Friday, November 11, 2016

Nobody Knows Anything

It was legendary screenwriter William Goldman who said, famously, of Hollywood, "Nobody knows anything." Tuesday's Trumpquake suggests the same is true of America's punditocracy.

Like most other supposedly keen political observers, I delivered the worst prediction of my long career when I wrote in this space that Hillary Clinton would be elected president. It was only the third time in a dozen tries that I've gotten a presidential election wrong, but if I were you, I'd never want to hear my opinion again. So you're excused from reading the rest of this blog, without prejudice or penalty.

If you are still here, I do have a few thoughts. observations and nuggets of data (similar data were proven worthless Tuesday, so take these with a giant land mass of salt) to share.


That's a question I've been getting a lot, and one that I admit I've been asking myself, since I didn't predict it would. Trump's victory didn't take me by quite as great a surprise as it did others; I did predict almost a year ago that he would be the Republican nominee, provoking tremendous consternation and derision among many of my readers. But I also have believed for at least four years that Hillary Clinton was a prohibitive favorite to be our 45th president, and I didn't alter that assumption even in the face of mounting evidence to the contrary.

It's been clear for a long time that much of the American electorate remains restless, angry, disenchanted. Yes, the economy is, empirically, a far sight better than it was when President Obama took office. Unemployment has fallen from almost ten percent to less than five. Instead of losing as many as 800,000 jobs a month, the country is creating as many as a quarter million. The U.S. economy is growing at about three percent a year, instead of plunging four percent, as it did in the spring of 2009. The Dow Jones average has tripled, meaning Americans with 401ks and other retirement and investment accounts have seen them more than recover from the Great Recession. But there are millions of people—no, tens of millions—who feel left behind by the recovery, whose paychecks are stagnant (or non-existent), who struggle to pay their bills and keep their credit card debt in check, who can't afford the still-rising cost of their health insurance premiums and their medications and their kids' education. They see factories close and jobs disappear and Mexican and Chinese and Somali and Puerto Rican immigrants arrive in their towns and they feel like they've lost their place in line, that someone else is stealing their shot at the American Dream, that they've been told to sit down, shut up, pay their taxes and take their lumps. They see the coastal elites with their app-driven lives and cars from the future and organic meal deliveries and wonder why everyone but them seems to be moving ahead. They live in places like Lycoming County, Pennsylvania and Manitowoc, Wisconsin and Hancock County, Ohio and Mariposa, California. The Democrats who live in these places voted for Bernie Sanders in the primary, not Hillary Clinton. And on Tuesday, most of the people who live there voted for Donald Trump.

Most of them don't like Donald Trump. They don't admire his behavior. They don't think he's the kind of man who should occupy the Oval Office. But they don't think much of Hillary Clinton either. These are not people who vote party, or even policies. They vote personality. They're not sure what Trump's beliefs on most issues even are (is anyone?), but the ones they do know about, they probably don't share. It doesn't matter. They're dissatisfied. They want change. They've wanted change for years, decades even. They voted for Barack Obama in 2008 and 2012, because he was new and different, and a more compelling personality than John McCain or Mitt Romney. They voted for George W. Bush in 2000 and 2004, because he was from Texas, not Washington, had some folksy charm and was more engaging than Al Gore or John Kerry. They voted for Bill Clinton before that, for similar reasons. And Ronald Reagan before him. And Jimmy Carter. You get the idea.

I interviewed quite a few white folks who voted for—even volunteered for—Barack Obama, but this time, pulled the lever for Trump (or punched the hole, or filled in the arrow, or touched the screen). That is hard for liberal Democrats to comprehend. How could anyone—anyone—help elect the first black president, a man considered so progressive that his opponents slurred him as a socialist, and then turn around and vote for Donald Trump, a man who questioned Obama's legitimacy, aligned himself with white supremacists and called for a roundup of Muslim Americans and illegal immigrants?

