I’ve found that in order to successfully cover politics — and campaigns in particular — it’s necessary to trust two things:
1) My own political instincts, so I can ignore “conventional wisdom” and figure out what’s really going on.
2) Outside data collected by experts, so I can quantify how voters think and feel about various subjects.
Balancing these imperatives — stats vs. gut, head vs. heart – is always extremely difficult. But in covering the 2016 presidential campaign, I found them to be irreconcilable. There was no convenient marriage of the two; a choice had to be made. And when ultimately forced to make that choice, I trusted data over instinct. This decision led me to reach the wrong conclusion about this election, but looking back, it seems unavoidable. In fact, given the sum of information that was available, it’s one I would probably make 100 times out of 100. Let me step back and put that in perspective.
Over the past two years I’ve traveled to every corner of the country and observed the Trump phenomenon in my professional capacity as a reporter. I’ve also witnessed it back home in Michigan with an informal focus group of friends and family members whose ideologies, cultural preferences, and political inclinations I’m familiar with. Wherever I went, whoever I’d speak with, the conversations were always pretty much the same: People were fed up with politicians, furious with both parties, and ready to stick it to Washington. Trump was their mechanism for doing so. They didn’t much mind that he was vulgar and provocative (even if they were self-identified Evangelical Christians); if anything his bluster reflected their exasperation and made him all the more appealing. This attraction was made possible by his celebrity status and relentless free media exposure, no doubt, but voters I spoke with also gravitated toward Trump because of his hardline, non-traditional positions on two critical issues: trade and immigration.
I’ve always liked the quip, “anecdotes are not data,” because it emphasizes that individual conversations aren’t necessarily representative of bigger trends. (IE: Meeting one black person who supports Trump should not inform broader conclusions about the black community’s feelings toward Trump; exit polls show he won eight percent of black voters nationwide.) That said, in the aftermath of last Tuesday’s election I have found myself remembering two years’ worth of anecdotes and regretting that I did not give them — and my underlying instincts that they spoke to — greater weight in my final analysis of the electoral map and of Trump’s prospects to win the presidency.
There are too many experiences to list here, but one looms large in my memory. A few days after Thanksgiving last year, my older brother got married in Michigan. The wedding was at a hotel in Ann Arbor — a liberal enclave of liberal Washtenaw County — but was attended by a cross-section of the electorate. There were older church-goers and younger party animals; white-collar automotive executives and blue-collar machinists; conservative Republicans and liberal Democrats. They had one thing in common: Everyone wanted to talk to me about their new obsession, the candidacy of Donald Trump. Had I met him? What did I think about his chances? Was he going to stay in the race? Could he win the nomination? Could he win the White House? And if so, could he really dismantle our existing free-trade agreements?
As I think back to those conversations, two things jump out. First, many of the conservative-leaning people who talked excitedly about Trump had no history of enthusiastically backing other politicians (at least not that I could recall.) Second, the liberal-leaning people who talked excitedly about Trump gave me a you-can’t-be-serious look when I asked them about Hillary Clinton. Almost uniformly they said they preferred Bernie Sanders, and suggested that if Clinton were the Democratic nominee they would not vote for her.
I remember feeling conflicted. On one hand, these sentiments squared with those I’d gathered during reporting trips across the industrial Midwest, particularly from white working-class voters. On the other, polling showed that Michigan — and Wisconsin, and Pennsylvania — wouldn’t be competitive for Republicans in 2016. These states were part of the Blue Wall, having been carried for decades by Democrats in presidential elections. No Republican could break through — especially not one with historically high unfavorable ratings, a lack of campaign infrastructure, and years of personal and professional baggage that Clinton could use to devastating effect.
I struggled with these incompatible assessments for many months. Even as Trump won New Hampshire and South Carolina and Nevada and a host of Super Tuesday states, demonstrating a diverse appeal and driving historic turnout, I resisted the impulse to articulate how he might — might — be uniquely suited to win the presidency not by taking the GOP establishment’s preferred route of winning Hispanics and suburbanites in Colorado and Virginia and Nevada, but by ending Democrats’ monopoly on the Rust Belt.
