are difficult, some might say impossible, to determine what they are with certainty. Elections can involve hundreds of millions of voters, each one having many different ideas about why they are voting for or against a particular candidate. A vote is a blunt instrument. Out of all of this, when so many voters were saying so many things, how can anyone reliably say that what the collective message was or that there was anything close to a coherent collective message that leaders could discern?

Indeed, errors are easily and often made in interpreting elections. For example, one notable interpretation of the 2016 election erroneously dismissed the possibility that anger in the electorate played a major role in deciding the election’s outcome (Sides, Tesler, and Vavreck 2017), p. 35). The authors incorrectly reasoned that voter anger did not make a difference because, as they put it, “there was no increase in ‘anger’ among Americans heading into 2016.” Of course, anger does not need to increase for it to remain a potent motivation. Change in anger levels is irrelevant. What is relevant to affecting and interpreting an election is the level of anger and what that anger is about. As many of the skeptics of mandates contend and as this example illustrates, interpreting elections is difficult. That interpreting elections is difficult, however, does not mean their interpretation is any less important.

Some have gone further to claim that mandates do not exist, that “election outcomes have little real policy content (Achen and Bartels 2016), p. 319).” This view asserts that citizens are too ignorant, misinformed, and politically uninterested for election outcomes to have any substantive meaning. Lacking minimal civic competence in the public a representative democracy and substantively meaningful elections amount to merely a romantic myth. Mandates can not be accurately deciphered, according to this view because voters do not imbue elections with any substantive meaning.

These extreme claims, however, fall far short of being credible.^[8] The deficiencies of citizens have long been well documented. Undoubtedly, many come up well short of what might be expected. However, while some votes are cast for frivolous reasons, most are not. Voters are certainly imperfect, but most do not act randomly or in lock step with those with whom they share some socio-demographic characteristic. They have opinions and preferences – many of which we are able to measure fairly well on surveys.^[9] As an abundance of voting behavior research has well established, a great many votes are grounded in substantive reasons and some of these affect how elections are decided (Campbell et al. 1960); Key 1966; Kelley 1983; Lewis-Beck et al. 2008)

The most important and understandable reason for skepticism about electoral mandates is how they have been commonly defined. Though there is no settled definition, as frequently used, mandates require declarations from the public that are unrealistically strong, clear and specific and the standards determining the strength, clarity, and specificity requirements are frustratingly indefinite. In some cases, mandates are accepted as the claims of what elites declare them to be rather than as what moved the critical number of voters to decide the election’s outcome (Conley 2001; Grossback, Peterson, and Stimson 2007; Azari 2013). A more useful, focused, and realistic definition of an electoral mandate is the reason or reasons why the electorate decided as it did. By this definition, mandates are drawn from the electorate, not from what elected leaders or anyone else declares about the electorate’s motives. The voice of the people can be as strong or weak and as specific or general as they want it to be.

The Components of Mandates

Even with this clearer and more realistic but also more expansive definition of mandates, deciphering what the election’s result “turned on” is exceedingly difficult. Many modern efforts at interpreting elections are somewhat ad hoc and not well structured by prior voting research.

This is not far from where we stood 60 years ago when three of the four members of The American Voter team of election pioneers wrote in their article on “Components of Electoral Decision” (Stokes, Campbell, and Miller 1958) about the question “What combination of forces elects a President?” and the chaotic array of answers to it when applied to the 1956 presidential election. They proposed a framework to their analysis of what elections mean. That structure consisted of six broad partisan attitudes that organized what voters considered in making up their minds about their individual vote choices and what in the aggregate they were conveying to the leaders they elected (Campbell et al. 1960); Stokes 1966).

