So I, like everyone else except for some comedians, got the election wrong. Even though Hillary is leading Trump by over 1,000,000 votes (61,913,199 votes (47.9%) 60,911,924 votes (47.1%) as of 11/16), Trump won the election with a strong electoral college showing, 306 to 232.
First, it is really important to note the obvious, which many in the media seem to be missing here. With the exception of Florida–a perennial tossup–all of Trump’s gains were in the Rust Belt. Hillary’s support in many western states was either as good or better than Obama 2012, she made gains in Texas (almost 600,000 more votes than Obama 2012), forced Trump to win by plurality in Arizona (although this is arguably more the product of Gary Johnson’s vote share) and only slightly fell behind Obama’s high water marks (2008 or 2012) in Georgia and South Carolina, and had a wider margin of victory in Virginia. The Northeast trended toward Trump, but only by a few percentage points (for example, Hillary lost 3 percentage points on Obama’s 2012 margin in NJ, even amid increased turnout). The only exception in this region is Upstate Maine, which swing to Trump by a net spread of 20 points. In short, in the aggregate the West, South, and Northeast did not shift very much in this election. What did shift was the formerly industrial Great Lakes/Appalachian states that moved heavily toward Trump. A combination of Trump turnout surge among uneducated white workers, lack of turnout among urban African American voters, and suburban Obama voters fleeing the Democrats explain the general political terrain in these states.
Why did suburban voters switch from Democratic support to pro-Trump?
Trump won 50% of the suburban vote, while Hillary won 45%, a pretty sizeable margin in the largest geographic electorate (49% of the electorate, compared to 34% in the big cities, and 17% in rural America. Why did this happen? This one really comes down to the fundamentals, and speaks the least to either of the two candidates unpopularity or transformative campaigning. Many forecast models actually had a generic Republican beating a generic Democrat in this election, based purely on a handful of variables. Among them, the two most important predictors are how long the party in power has held office, and the economic growth rate leading into the election. Well, the Democrats have held the executive office for eight years, which generally favors the out-party to gain the presidency. The only exceptions to this in the 20th century are Taft following TR’s two incomplete terms, Truman winning in 1948 after assuming the presidency after FDR passed, and George HW Bush winning his solitary term after Reagan’s 8 years (some consider only the latter to be analogous to the current situation given the shortened time-frame of the former, and Truman’s station of VP in the latter). Economic growth rates have been steady, but low in 2016. The first quarter had a recorded growth rate of 0.8%, the second 1.4%, and the third (ending with September) had 2.9%. Except for the last quarter, this country has not seen very much economic growth this year. While it is always unclear on what basis people feel or understand economic conditions in their everyday lives, it has historically been a good indicator into the public mood on staying pat or changing leadership.
For these basic reasons, in all likelihood suburban voters (“middle America”) were going to swing to some degree to Trump. America does have a strong tradition of switching party in power following a two-term president, and these are the types of people that generally see to that. It is also important to note this category of voters is the least likely of the three (with high white voter turnout and decreased Black turnout) to be instructed by aversive, reactionary racism. Although Trump made both latent dog-whistle (“law and order candidate,” “Make American Great Again,” etc) and overt racist pleas (banning Muslims and portraying Latin American immigrants as criminals), these voters predominantly voted for Obama in not just 2008, but also 2012. The racial backlash against the president argument and nativist appeals may have some import for this voting bloc, but it is no way the dominant explanation for the suburban switch to the GOP.
The states in which this was the primary cause of the shift are Pennsylvania and Ohio. Although Iowa is not generally conceptualized as a suburban state, the bellwether facet to this category does apply to the Iowan electorate.
Why did Black voter turnout decline so dramatically?
Although turnout was actually higher in absolute terms this election that 2012 (something pundits continue to get wrong), this aggregate trend belies group dynamics. I have yet to find a good metric for white or Asian turnout, but it is clear Latino turnout was up and African American turnout was down. Texas, Arizona, and Nevada all trended more heavily towards Democrats than expected, almost exclusively due to the rise of Latino mobilization. However, the gains in the West were more than offset by the losses in major Midwestern urban areas, such as Wayne County in Michigan, Milwaukee County in Wisconsin, and Cuyahoga County in Ohio.
The margin of victory for Trump in Michigan (+10,000) and Wisconsin (+25,000) were notably much smaller than the difference between Obama and Hillary’s vote share in the two biggest metropolitan areas. Turnout declined in Milwaukee County from just under 491,000 in 2012 to just under 430,000 in 2016. Notably, Hillary won the same 66% of the vote in the county as Obama in 2012. In Wayne County, total votes cast went from 814000 in 2012 to around 766000 in 2016. Hillary did receive less support than Obama by proportion (67% to 73%), but had turnout been the same as 2012, she would have carried both states.