It's too easy for outsiders to dismiss these voters as crazy, racist, sexist, xenophobic, illiterate misogynists. You know, your standard basket of deplorables. But even Hillary Clinton recognized that many of them are not that at all. Some have college degrees. Some went to graduate school. Many are women. Some are people of color. What they share is a sense that things still aren't getting better, that the country is moving in the wrong direction, and the people in charge just don't get it.

Even so, there weren't quite enough of them to elect someone as reprehensible as Donald Trump. That's not my characterization of him: the exit polls (if you still believe in any polls of any sort, which I'll get to below) reveal that 60% of the voters think Trump is not qualified to be president, and 63% say he doesn't have the right temperament for the job. Even a quarter of those who voted for Trump feel this way. So how in the world did he win? Because Hillary Clinton failed to inspire, and motivate, and mobilize, the voters everyone assumed she would turn out to win. I've seen some articles talking about the "Trump wave" that we journalists missed. There was no Trump wave, at least not nationally. Trump is going to end up with slightly more votes than Mitt Romney got in 2012, about 61 million or so. That's what we expected him to get. We anticipated an electorate of roughly 130 million people, and that it would take somewhere close to 65 million votes to win.

What we did not expect—and no one on the Clinton team did either—was that Hillary would not even come close to getting as many votes as Barack Obama did. Obama won in 2008 with a record-shattering 69.5 million votes, more than seven million more than George W. Bush got four years before, when he set a new record. Obama's support fell in 2012, when he won with about 66 million votes, to Romney's 61 million. Clinton is on track to finish with somewhat more than 63 million this time. She should end up about 1.3% ahead of Trump in the national popular vote. We all thought there would be a massive turnout of women to elect the first female president, and a surge of Latinos to keep the wall-building Trump out of the White House. Neither materialized. The female share of the electorate actually fell one percent from 2012, and Clinton did only slightly better among them than Obama did. She outpolled Trump with women, 54-42%, while Obama beat Romney among women 55-44%. Only one percent more Latinos voted this time than did in 2012, and Trump did better among them than Romney did (Trump won 29% of the Latino vote vs. Romney's 27%). Clinton didn't come close to matching Obama's appeal to Latinos—Obama won 71% of Hispanic voters, to Clinton's 65%. White women actually preferred Trump, with 53% voting for him and only 43% voting for Hillary. We knew Trump would beat Clinton among whites overall, but he did it by a record-setting 58-37% margin. As expected, Clinton did not do as well with black voters as Obama did, but we thought she might make up that difference by attracting new Latino voters and widening the gender gap. It didn't happen.

Another myth: Trump brought out many new voters, and also won over the middle-of-the-road folks. Nope. Only ten percent of the voters were first-timers. And, tending to be younger, they voted overwhelmingly for Clinton, 56-40%. And self-described moderates preferred Clinton, 52-41.

So, she won the fence-straddlers in the middle, and she won the new voters, so why did she lose? Well, first of all—she didn't, at least not in the purest sense. Clinton won a plurality of the votes, and in fact, will have the widest margin of victory in the popular vote of anyone who didn't win the presidency since Samuel J. Tilden in the notorious disputed election of 1876 (I wrote a term paper on that one in college, if anyone is interested in digging that up). But she didn't meet turnout expectations in a few key places: Detroit, Philadelphia, Miami, Milwaukee. Even though she won in the cities, 59-35%, and Trump narrowly carried suburban America, he thumped Clinton in rural areas, 62-34. A relatively small number of white male Obama voters in those places I mentioned above gave Trump the narrowest of margins in three key states—Wisconsin, Michigan and Florida—and that gives him an Electoral College victory. At this hour's counting, Trump carries those three states by a total of 59,000 votes, or less than one percent. Toss in Pennsylvania, where Trump won by about one percent, and we're talking about less than one-tenth of one percent of the total votes cast Tuesday determining our next president. If Clinton had turned out literally another 0.1% of voters in the right cities, she'd have won those four states (technically, Michigan is still too close to call, with Trump ahead by less than 12,000 votes), which would have given her 303 electoral votes.