The six partisan attitudes, examined to understand the most proximate thinking about the vote choice and election outcomes (near the tip of The American Voter’s “funnel of causality” rather than more remote circumstances and predispositions), spanned the range of substantive concerns. These were drawn from responses to a battery of general open-ended questions asking respondents about what they liked and disliked about the parties and candidates. The six partisan attitudes composing this classification of responses included attitudes related to: (1) the Democratic candidate as a person; (2) the Republican candidate as a person; (3) group-related associations; (4) domestic policy; (5) foreign policy; and (6) the general performance of the parties and candidates in managing the nation’s affairs.^[10]

The six partisan attitudes affecting the vote and ultimately election outcomes are comprehensive, but not drawn from an organizing principle. The categorization, as a result, has a somewhat ad hoc quality to it. A natural organizing principle for making sense out of the many considerations of voters can be found by going back to the first principles of campaigning: the primacy of “the message” (as in James Carville’s admonition in 1992 that “it’s the economy, stupid”). In broad terms, the message of a campaign amounts to the reasons forwarded by the candidates or anyone else for why a prospective voter should vote for one candidate rather than another. These proposed reasons fall into three categories: past performance (the record, outcomes, retrospective evaluations), values (promises, policy, prospective outlooks), and leadership (the messenger). Strictly speaking about the message, prospective voters are asked to vote for or against a candidate or party because of what they had done in the past (as Ronald Reagan asked voters pointedly in 1980, “are you better off today than you were 4 years ago?”) or because of what they propose or are likely to do in the future. But voting for a candidate also involves a leap of trust in the honesty and competence of those asking to be elected. The elements of leadership are also important components of the vote. The messenger as well as the message is part of the package voters are asked to decide upon.

So how did the trinity of the electorate’s fundamental partisan attitudes-performance, values and leadership-lead to the outcome of the 2018 midterm?

Three Partisan Attitudes in 2016

The electoral verdict of 2018 cannot be understood without first understanding why the electorate decided 2016 as it did. Why did voters elect Donald Trump to the presidency in 2016 and with him 241 Republicans to the House? To clarify, this is not to ask why fewer than eighty thousand voters in a three states (Pennsylvania, Michigan, Wisconsin) tipped the electoral vote majority to Donald Trump. It is instead to ask why the election was close enough that so few votes were able to decide it. Hillary Clinton was a candidate with extensive political experience and faced an opponent with none. She and her supporters outspent her opponent and his supporters by two to one (Center for Responsive Politics 2018). Although she received plenty of bad press, it was much less unfavorable than her opponent’s press (Patterson 2017). Clinton faced a stiff challenge from Sanders for her nomination, but Democrats were far more unified around her than Republicans were around Trump. Yet despite all of these important advantages, she was only barely able to win a plurality of the popular vote and lost the electoral vote to an opponent who was widely thought to be unelectable, who frequently engaged publicly in embarrassing gratuitous feuding, and who had earned the public disdain of most of his party’s establishment. Why did this happen?

There is no shortage of explanations for what happened in 2016. Without taking these on one at a time, a task requiring many more pages than available here, we can safely set aside or dismiss outright the ideas that the outcome hinged on Russian hacking of the emails of prominent Democrats, FBI Director Comey’s late in the campaign reopening of the investigation into Clinton’s mishandling of classified emails, voter suppression, an increase in racial resentment, or any number of other theories.

The outcome of the 2016 election involved each of the three fundamental partisan attitudes. First, the performance of the in-party favored the Republicans. The electorate was not pleased with the record of the Obama administration.^[11] In the 2016 exit polls, 62% said the economy was “not good” or “poor.”^[12] A regression of real GDP growth in the quarters during the Obama presidency after the “Great Recession” ended indicated no trend of improvement. Real GDP growth had flat-lined at just over 2% growth, about a full percentage point below the median growth rate in the post-WWII period (33% below average). In the four quarters of 2016, real GDP growth exceeded two percentage points on only the second quarter and then fell short of 2.5% growth (Bureau of Economic Analysis 2018). Most tellingly, the economy was thought to be so fragile in 2016 that the Federal Reserve Board’s monetary policy maintained its low pro-stimulus “discount and advance rates” begun during the economic collapse of 2008.^[13] The Fed did not regard the economy as stable enough to risk the easing back to higher interest rates. The electorate concurred. Advantage Republicans.

The second partisan attitude in play in 2016 was leadership. Throughout the election year, poll after poll reported a majority of respondents holding unfavorable opinions of both Clinton and Trump, but with more prospective voters holding Trump in low esteem. In Gallup’s daily tracking polls in mid October, Trump’s unfavorables averaged about ten points higher than Clinton’s (about 65% to 55%). The exit polls asked about the candidates’ leadership traits-honesty and trustworthiness, temperament, and whether they were qualified to be president. Clinton fared better than Trump with voters on each of the leadership qualities (though just barely on honesty and trustworthiness). Advantage Democrats.