It should be noted Black turnout was a bit higher in places the Atlanta metro area, greater New Orleans, and Philadelphia, and also lower in Southern Florida, northeast North Carolina, Wyandotte County in Kansas, Hennepin County in Minnesota, and Shelby County in Tennessee. So what explains this variance in Black mobilization, since a clear geospatial pattern is not immediately clear?
As of now, I have three leading hypotheses. First, the obvious: Obama was a Black male, while Hillary is a white female. The racial distinction is self-evident–plenty of social science scholarship has demonstrated all people, and especially African Americans, tend to show higher levels of support, trust, and efficacy towards officials that share their descriptive features. It is expected that Hillary would lag behind Obama for this reason alone. But importantly, Hillary’s gender could have been an impediment in the Black community, which is not exempt from chauvinism. However, it is unlikely that descriptive features alone explains such a steep, concentrated decline, let alone the regional variation of the decline.
The second hypothesis is Hillary Clinton’s immense unpopularity ended up mattering a lot more than Trump’s even higher unpopularity. According to Pew in late October 2008 Obama had a favorability of 60%, while Hillary in late October had 43% favorability. Specifically to the Black community, the ubiquitous feelings of support among voters for Obama during his initial election could only be made by Hillary if Hillary was Black. Other scholarship has shown the Black voters rally around Black candidates under fire, but Hillary’s whiteness impedes a similar steadfast support for her candidacy (Clintonian honorary Blackness notwithstanding). Clinton’s inability to craft a convincing message in decreasing police violence against African Americans, lack of attention to employment strategies, and lingering questions about her loyalty to egalitarian change could also help explain her lack of standing in the Black community. But this approach is a very coarse measure, and explains none of the variation manifest in the maps.
The third, and perhaps leading hypothesis, is the role of heavy-handed voter ID laws across the nation, but most concentrated in the Rust Belt states Donald Trump flipped.
Obstructive–and biased–impediments to vote, such as voter ID laws, tend to favor Republican causes, since the people most affected by ID requirements tend to be less economically secure, which is more common in the Black community, for immigrants, the poor, students, urbanites, and the elderly. Noticeably, only the very last constituency is even remotely pro-GOP. So if voter ID laws tend to decrease the Democratic electorate, and frequently African-American voters, is it possible these laws had some effect in this election? The answer is yes.
Most of the states with the most pernicious voter ID laws–Wisconsin, Michigan, Indiana, Ohio, Kansas, Mississippi, and Tennessee–saw a decline in Black turnout during this election. The only two exceptions to this rule are Georgia and Virginia, which makes some sense: the Clinton campaign spent a lot of time encouraging turnout in and around Atlanta, while Virginia has a Democratic governor that went so far as to pardon incarcerated members of the population with an important election in sight. Additionally, Texas had an increase in turnout, but it is unclear whether Black turnout was up with Latino turnout, or if the latter simply masked a decline in the former.
The only ways to truly understand whether voter ID laws played a role in the election, beyond the recognition of a pattern, is twofold. First, interviews with voters that can testify to the increased hardship in voting and interviews with those that did not vote and their reasoning. Second, a calculation of IDs issued based on demographics would confirm a bias in which groups failed to gain the necessary IDs to vote.
States that were most affected by the decrease in Black mobilization are Wisconsin, Michigan, and Ohio (although the latter would not have been a Clinton state even with more Black support–see suburban voter explanation).
Why did working class whites support Trump?
The most common explanation for the outcome is that disaffected white, working class voters with low educational attainment broke hard for Trump, and moreover, they turned out in high numbers to voice their displeasure with the status quo. Importantly, neither of these points are proven with the data, even if they are true. White voters made up 70% of the electorate in the election, their lowest number on record (down from 72% in 2012 and 74% in 2008). Moreover, according to exit polling Trump won about the same proportion of whites as Romney did in 2012 (Romney won 59%, Trump 58% with more third party candidate share). What about turnout? It is actually very difficult to find counties with mostly white residents that increased in turnout. For example, turnout was up by 2.5% in Hall County, Georgia, but although this is a heavily Republican county, whites only constitute 2/3 of the county population. Since both the Ds and Rs gained around 4,000 votes, it is not clear the white backlash thesis is correct.
So while it is plausible that uneducated whites in the Rust Belt feel unrepresented by both parties, leading many to favor the outsider candidate, the evidence does not show increased white support for Trump, either in the form of preferences or turnout. If anything, Trump’s 1-3% gain in vote share among Hispanic, Black, and Asian voters, and decreased turnout among Black voters, seem more consequential.