So, as you can see, it's not that Trump won this election as much as Clinton lost it. For all her vaunted advantages in fundraising and ground game and party support, she couldn't overcome the perception among white voters in rural America that she's a dishonest, lying, corrupt part of an elite political establishment that doesn't care about them and takes them for granted, if it thinks about them at all, a perception driven home in the campaign's final ten days by Trump's relentless ads portraying her as exactly that, and by FBI Director James Comey's untimely (for Clinton) announcement that he was investigating additional emails that might incriminate her, which ultimately amounted to absolutely nothing. Voters didn't turn to Trump in massive numbers in the closing days. They simply turned away from Clinton, enough to deny her victory in a handful of battleground states (Trump did do significantly better than Romney in Michigan, Pennsylvania and Florida, but Clinton surpassed Obama's totals there, too, because the overall turnout hit record levels in those three states. In Wisconsin, Trump matched Romney's vote total, while Clinton fell about 200,000 short of Obama's).

Although I do think it's risky to write off the Trump Voter as uneducated, Trump did score an historic edge among voters without a college degree. In 2008, Obama and Romney essentially split college graduates, and Obama won among those without their degrees. This time? Clinton won the educated voters, 52-43%, while Trump won the less educated ones, 52-44. And among white voters who didn't finish college, Trump won resoundingly, 67-28. Those are unprecedented gaps.

So this election, like virtually every election without an incumbent president in recent times, was about Change. New. Promise. Different. When the first exit polling data came in Tuesday afternoon, what did voters say was the quality they were looking for most in their next president? "Can bring needed change." Uh-oh. Alarm bells went off at our KCBS Election Desk. That was not a good harbinger for Clinton. Two-thirds of the voters told the exit pollsters the country was on the wrong track. Yikes. That boded well for Trump. It wasn't until later that we got more data: A whopping 83% of those who said "bring needed change" was most important, cast their ballots for Trump. Even though Clinton won 66% of the people who said "has good judgment" was most important, and 90% of those who were looking for someone with "the right experience," there weren't as many of those voters in the mix. Which brings us to one final question...


Here, I posit the theory, supported by the monograph's worth of data I just laid on you, that we didn't, really. That's not a copout. I said Hillary would win the popular vote by more than five percent. I am an idiot. But the final polls all coalesced around a Clinton win by about 3 or 4%. I know when people hear, "Clinton will win by four percent" and she wins the popular vote by only one percent, they think, "Boy, the pollsters really blew it." But polling is rough science. Pollsters extrapolate results from small samples. Their sample sizes can be off. The way they weight the data they collect to reflect what they think the makeup of the electorate will be can be wrong.  But even so, missing the final popular vote by 2 or 3 points is well within most polls' margin of error. A poll that says Clinton will win by three points means in fact, she could win by six, or it could be a tie, or the final tally could fall anywhere within that range. Which it did.

So where the pollsters (and I) really did blow it was in the electoral vote. But again, as noted above, Trump won that by the slimmest of margins, by denying Clinton a fractional percentage of votes in states you can count on one hand (I suggest using the thumb for Michigan). And really, the only places the polls were off significantly were Wisconsin, Michigan and Pennsylvania. The final four polls in Florida had Clinton up two, Trump up three, and two were a tie. Trump won by about one percent, right in the middle of that spread. The polls in Wisconsin had Clinton up six to eight points, but no one bothered polling there in the last week of the campaign. The needle was already moving to Trump—or, more accurately, away from Clinton—but there was no one on hand to record it. The same was true in Michigan, where Clinton was consistently ahead by four or five points and it ended, essentially, in a tie. Perhaps in 2020, the pollsters will conduct more surveys, and later ones, in the key battleground states.