The dam finally broke on March 8. On that night, when Trump won Michigan’s primary — and Clinton lost in a stunning upset to Sanders — I wrote a story entitled, “Michigan Shows Trump Could Redraw Electoral Map vs. Clinton.” Some highlights from that piece:
Sixty-four electoral votes. That’s how many Mitt Romney fell short of the 270 required to win the White House in 2012. Sixty-four electoral votes. That’s the sum of four Rust Belt states Romney lost: Pennsylvania (20), Ohio (18), Michigan (16), and Wisconsin (10). Donald Trump says he can win back these states as the GOP nominee … Between his comfortable victory fueled by blue-collar white voters, and Hillary Clinton’s defeat thanks to her vulnerabilities with that same group, it’s possible that a November matchup between the two front-runners could drastically redraw the electoral map.
Trump won the state easily, taking 37 percent to Ted Cruz’s 25 percent and John Kasich’s 24 percent. According to exit polls, half of Republican voters there were whites without a college degree. Trump dominated among that group, winning 46 percent compared to Cruz’s 25 percent and Kasich’s 19 percent. On a key question for that demographic, a majority of all Michigan GOP voters, 55 percent, said trade with other nations “takes away U.S. jobs.” Trump won 45 percent of those respondents, compared to Cruz’s 22 percent and Kasich’s 20 percent. …
… Bernie Sanders, who trailed in many recent Michigan polls by more than 20 points, stunned the political universe by winning the state Tuesday, taking 50 percent to Clinton’s 48 percent and beating her by some 20,000 votes. A plurality of Michigan’s Democratic electorate, 36 percent, were whites without a college degree. Clinton lost those voters badly to Sanders, 58 percent to 41 percent. An even bigger majority than in the GOP primary — 58 percent — said trading with other countries “takes away U.S. jobs,” and Sanders won those voters by double digits, once again taking 58 percent to Clinton’s 41 percent. …
… And in a state with a traditionally sizable bloc of swing voters, another statistic stood out. Whereas Trump won a plurality of the 31 percent of self-described independents in Michigan’s GOP primary, Clinton was trounced, 71 percent to 28 percent, among the 28 percent of self-described independents who participated in the Democratic primary. …
… If there’s reason to believe Trump can compete in Michigan — a state Obama carried by nearly ten points in 2012 — then there’s reason to believe he can compete across the industrial Midwest. After all, Obama’s margins that year were even slimmer in Ohio (three points), Pennsylvania (five points), and Wisconsin (seven points), and those three states are demographically similar to Michigan. Their electorates are all overwhelmingly white (Michigan’s 2012 electorate was the most diverse of the four at 77 percent white), and they are all predominantly blue-collar (a majority of voters in all four states lacked college degrees in 2012). …
… Moreover, Republicans are counting on lower minority turnout this fall without Obama atop the Democratic ticket. Any meaningful drop-off would radically alter the electoral complexion of those four Midwestern states, each of which has major urban centers where minorities turned out en masse for Obama in 2012.
That article went viral the morning of March 9 and generated tons of attention in Republican circles. Most of it was negative bordering on antagonistic: Plenty of smart reporters and operatives whose work I respect made critical comments on Twitter or emailed me with dumbfounded reactions. Didn’t I know that Pennsylvania and Michigan and Wisconsin were part of the Blue Wall? Didn’t I know they were fool’s gold for the GOP? Didn’t I know that Republicans had a reputation for winning those states in midterm cycles but losing them in general elections when participation among low-propensity voters is greater?
Of course I did. And I had hedged my language in the story to acknowledge those realities. But to me, the math was simple: If Trump maximized turnout from his white working-class base, and Clinton failed to mobilize black voters in urban centers, Republicans could suddenly seize control of the Rust Belt. I felt comfortable publishing that argument — and dealing with the pushback — for one simple reason: It was no longer just my political instincts telling me Trump could pull it off; it was empirical evidence in the form of exit polls and primary votes.
As the campaign dragged on, however, that empirical evidence began to dissolve. I still had plenty of anecdotal fodder to suggest that he was singularly competitive in the Rust Belt. But the data wasn’t holding up. Trump was looking sufficiently strong in the state of Ohio. But his numbers at the time of the Republican convention were brutal in Pennsylvania, Michigan, and Wisconsin, and they only got worse a week later in the wake of the Democratic convention.