Trump did improve on Romney’s figure in regards to those that made under 30k a year–from 35% to 41%. If we assume many of these poorer voters were white (which is a hefty assumption), there is a good explanation for this change, especially in the Rust Belt. Trump preached isolationism and the false promise of being able to hedge job loss in manufacturing, and even restore many of these jobs. This plea to voters would be a successful frame, given the job loss in heavy manufacturing and lack of trade adjustment funding for jobs training with each successive free trade agreement. However, Trump will not succeed on this promise unless the labor force in the region is willing to work for 4 dollars (or less) an hour. Tax incentives are not enough to reverse private sector-led globalization and deindustrialization that has been occurring since the 1970s. The government can certainly facilitate the flight of business to other countries, but reversing that trend would require government-funded factory construction and jobs training, which there is little evidence any politician favors. The Democratic alternative–at least among the left side of the party–is economic diversification of the region, such as making education attainment higher in the region, jobs training in high-tech manufacturing, and increased demand-side stimulus to allow poorer citizens to use federal funds to redistribute wealth in their local economy.
This explanation is the most shaky for Trump’s victory. It is not clear there was an appreciable increase in poor white turnout in this election. Absent state-level exit polls in the Great Lakes, the data will not prove this point. If poor white voters were consequential in any states, it would have been predominantly the Great Lakes and Appalachia, both of which have been trending GOP for quite some time.
What about race?
With the immense amount of race-hate spewed by Donald Trump throughout the election, it is certainly plausible it activated, or made manifest, racism in the electorate. The endorsement by KKK and neo-Nazi groups of Trump, violence against racial minorities at Trump rallies, and rise in hate crimes are all reasons to assume race played a seminal role in this election. Moreover, the post-election ascendance of Steve Bannon to the top of the executive branch shows there is some work to promote white supremacy in government, if only descriptively and not substantively (although there is definitely room for both in a Trump administration).
However, the likelihood of either racial backlash against America’s first Black president or activated white supremacy by Trump being the defining facet of this election is very low. Not because it does not matter–which it obviously does–but because there are ample legitimate reasons to feel that the country is going in the wrong direction and that government is not responsive to the needs of the disaffected. Essentially, it is true some voters were mobilized by Trump’s white nationalist sentiment, but this race-centric theory fails to explain much of the suburban shift from Obama to Trump.
What can be stated about race is that racist rhetoric was not refuted by the public at-large or Trump supporters. Although we cannot know if Trump’s supporters were motivated by racism, we can safely say they were not deterred enough by Trump’s racism to vote for another candidate. This is kind of an odd dynamic, but in essence, we can say Americans are not racially liberal enough as a whole to rebuke a clearly racist political campaign. It should be noted there are probably many conservatives, like Mitt Romney and many in the Mormon community, that decided supporting another candidate (perhaps one less bigoted like Evan McMullin) was the thing to do. (Side note, I am a bit disheartened that so many Mormons decided to stick with Trump, as I suspected he would be the least attractive candidate of the main three Utah was considering.)
What about gender?
Gender is another important variable, especially with the disproportional negative media and political attention Hillary has experienced since the 1990s. She is perhaps the most scrutinized political figure in American history. Trump’s entire stamina critique of Hillary seemed to be about gender, veiled in her health episodes. It is safe to assume her gender is part of this, as is her relation to Bill’s promiscuity, her opportunistic position-taking and her email server issues. There are both legitimate and illegitimate reasons to be hesitant to support Hillary Clinton. As it turns out, Hillary’s support among men and women is nearly the same proportion as what Obama achieved in 2012 (the only big difference is the presence of third party candidates). Is it possible patriarchy is so ubiquitous in American society that women were self-policing and hyper-critical of Hillary in a way that would not occur if she were a man? Sure. Again, until we have enough interviews of women saying any woman is not well-suited to be president, we cannot posit out thin air that gender discrimination is a dominant explanation for what happened in this election.
What about immigration?
Immigration definitely mattered in this election, arguably more than race and gender. Many of the suburban voters claimed to be motivated by immigration related issues, such as building the wall or deporting undocumented residents. The anti-Latino sentiment of Trump clearly worked to mobilize Latino communities in Nevada, California, Texas, and Arizona. And unfortunately, Trump’s policy details are arguably most developed on the immigration question–a low bar, but true nonetheless. Luckily, it appears House Republicans are less apt for mass deportations than what many previously expected, so maybe a combination of meaningless fence construction and some path toward citizenship will occur. According to exit polling, those that claimed immigration was the most important issue supported Trump 64 to 32. However, among the four issue options, immigration was tied for last with foreign policy at 13%, while terrorism received 18% and the economy reached 52% (Hillary decidedly won the economy and foreign policy, while Trump won terrorism and immigration).