The values of the electorate, reflected in their ideological polarization, was the third partisan attitude affecting the 2016 outcome. In their direct effect on the election, values slightly favored Republicans-tilting by two percentage points toward a more conservative president than the incumbent and by five points toward a less active government. But it was the interaction of values and political circumstances that was most important.

Being out of power when politics are so polarized is particularly frustrating and the anger on the right had been building up for some time. From President George H.W. Bush reneging of his “read my lips, no new taxes” pledge, to the big spending “compassionate conservatism” of George W. Bush, to the maverick conservatism of John McCain and the self-described “severely conservative” Mitt Romney, conservatives felt unrepresented in the system. Even when they won big, as they had in the 2010 midterm, they had little in the way of policy to show for it.

Their anger came through most clearly in a Pew Research Center study conducted in late 2015. It asked whether “on issues that matter to you in politics today, has your side been winning or losing more often?” Only 44% of liberal Democrats thought they were usually on the losing side. In sharp contrast, 81% of conservative Republicans said they usually got the short end of the stick (Pew Research Center 2015), p. 10). Many conservatives felt unrepresented by the system and by their own party. In the exit polls during the primaries, about 41% of Republican voters on average said they were “angry” about the federal government and another 49% said they were “dissatisfied.” In those same primary exit polls, about 54% of Republican primary voters reported feeling “betrayed” by Republican politicians (CNN 2016b).

By 2016, conservatives were livid. Being on the outs in a highly polarized environment made them determined to prevent another 4 years of a liberal in the White House, but many faced a tough choice. According to Gallup (2016), better than a third of Republicans in mid-October held unfavorable views of their own party’s candidate.^[14] They thought Donald Trump was unfit for office.^[15] Despite Clinton being unacceptable to them ideologically and having her own perceived liabilities in leadership, a significant number of conservatives were in sympathy or were part of the “Never Trump” movement.^[16]

But as Election Day approached, an unpleasant choice forced by the anger and the ideology of many “Never Trump” conservatives was clear. It was not a vote they felt comfortable about, but the alternative was worse. They were sick and tired of losing. They wanted desperately to win. If nothing else, Trump was all about winning.

Many would rationalize their views of Trump as they reluctantly came to the decision to vote for him. According to Gallup (2016), Trump’s unfavorables among Republicans declined by about 6 or 7 points from mid-October to Election Day (34–35% to 28%). This accounts for much of the overall 5 point decline in unfavorables in the broader electorate (65% to 60%), and the two point increase in Trump’s share of the national two-party vote preferences in the last 2 weeks of the campaign.^[17] Several polls conducted during the campaign found Republicans to be three to eight percentage points less loyal in their vote intentions than Democrats. By the time of the election, this gap had been eliminated.

In the end, a fifth (based on the exit polls) to as much as a third (based on mid-October polls (Gallup 2016)) of Trump’s total vote came from voters who held unfavorable views of him. Of those simultaneously holding unfavorable views of both Clinton and Trump (about 18% of voters), Trump beat Clinton by 18 points (47% to 30%).^[18]

The bottom line of the 2016 election is that evaluations of performance (primarily the economy) favored the Republicans, but evaluations of leadership favored the Democrats (the perceived deficiencies of Trump as a potential president). The later were muted by considerations of values (some conservatives would set aside their disdain for Trump to avoid Clinton). In a word: polarization. This left the election closely enough decided that a relatively small number of votes in the industrial midwest could swing an electoral vote majority to Donald Trump.

From 2016 to 2018?

So what changed or perhaps did not change between 2018 and 2016 that would account for the reversal of fortunes in the House?

The partisan attitude of the parties’ performance on the key element of the economy was pretty much unchanged between the elections. The economy changed, it improved a good bit, but because of the change in presidential party (from the Democrat Obama to the Republican Trump) the partisan consequences did not change. In 2016 real GDP in the first three quarters the election year grew at weakly positive rates: 1.5% Q1, 2.3% Q2, and 1.9% in Q3 (Bureau of Economic Analysis 2018). The median growth rate in quarters since these data first became available in 1947 has been 3.1%. As presented in Table 2, this sub-par performance was accurately perceived by voters, with 62% in the exit polls reporting negative opinions about economic conditions. This favored change to the out-party Republicans.