In late August, with Labor Day right around the corner, I wrote a piece examining how Trump’s prospects in those states had withered:
This year, it appeared that Trump, by virtue of his strength with working-class white voters, was uniquely positioned to put Pennsylvania in play for Republicans. But the numbers simply aren’t bearing that out. Every respectable poll taken since the conventions in late July shows Clinton with a hefty lead, ranging from nine to eleven points. …
… In four Michigan polls taken since the convention, Trump trails by nine, ten, ten, and seven points. What’s notable here isn’t the margins, but Trump’s unpopularity. In all four polls, he fails to reach 40 percent against Clinton — a stunning display of weakness for a major-party nominee in a supposedly competitive state. …
… Trump has visited Wisconsin several times, but Republicans don’t even bother to argue that he’s competing there. The state has been scarcely polled, but this month, the highly respected and historically accurate Marquette Law School survey showed Clinton leading Trump by 15 points (52 percent to 37 percent) among its likely voters. …
… Without a sudden and extraordinary change of trajectory, Trump is almost certain to lose Michigan, Pennsylvania, and Wisconsin, and he’s currently a slight underdog in Ohio. He did at one point appear to have a plausible path through these four states to the presidency, but that path has since disappeared.
What changed between early March and late August? My long-held instinct — that a Trump-Clinton matchup carried the unique potential to put the Rust Belt in play — remained in tact. But it was being drowned out by the data. There was little room left to trust anecdotal evidence of a historic pro-GOP shift among working-class whites coinciding with an enthusiasm drop-off among Democrats. At least, not with the polls showing a sizable Clinton lead in all three states.
Indeed, this must be emphasized: It wasn’t as though Trump trailed by three or four points in Michigan throughout the summer and stayed within striking distance in the fall. He was polling at 32 percent in early August and 31 percent in mid-October. Thirty-one percent. In a presidential campaign. Weeks away from Election Day. It’s true that the race tightened in Michigan in the run-up to November 8. But it’s also true that two major surveys taken less than a month out showed Trump losing by double-digits. And it wasn’t just polling that was on Clinton’s side. She supposedly boasted a sophisticated grassroots organization and voter-turnout operation that, in any normal election, should be worth two to three points in any given state.
The bottom line was this: Even with last-minute evidence that Michigan had become competitive (such as Clinton’s November 7 visit) there was nothing to suggest she would lose the state.
But I still couldn’t shake my instinct. I woke up on Election Day and re-read my piece from March 8 about the Rust Belt. I thought about the nationwide trend of black voter participation falling off in the first post-Obama election. And then I decided to write something putting it all in perspective:
Trump’s message on trade has a unique resonance in the state. That said, Michigan has been solidly Democratic for decades on the breadth and sturdiness of its electoral coalition: a supermajority of blacks, a good chunk of Metro Detroit’s college-educated suburbanites, and a decent share of the state’s sprawling class of white, non-college, blue-collar workers. There’s no question that Trump is going to overperform with that third group. The second group will likely drift somewhat leftward, though not nearly enough to compensate for the rightward movement of the third group. That leaves black voters, the cornerstone of the Democratic coalition, as the question mark. There’s no doubt they’ll vote overwhelmingly against Trump; the question is how many of them will vote, period.
I concluded with this prediction:
Black turnout drops enough to make things interesting, but not enough to give Trump a chance. He’ll come closer than Romney, but still lose the state by four to seven points.
I was wrong. Black turnout did drop enough to make things interesting; but it was the overwhelming support of the state’s white working-class that appears to have put Trump over the top. (Michigan’s final result is still being tallied; he currently leads Clinton by roughly 12,000 votes.) While Clinton suffered from a significant turnout decline in Wayne County (home to Detroit) and Genesee County (home to Flint), her fate was likely sealed in Michigan’s swath of mid-sized, blue-collar counties that were competitive in 2012 but went overwhelmingly for Trump in 2016. Look at the two maps:
It’s the same story in Wisconsin:
To watch the returns in those two states — and in Pennsylvania, which Trump also won and where similar dynamics were at work — was to be simultaneously shocked and unsurprised.
As a professional consumer of political data, it was stunning to witness Trump carry states where he’d never once led Clinton in a single nonpartisan poll of the general election horse-race. (It wasn’t just the public polls, either; senior Republicans say that while internal surveys showed the race tightening in those states, prompting Trump to invest more time and resources there, they never showed the Republican ticket actually winning.)
But as a middle-class Midwesterner who grew up around Reagan Democrats — and who knew former auto workers that had in their mid-forties begun waiting tables because their jobs had been shipped to Mexico — it wasn’t exactly staggering to see voters embrace a populist outsider running on a protectionist platform.
In the end, perhaps I was too deferential to big data and too dismissive of my own instinct. But as I wrote at the outset, if given the chance to cover this campaign all over again, I’m not sure how I — or anyone — could ignore the mountain of quantitative evidence pointing to Trump’s defeat.