What about the urban-rural divide?
This is pretty clear as an important dynamic, with Trump setting records in rural areas. The resentment rural voters feel towards cities is palpable and somewhat justified: governments are located in cities, and government services are best delivered in cities. This dynamic often leaves the countryside feeling alienated from spending decisions, with little to show for their tax dollars. If the conversation ended there, an anti-establishment vote would be legitimate and easy rectify (show a new commitment to delivering services in the country). However, it is much more complicated, and heavily tied up into stereotypes and anachronistic notions of what modern governments should do. Many people in rural America have a skewed perception of what city dwellers are like. Sure they like lattes and ombre haircuts, but dependency on government support is not nearly as ubiquitous in cities as one might expect. In fact, the opposite is actually true: in what some term “red state socialism” many rural states receive more federal funds per outgoing tax dollar that do metropolitan states (New Jersey perennially getting the least for what they pay). Moreover, antagonism towards government in the countryside seems to deny the presence of social problems unique to cities that require collective governmental action, such as housing segregation, concentrated poverty, crime, and infrastructure maintenance. While the cultural divide between rural and urban folks is unlikely to get resolved, it is probably not a good sign to see the parties reshaped as metropolitan versus agrarian, as both geographic locations stand to gain from concerted government action to address the struggles in each environment.
Concerning this election, there is ample evidence that rural resentment of urbanites spurred support for Trump. There is some multicollinearity here, though, since race, class, partisan sorting, and ideology are correlated with settlement type, meaning it is difficult to ascertain the causal power of urban-rural cultural divide separate from those variables.
What about the media?
The media is culpable for this outcome in myriad ways. First, the unfettered coverage of Trump’s every move, from eating pizza with a fork and knife to taking a shit at 3am is a sign of the repugnant state of sensational, now tabloid, journalism. Making money is a necessary means to finance a news operation that allows for extensive investigative journalism, but money-making cannot be an end in itself. Trump should get a lot of credit for running a staff-less campaign and his innovative use of directly calling into news shows, but breaking regularly scheduled programing to cover one of his many rallies became gratuitous and transparently about ginning up the horse race.
Second, the lack journalistic push-back on Trump’s many false statements enigmatically fits in with the dominant to trend to draw a false equivalency on all sides of a debate. Hate speech cannot be covered as anything other than hate speech. Although the media was by-and-large critical of Trump, much of it was less fact-based and more focused on pot-shots and sensationalized quotes.
Third, the over reliance on tracking polls to explain dynamics on the ground directly contributed setting up high expectations for Hillary. There are examples of celebrities and journalists traveling around Michigan and perceiving it to be a Trump state, but the media did not seem aware of this sea change due to stable poll results and a lack of care for understanding Trump supporters.
A media that is solely concerned with ratings, and sanctimoniously dismisses a candidate that continues to beat expectations is a recipe for disaster. I do not think it is fair to claim the media should have known Trump could actually win–that is way to much to ask of anyone–but a more nuanced coverage of his bases of support would have changed expectations going into the election.
Did James Comey cost Hillary the election?
In the immediate days after the election, I would have summarily dismissed this claim. The polls showed very little movement beyond the pre-existing trend toward Trump because of the FBI reopening the investigation of Hillary’s emails. However, exit polling tells a different story about the effect of the very late announcement by Comey, which was only a week and half before an election. The weekend before the election the FBI concluded no further action would be taken on Hillary’s email scandal (although Anthony Weiner will surely be less lucky). Exit polling shows a trend: Clinton did better with voters that decided before September, while Trump did better in September and October. Importantly, those that decided to vote in the last week (after Comey reopened the investigation) supported Trump 50 to 38, while those that made up their minds in the last few days (after Comey cleared Clinton) supported Trump 46 to 44.
Hillary picking up supporters after the Comey clearance can really only be explained by two answers. It is possible as the weight of the decision to support one of the candidates became more salient, voters decided supporting Trump was a less responsible move than they had previously felt. Or, Hillary was gaining steam after a lackluster October and the emails derailed some of her “momentum.” Both could be true at the same time, but this pattern in the exit polling suggests the emails might have had some effect. In either case, Comey’s meddling in the campaign and lack of control of his own agents at the FBI (with all the leaks), suggests he has lost institutional support for his leadership. Therefore, Comey should resign effective immediately, since he is neither serving the public nor FBI interests, but is strictly looking out for himself.
While it seems likely the Comey fiasco had some effect on deterring support for Clinton, it is still unlikely that the margins are perfectly correlated with areas where she needed more support, like in the Great Lakes. Until we see evidence from voters in that region that the emails mattered on a large scale, this episode will remain a stain on the cycle, but not a determinate one.