By 2018, the economy had improved. Although the year got off to a ho-hum start, it picked up in later quarters: 2.2% Q1, 4.2% Q2, and 3.5% in Q3. Voters noted the improvement. In the 2018 exit polls, 68% reported positive impressions of the economy. An electorate pleased about the state of the economy was good news for the now in-party Republicans. As such, and with no other burning performance consideration in 2018, performance considerations do not help to explain why President Trump’s approval rating was low and why Republicans lost so many House seats.

The explanation of the blue wave in 2018 is found in the electorate’s evaluation of leadership. A substantial majority of the 2016 electorate, as already noted, did not think highly of Donald Trump’s leadership potential. Large majorities thought he was not honest or trustworthy (64%), not qualified (61%), and lacked the proper temperament to be president (63%). A large majority held overall unfavorable views of him (60% in the exit polls). Many holding these less than flattering views of Trump were conservatives and moderates and many of them ended up voting for Trump anyway because they faced a choice of either Trump or Clinton and saw Trump as the “lesser of two evils” and of a polarization-fueled anger of being too often on the losing political side.

Opinions of President Trump’s leadership had not changed much by 2018. An average of ten polls conducted between April and October of 2018 indicated 57% of Americans held an unfavorable view of President Trump (Polling Report.com 2018). According to The Fox News Voter Analysis (2018) around Election Day, 53% of voters held unfavorable views. As in 2016, large majorities continued to think Trump was not honest or trustworthy (62%) and lacked the proper temperament to be president (64%). The Fox study also found that 58% of voters in 2018 did not believe President Trump cared about people like themselves.

These views of leadership in 2018 were basically unchanged from what they had been in 2018, but what had changed were two important contexts and they made a huge difference. First, since they do not involve the election of a president, midterm elections are more readily regarded by voters as a referendum on the in-party and its president than a choice between parties and their leaders. In a sense, some voters holding Trump in low esteem in 2016 may have felt the pressure of making a choice between him and his liberal and flawed opponent. They eventually came back to their party and its candidate. In 2018, without the presidency at stake, that pressure was relieved.

Second, the motivating anger of polarized voters who were out of power shifted from the right to the left as a result of the controversial 2016 election. In 2016, conservatives who felt disenfranchised were desperate to get back into power. Many would overlook or put up with what they believed were the many leadership shortcomings of Donald Trump if he could help them prevent the continuation of what they say as the radical misrule of the Obama years. The results of the 2016 election offered them relief, an uneasy relief given Trump’s leadership style, but nonetheless a relief.

The anger of being polarized and out of power now shifted to the left for 2018. Many liberals found a Trump presidency to be an abomination. As the right saw the election of liberals as the result of voter fraud, the left now saw a Trump victory as rigged by voter suppression and Russian “meddling.” Their anger showed up in the high voter turnout and campaign contribution numbers as well as in the 2018 exit poll and the Fox News Analysis Survey. Most of those holding negative opinions of President Trump – the 54% who disapproved of his performance in office and the 53% who held an overall unfavorable opinion of him – reported holding these views “strongly.” With respect to both approval and favorability, about 44% said their impressions were strongly negative and they voted overwhelmingly (about 90%) for Democratic House candidates. In contrast, only about 30% of voters said they had strongly positive impressions of the president.

The 2018 midterm election turned on the partisan attitudes about leadership, again interacting with values-fueled anger on the left in this election. Table 3 displays the crucial clue to interpreting 2018. The exit polls and Fox News Analysis Surveys asked voters whether a reason for their vote had been to express either support for or opposition to President Trump? Nearly two-thirds (63%) claimed their vote had been directed at President Trump and a large majority of them said their vote was cast as expression of opposition to Trump. Looking at this from another angle, if Trump was the focus of the vote, as it was for a majority of voters, congressional Republicans lost to congressional Democrats by 16 points. Among those whose vote was not about Trump (about a third of voters), congressional Republicans beat congressional Democrats by 8 points.

There had been two components to the mandate of 2016: restore greater vitality to the economy and govern in a way that Americans can feel proud of or at least feel comfortable with. Republicans appear to have delivered fairly well on the first component, but not on the second. Trump was still being Trump and, though that sells well in some corners, overall it is a political liability and was one in 2018.