The other way we blew it was by how we read the polls we did get. Most political pundits, and even most people within the Clinton and Trump campaigns, were pretty sure Hillary was going to win. All the data, until the Comey letter came out on October 28, said the race was over. My interactions with voters in Ohio convinced me Trump would win there, and the polling supported that. If I had seen consistent survey data showing Clinton ahead in Ohio, I would have reconsidered my conclusion. But I didn't. So I put the Buckeye State in Trump's column in my final prediction (I remain astounded by how many pundits said Clinton would win Ohio. There were literally zero data to support that). But the polls were steady for Clinton in Wisconsin, Michigan and Pennsylvania. There was no reason to suspect a Clinton collapse there. So when we looked at the final polls, both nationally and in swing states, we tended to believe the ones that supported what we already believed to be true: that she would win. This bias reinforced a flawed conclusion. We disregarded polls that deviated from our expectations. They must be outliers. They can't be right. Given a choice between surveys that bolstered our preconceptions and ones that challenged them, we embraced the former and scoffed at the latter. This wasn't because we wanted Clinton to win; it was because we thought she would. I didn't want Trump to win the Republican nomination, but I predicted he would because all the available data, and my own observations of voters, told me that he would. In this case, even though I had ample contact with Trump supporters, I didn't believe there were enough of them to overcome what I thought was the larger universe of Clinton voters, a universe that collapsed into a black hole for Democrats on Tuesday.

I called three states wrong on Tuesday (four, if Michigan ends up a Trump state). I got two of the U.S. Senate races wrong. That's not bad, I suppose, but it falls far short of my usual standard, and bottom line, I PREDICTED THE PERSON WHO LOST WOULD WIN. Please allow me to hang my hat on this caveat that I included in Tuesday's prediction:

There could be a hidden pool of Trump voters who aren't showing up in the polls. There could be a Trump Effect, with respondents afraid to tell pollsters they're secretly planning to pull the lever for Donald. There could be a huuuge turnout of white men without college degrees who descend on polling places en masse to Make America Great Again.

As things turned out, that's exactly what happened, and in just enough critical places where Clinton failed to make up the difference.

Okay. That's a really long explanation of how Trump pulled off the biggest upset in modern political history, and how I missed seeing it coming. I underestimated the level of antipathy for Clinton, and I overestimated her ability to get her voters to the polls. The lesson for Democrats next time? Nominate someone who can run as an outsider, as a legitimate agent of change (a la Barack Obama or Bill Clinton), especially if President Trump (anybody used to saying that yet?) fails to deliver. The people who switched to him this time won't be patient with him for long. Certainly not as patient as you've been with me.

Final note regarding the exit polls: You'll notice I treat this data as if it were handed down on stone tablets. I am always amused by how we deride the pollsters for being so off the mark before an election, and then regurgitate the exit polling data as if it were stone cold fact. It isn't. It could be wrong too. But—there are some critical differences. Pre-election surveys are based on an estimate of what we think the electorate will look like, a best guess of who's likely to vote, and then a weighting to reflect the expected demographic breakdown. Exit polls are a measure of people who have actually voted. We don't have to guess how many Latina women over 50 will vote; we can count the ones standing in front of us coming out of the polling place. And the sample size is much larger. Instead of a survey of one or two thousand people on the phone, the National Election Pool (a data-sharing consortium of which CBS is a member) interviews more than 100,000 people, mostly in person. The response rate is much, much higher, and the margin of error much, much lower. People could still lie about how they voted, but they're less likely to do that in a face-to-face interview. This data is how we call the results in states long before the votes are counted. When the returns start to come in, if they hew closely to what the exit polling suggests they will be, we can deduce that the actual result will be what the polls say. If a state is very close, the polling data may not be enough, delaying a call. This is also how we knew something was amiss in the "butterfly ballot" counties in Florida in 2000. The exit polls said Al Gore had won Florida, because voters told the pollsters they'd voted for him, when in fact many of them had cast ballots for Pat Buchanan by mistake. So when their votes were counted, they weren't for Gore, confounding the exit poll data.

There can still be weighting errors in these polls of course, and people who refuse to answer can throw things off, but in general, we believe the exit poll data to be fairly reliable. Of course, we thought Hillary Clinton would win Pennsylvania, too.

No comments: