Why I Disagree with Most National Polls: My 2016 Final Predictions

Published November 7, 2016

National Review Online

All in the Family and Maude were two of the most popular television shows in the early 1970s. Both were the brainchild of liberal producer Norman Lear, and each sought to argue — some might say propagandize — in favor of liberal politics. The main character in All in the Family was a working-class man from New York named Archie Bunker, while the eponymous lead in Maude was a suburban, college-educated woman from New York named Maude Findlay. Archie was ignorant, bigoted, and vehemently conservative while Maude was talkative, ostentatiously “liberated,” and vehemently liberal.

It doesn’t take a rocket scientist to realize that our current contest is, in the eyes of the national media, a rerun of this TV contest. Both candidates live in New York; indeed, Clinton lives part time in Chappaqua, N.Y., mere miles away from Maude’s fictional home in the very real upper-income Westchester County town of Tuckahoe. In this view, Trump is simply a more vulgar Bunker while Clinton is an ethically challenged Maude. And was the case 40 years ago, it’s clear today that those who try to determine such things want the rest of us to prefer Clinton/Maude to Trump/Bunker.

Back then, America didn’t go along with the elites’ judgment. Blue-collar whites loved Archie and made the show the most watched in America for an unprecedented five straight years. Maude debuted in 1972 to excellent ratings but could never dislodge Archie from his perch. After four top-ten finishes, Maude quickly slipped and was canceled after six years. All in the Family, meanwhile, never dropped below the twelfth spot, and a sequel featuring Archie alone ran for another four years with superb ratings throughout. Thirteen years after his first appearance, Archie Bunker still attracted tens of millions of viewers each week.

Donald Trump has spent the last month crisscrossing America and stopping at places that normally never see presidential candidates, like Eau Claire, Wis. All the off-the-beaten-track places where he’s campaigning are bastions of whites without a college degree, the modern-day equivalent of the people who flocked to Archie. He’s banking his entire candidacy on the idea that the Bunker family is still there and that, once mobilized, they can make the veteran TV entrepreneur the Number One show in the country.

Most national polls suggest that the race is over, with Clinton headed to a firm win. I disagree. I think it is likelier that she will win, but she will do so very narrowly in both the Electoral College and in the popular vote. It won’t take much improvement over my final projections for Trump to pull this out narrowly, and I think it is likelier that he wins than that Clinton wins by 3 or more.


Final Predictions

Popular Vote

Clinton 48 (range 46–48.5)

Trump 47 (range 44–48.5)

Johnson 3 (range 2–4)

Stein 1 (range 1–2)

Others/Write-ins 1 (range 0.75–1.5)


Electoral College

Clinton 278 (range 248–323)

Trump 260 (range 215–290)

States/Districts That Switch to Trump: IA, OH, FL, ME-2

States/Districts That Switch to Clinton: None


Senate Breakdown

R 51, D 49 (King and Sanders caucus with Dems)

States Switching to Democrats: WI, IL, PA

States Switching to Republicans: None


House Breakdown (off of 247 R, 188 D)

R 237 D 198 (D +10; range D +8–+15) (Babinec wins NY-22; caucuses with R)

Seats Switching to Democrats: FL-10 and VA-4 (redistricting); FL-13, NV-4, NH-1, CA-25, CO-6, FL-7, IL-10, MN-2, NV-3, NJ-5, PA-8, VA-10

Seats Switching to Republicans: FL-2 (redistricting); MN-8, NE-2, FL-18


Why Trump Will Close

Clinton leads the RealClearPolitics poll average as I write this on Sunday afternoon by a little more than 2 percentage points in the four-way race. Only the IBD/TIPP poll has him ahead (The LA Times poll having been dropped from the average), and four of the nine polls in the average have her ahead by 3 percentage points or more. So why am I so confident it will be a nail-biter?

It all comes down to what you think those who say they are undecided or voting for third-party candidates will do. We have years of evidence that even final polls overstate the number of people who actually vote for third parties by 33 to 50 percent; and this evidence shows that voters who back away from voting for third parties tend to break back to the party they normally favor. We also have years of evidence that undecided voters tend to break against the incumbent where there is one. Both factors favor Trump over Clinton.

Most third-party voters are either backing Gary Johnson or Evan McMullin, or they plan to vote for a write-in candidate. These voters tend to come from Republican-leaning backgrounds, and hence it follows that as they back away from throwing their vote away, they will return to the Republican nominee. In contrast, fewer than 2 percent currently think that they will vote for the Green Party nominee, Jill Stein. Even if she loses half of her vote, that will add less than 1 percent to Hillary Clinton’s vote share.

I also consider Clinton to be the de facto incumbent for purposes of this race. She has been in the public eye for more than a quarter century; she served in President Obama’s administration and clearly represents continuity with current policy. It’s reasonable to think that if voters aren’t already with her, they are likely to plunk down against her, as is the case in most races involving well-known incumbents.

In coming up with my final predictions, I needed to allocate these voters to the major-party candidates. We have had some similar races recently where polls indicated much higher third-party support than eventually turned out. Two prominent ones were the 2013 Virginia governor’s race (McAuliffe versus Cuccinelli) and the 2014 North Carolina Senate race (Tillis versus Hagen). In each case, a Libertarian was the primary third-party candidate in the pre-election polls, and in each case, the GOP candidate outperformed his or her final poll average by about 5 percentage points (e.g., Cuccinelli was polling at 40 percent but received 45 percent). This meant that between 70 and 75 percent of voters who had said they would vote third-party or were undecided but who did not end up doing so in fact opted to vote for the Republican.

I am assuming that 5 percent of 2016’s voters will cast a ballot for one of the third -party or write-in candidates, leaving 8 percent of the total vote to allocate to Clinton or Trump. Further assuming that 70 percent of that 8 percent vote for Trump narrows the race considerably.

I used racial breakdowns to arrive at my final numbers, as the exact share of the vote cast by whites and nonwhites will be crucial to determining the outcome. The five national polls for which I could find breakdowns for whites showed Trump winning 51–36. The four national polls containing breakdowns for nonwhites showed Clinton winning 71–16. Those polls estimated that whites will be 73 percent of the electorate, nonwhites 27 percent.

Assuming Trump gets 70 percent of the remaining 8 percent of whites and the nonwhites I estimate will not vote for a third-party candidate means that the race is effectively tied. Giving Trump 70 percent of the nonwhite share, however, seems too generous. It would also lead to his getting 22 percent of the nonwhite vote, significantly higher than Mitt Romney’s 19 percent. So I relaxed that assumption and gave him only half of the nonwhites to be allocated, giving him 20 percent of the overall nonwhite vote.

My final racial breakdowns, then, have Trump winning whites 57–38 and Clinton winning nonwhites 76–20. Applied to a 73–27 white/nonwhite electorate, this yields a 48.3 to 47 Clinton victory.

Democratic readers will object to the 73–27 breakdown. The exit poll showed a 72–28 breakdown in 2012, and nonwhites have gained an average of 2 percent in total electorate vote share every four years for the last 20 years. Democrats usually assume this will continue and that the 2016 electorate will be 71/­29 or 70/­30 white/nonwhite. Under this scenario, Clinton wins easily.

I don’t think that will happen this year. Half of the total 4-point increase in the nonwhite vote share between 2004 and 2012 came from record high turnout among African Americans. Without Barack Obama on the ballot, there is already strong evidence that black turnout is down. (This is also why Trump may slightly exceed Romney’s share of the nonwhite vote, as 94 percent of blacks voted for Obama.) Additionally, white turnout was down slightly in 2012 over 2008. Trump seems to have energized white voters (for and against, more about that later), and I think it likely that white voting will be up this time. Combining the two trends, offsetting some of the black voting decline with an increase in Hispanic and Asian voting as their voting-aged populations rise, means I think 73–27 is the likeliest electorate outcome.

Note how small changes in these assumptions can lead Trump to victory. Suppose he gets 1 percent more of the white vote, winning 58–38. That extra 1 percent gets him nearly 0.75 percent of the total electorate, nearly closing the gap. And if he wins 58–37, then he, not Hillary Clinton, wins.


Winning Downstairs, Losing Upstairs

Trump’s overall margin among whites appears to be quite similar to Romney’s, but that coincidence masks a huge difference. Trump is winning much higher shares of whites without a college degree than Romney did, but he is running behind Romney among whites with a college degree. Since the two groups each make up about half of the total white vote, Trump’s gains are balanced by his losses, keeping him from winning.

These factors are slightly changing the battleground map. Iowa has titled slightly Democratic in each of the last three elections, but its electorate is 93 percent white, with a tilt to those without a degree. Without nonwhite votes to prop her up and with Trump running away with non-college whites, Iowa looks set to swing sharply toward the GOP this year. Ohio also has a large share of non-college whites and very few Latinos or Asians. While Ohio is much closer, declines in black turnout and increases in white turnout will probably shift Ohio toward the GOP (Ohio was the only swing state where fewer votes were cast in 2012 than in 2008, with larger declines occurring in less-educated white areas).

This is also why I am calling Maine’s second congressional district for Trump. Maine gives the winner of each congressional district one Electoral College vote and Maine’s second district consists of the rural and less-educated area of the state. It is a bastion of white, working-class, and poor voters, and all polls show Trump holding firm. In the 2014 gubernatorial race, the second district firmly backed Maine’s Trump-like GOP governor Paul LePage, giving him all of his 29,000 statewide margin and then some. I suspect that the surge to Trump will occur most strongly in places like this, giving Trump a crucial extra Electoral College vote.

Florida and North Carolina do have significant shares of college-educated whites, but in each case there are large number of non-college whites to balance them. Each has a significant black population, and black turnout in each is down in early voting, sharply so in North Carolina. Increased Hispanic turnout in Florida is making up a bit for that, but if the white vote does break sharply for Trump, then Clinton will not carry either state.

That means the election will come down to Pennsylvania, New Hampshire, Colorado, and Nevada. Winning Pennsylvania or Colorado alone, or both of the other two, gives the White House to Trump. (If Trump wins Colorado plus ME-2 plus all the other states listed above, he has 269 electoral votes and the Republican-controlled House would select the president unless one of his electors defected to Clinton.)

In the first three states (Pennsylvania, New Hampshire, Colorado), however, the share of college-educated whites is higher than in the other states tilting toward Trump. These whites are also slightly more socially liberal, upscale, and less religious than those in Iowa, Ohio, North Carolina, and Florida. They live in the Boston, Philadelphia, and Denver suburbs, and they chose to settle there for a reason: Those cities’ cosmopolitan values better reflect their beliefs than do the more conservative and parochial cities that characterize the South and Midwest (perhaps Miami excluded). Polls show Trump badly losing the Philadelphia suburbs (where he must run roughly even to win), drowning out the record support he is winning from rural and small-town Pennsylvania. If Trump does get only 57 percent of the white vote, it will because he lost suburban voters who voted for Romney four years ago but just cannot stomach him.

Thus, I am predicting that Trump will narrowly lose New Hampshire by a bit over 1 point and lose Pennsylvania and Colorado by the slightly larger margins of 2–3 and 3–4 points.

Nevada is different, because its white population skews older and less educated. It has a burgeoning Hispanic population, however, and those voters are mainly of Mexican descent. They seem to be turning out in droves, and the polls suggest they are not in favor of the man who disparaged their nationality and wants to send illegal immigrants home. Blacks are only a tiny share of the Nevada electorate, so their drop-off in voting won’t affect tallies much, and Mormons (who notoriously do not like Trump) make up about the same share of the vote as blacks. Romney lost Nevada by 6.5 percent, and while Trump should close that gap, it is highly unlikely he can pull off a win.

Nevada and New Hampshire are also states with histories of poor public polling. In 2010 public polls showed that Republican Sharon Angle would defeat Democrat Harry Reid in the Nevada U.S. Senate race. Reid won handily, running nearly 5 points ahead of his final polling average. In 2012, President Obama ran nearly 2 points ahead of his final average, turning a projected 2-point win into the comfortable winning margin of 6+ points. There’s no reason to think public polling has improved, meaning it’s likely again to understate the Democratic advantage enjoyed among Spanish-speaking Hispanics.

Similarly, New Hampshire state polls understated the Democratic vote in both the 2014 U.S. Senate and the 2012 presidential races by about 2 points. In the latter case, the polls also overstated Romney’s share of the vote by nearly 2 points as well. I would not doubt that this is again the case, meaning that Clinton is ahead, not tied, in what will likely be the closest state she carries on Election Day.

Colorado also has a significant and growing Hispanic population, a miniscule black population, and a Mormon vote roughly of equal size to the black vote. These factors also have a similar effect on Trump’s chances — he won’t benefit much from black turnout drop-offs and should be hurt by Mormon defections and increased Hispanic voter turnout.

Trump is trying to make up ground by campaigning in Michigan in the last week, but this has the feel of a desperate move. Michigan polls have shown Clinton well ahead all year, and it is consistently slightly more Democratic than Pennsylvania is. Trump may do very well indeed in rural and small-town Michigan and in depressed auto towns, but he will lose votes in wealthy Oakland County and the college counties of Ingham and Washtenaw. With the one exception of 1976, when native Gerald Ford headed the ticket, Michigan always votes with Pennsylvania: It would be a miracle if Trump were to break that pattern, which had held for more than 70 years.

Trump thus suffers from the opposite problem that sunk Romney. Romney did well among upper-income and educated whites but failed to excite non-Evangelical less-educated whites. Trump excites the latter but repels many among the former. Somewhere there’s a winning formula in here. Anyone see what it is? Bueller?


The Senate and the House

The House is relatively easy to predict. Democrats were unable to place a large enough number of districts in play to effectively challenge the control of the House. Despite many boomlets in press stories about the Democrats’ chances, no neutral observer has ever predicted that Democrats would gain anywhere near the 30 seats they need for a narrow one-seat majority. Democrats could gain as many as 20 seats if they have a much better night than I, or most other observers, expect. As it is, their gains will come mainly in seats favoring them that Republicans captured in fluke campaigns, and in some more educated marginal seats where sentiments are likely to run against Trump.

The Senate is harder because many of the states in which Republicans are defending their seats are battlegrounds. My rule of thumb for most of the year was that GOP incumbents could run only so far ahead of Trump, because very few voters split their tickets any more. Indeed, in most states, the polls have borne this out: GOP incumbents have run behind when Trump has been running poorly, and ahead when he has been running well.

The most recent polls support this observation, as no Republican candidate still with a competitive race is running 4 or more points ahead of Trump. New Hampshire’s Kelly Ayotte, Wisconsin’s Ron Johnson, and Marco Rubio lead the pack by running 3 or more points ahead of Trump, while Pennsylvania’s Pat Toomey and Nevada’s Joe Heck are running within a point in either direction of him.

I am therefore calling the Senate races based on how I think each state will go. Clinton will carry Illinois by a large margin, dooming Senator Mark Kirk to a sizeable defeat. Clinton also looks likely to win Wisconsin by more than 4 points (Trump is stopping in neighboring Iowa and Minnesota, but not in Wisconsin, in his last 48-hour cross-country push), dooming Johnson to a narrow defeat. Toomey is also unlikely to win unless Trump carries Pennsylvania, something that looks unlikely.

That drops the GOP to a narrow 51-49 advantage. Nevada is the only seat the GOP can pick up, and that looks to be a nip-and-tuck race. Since I think Clinton will prevail, however, I am calling this seat for Democrat Catherine Cortez-Masto.

Rubio has been leading in the polls since he unexpectedly re-entered the race and national Democrats pulled out of here weeks ago. He might not win pretty, but a 3 or 4 point win is still a “W.”

Ditto with recent close races in Indiana, Missouri, and North Carolina. In the latter two races, Democrats mounted late challenges against vulnerable incumbents, but in both cases the Republican recovered and is now leading. Trump will carry Missouri easily, and I think he will retain North Carolina, and nothing suggests that either GOP incumbent will run far enough behind Trump to lose. Former senator Evan Bayh made his own late entry into the race, returning to his home state and putting what was thought to be a safe Republican incumbent seat in play. Months of incessant hammering, however, have taken their toll on the moderate Democrat, and he now trails slightly in the most recent poll. Trump is likely to win Indiana by 10 points or more: Bayh remains very popular, but he won’t run 11 points ahead of the top of his ticket.

That leaves Senate control up to the New Hampshire Senate race. Ayotte has worked very hard to maintain a less partisan image, and despite tens of millions in attack ads tying her to Trump, she has consistently polled further ahead of him than any other incumbent. She will have a very close race, but so long as Trump loses by 2 points (or less), she should squeak through.


Rebuilding Reagan’s America

Much will be written about how Republicans and conservatives should respond to the outcome regardless of what it is. If Trump does win, he will effectively have launched a hostile takeover of American politics, hijacking one political party to his ends without owing anything to it. If he loses, his backers will probably blame recalcitrant Republicans for the defeat: Had they remained “loyal,” they’ll say, Trump would have beaten a very vulnerable Clinton. This new “Trump faction” will in all likelihood try to influence or take over the GOP or at least defeat a number of prominent anti-Trumpers. The GOP’s civil war is apt to get much bloodier before a winner emerges.

But it doesn’t have to be this way. Barry Goldwater’s nomination was similar to Trump’s in many ways. He mobilized discontented Republicans and independents to shock an establishment used to getting its way — many prominent officeholders refused to endorse him. Tempers were high after Goldwater’s defeat, and it seemed as if the GOP would either split or move sharply to the left in reaction.

It did neither, largely because all sides in the civil war cooled down the rhetoric. Establishment types decided not to purge the Goldwaterites. Goldwaterites such as William F. Buckley also changed their tune: They had refused to support Republicans who were insufficiently conservative, but henceforth, Buckley — who refused to back Eisenhower in 1956 or Nixon in 1960 — urged conservatives to support the most conservative electable candidate.

Ronald Reagan echoed that call in his own incomparable voice. He told conservatives that “conservative, moderate, and liberal” were merely labels foisted upon Republicans by their adversaries. All brands of Republican had more in common than any of them had with liberal Democrats, he said, and they must first unite and put their differences behind them if they were to retake their country. When he declared for the California governor’s race in 1966, this most prominent of Goldwater backers had hired Nelson Rockefeller’s political consultants to guide his campaign, and he made party unity, not revenge, his hallmark.

Reagan’s brand of unity also extended to those who were not Republicans. He reached out to Democrats and independents who shared what he called “individualist” values, telling conservative activists that someone who agrees with you 80 percent of the time is your potential friend, not an enemy.

The positive results were immediate. Reagan won in a landslide, uniting the upscale, traditional Republicans who had backed Rockefeller with the middle-class conservatives who had backed Goldwater. Millions of working-class white Democrats and independents who had backed LBJ just two years earlier voted for him. Reagan’s appeal even extended to the normally Democratic-voting Mexican-American precincts in Southern California. Campaigning on what united these voters rather than dividing people according to ideology, class, or race, Reagan set in motion the forces that would culminate 14 years later in his landmark election to the presidency. All he did then was change his country and the world.

How he did that and what lessons today’s conservatives can learn is the subject of my upcoming book, so I won’t spill all the beans now. But as we all contemplate how best to move forward, we should recall two of Reagan’s less-well-known sayings. He wrote one in these pages in December 1964 in a short, handwritten piece analyzing the Goldwater election. “Human nature resists change,” he wrote. “And it goes over backward to avoid radical change. Time now for the soft sell to prove that our radicalism is an optical illusion.” Yes, the man who preached “only bold colors, no pale pastels” argued that conservatives needed to lighten up if they were to win over the American voter.

The other saying is his epitaph, the last words he chose to have us remember him by. One might expect it to be one of his paeans to freedom or warnings about the evils of government. One could not be more wrong. Inscribed on his gravestone, resting in the shade of the California oak and warmed by the Pacific sun, Reagan’s last words were these:

In my heart I know that man is good, that what is right will eventually triumph, and that there is worth and purpose to each and every human life.

Americans of all ideological stripes want someone who can say these words with complete sincerity and show how they can live their lives in today’s America with purpose and dignity. The conservative who recovers this Reagan is the person who can unite the Reagan coalition again and once more put Americans all in the same family.

— Henry Olsen is a senior fellow at the Ethics and Public Policy Center, an adjunct professor at Villanova University, and the author of the forthcoming book Ronald Reagan: New Deal Republican. 

Most Read

This field is for validation purposes and should be left unchanged.

Sign up to receive EPPC's biweekly e-newsletter of selected publications, news, and events.


Your support impacts the debate on critical issues of public policy.

Donate today