Not the only “fundamentals”-based model to show a likely Republican victory this year. Alan Abramowitz’s “Time for Change” model also pointed to a probable GOP win based on things like the state of the economy, the president’s job approval, and whether the party in power has been there for two terms or just one. Abramowitz disowned his own model’s prediction this summer, though, on the theory that it only explains whether a generic Republican should defeat a generic Democrat or vice versa. When you have a nominee who’s not at all generic, who in fact much of the public deeply dislikes and has deemed unfit for office, then don’t be surprised if that affects the results. In other words, the “Time for Change” model doesn’t say who will win the election so much as which party should have won it. And therein lies the tragedy for Republicans in 2016.
This model, called the “Primary Model,” looks at different factors than “Time for Change,” although the amount of time that the incumbent party has been in power is relevant in both. Unlike Abramowitz, though, the man who developed it is sticking to his guns.
Helmut Norpoth has been predicting a Trump victory since early this year. His model currently projects a win for the Republican with a certainty of 87 to 99 percent…
Instead of opinion polling, Norpoth relies on statistics from candidates’ performances in party primaries and patterns in the electoral cycle to forecast results. The model correctly predicted the victor in every presidential election since 1996, according to the Daily Mail.
Running the model on earlier campaigns comes up with the correct outcome for every race since 1912, except the 1960 election.
Here’s what Norpoth wrote a few days ago at the Hill:
To start with something basic, opinion polls are really about “opinions,” not actions. At their best, they can tell us how people feel about political issues and personalities. Do voters, for instance, like or dislike candidates such as Hillary Clinton and Donald Trump?
Yet having an opinion and acting on it are two different things. Barely 6 in 10 voting-age American citizens turn out for presidential elections. Ascertaining the opinions of 100 citizens is just a start. Now you have to determine which 60 of them actually take the time to mark a ballot. They are the “likely voters.” They are the only ones that count. But to find them is no easy chore.
It is ingrained in all of us that voting is civic duty. So nearly all of us say, oh yes, I’ll vote, and then many will not follow through. Miscalculations of which respondents will turn out to vote can easily screw up a poll prediction, so would it be sure thing if we just dealt with actual voters?
He’s suggesting, in other words, that turnout on Election Day might look very different from how most pollsters seem to believe that it will. Maybe many Americans who only tepidly support Hillary will stay home and/or many who strongly support Trump but don’t usually vote will show up. That argument should be familiar to you by now — it’s the “undercover Trump voters” theory, that there are lots of people out there planning to turn out for Trump this year who aren’t showing up in polls for whatever reason. Either they’ve been left off of voter lists used by pollsters because they’ve registered to vote only very recently or they’re embarrassed to tell pollsters they’re supporting Trump so they lie and say they’re for Hillary or Gary Johnson instead. Anything’s possible, but pollsters are aware of the possibility that Trump will bring out “unlikely voters” who haven’t voted much in recent years and have spent the campaign trying to incorporate them into the data. The Clinton campaign has a sophisticated data operation that’s desperate to identify these people, I’m sure, so that they know how to allocate their resources in the final few weeks. If there are 200,000 white rural votes sitting out there for Trump in Pennsylvania that none of the public pollsters know about, Team Hillary will be doing everything it can to uncover them and then adjust its ground game and ad spending accordingly. Needless to say, with so many pollsters now pointing to a comfortable Clinton win, to miscalculate the number of likely voters so badly that a six- to seven-point polling advantage suddenly disappears on Election Day would be the most catastrophic failure in the history of American polling. A lot of statistical brainpower has gone into averting that humiliation and getting turnout expectations right.
In fact, the first inklings from early voting point more to the polls being right than wrong:
More than 3.3 million Americans have already voted. And among that group, Democrats have improved their position in North Carolina, Nevada, Arizona and even Utah compared to this point in 2012…
Democratic early turnout has stayed steady in North Carolina compared to 2012, while Republicans have dropped by about 14,500. In Nevada, Democrats have a smaller early voting deficit today than they did at this point in 2012. And Democrats are slightly ahead in Arizona in the early vote so far, though they are lagging Republicans in the tally of how many Arizonans have requested ballots.
Perhaps most surprisingly, Democrats improved their position in conservative and Mormon-heavy Utah, where recent polls have shown a tight race. At this point in 2012, Republicans led Democrats in early voting by more than 22,000 voters. But so far this year, the GOP advantage is only 3,509.
By Nate Cohn’s estimates, Trump has a five-point lead in North Carolina among absentee voters right now, which sounds good until you know that Mitt Romney had a … 33-point lead at the same stage of the 2012 election. Trump is way off the traditional pace. The early-voting numbers look much better for him in Ohio and Iowa, but that’s consistent with the polls too. Those two states have been his strongest battlegrounds all year. His path to 270 begins with winning those two, but he’d also need Florida and North Carolina to have any shot at the presidency — and as noted here, the early numbers in NC don’t look good.
As for the “Primary Model,” if I understand the methodology correctly, Norpoth is essentially extrapolating each nominee’s relative strength from how they did in the New Hampshire and South Carolina primaries. Trump won both comfortably but Clinton got blown out in NH before coming back big in SC. The idea, I guess, is that the candidate with the most enthusiastic base wins in November. Two problems with that, though. One: The New Hampshire primary was basically a home game for Bernie Sanders, who lives next door in Vermont. Hillary looked weaker there than she did in most other Democratic primaries this year because of his regional advantage. Two: Although both parties have partly fractured over their nominees this year, most polls at this point show Clinton winning more Democratic votes than Trump winning Republican ones. He’s the one dealing with the deeper party schism, especially since his “Access Hollywood” tape and the sexual-assault accusations started trickling out. I’m not sure how the “Primary Model” accounts for that. Trump’s supporters may be more motivated to vote than Clinton’s supporters, but if Clinton’s party — with plenty of “soft” supporters — is more motivated to vote on balance than Republicans are, what does that do to the model? And what about the data that shows, er, that primary turnout means nothing for the general election?
Incidentally, Norpoth notes in his column as an argument against trusting the polls that (a) the exit polls in 2004 were famously wrong in suggesting a 51/48 Kerry win and (b) Gallup was famously wrong in 2012 when it predicted Romney as the winner in its final national poll. Both of those points are true — but only partly. The preliminary exit polls in 2004 did indeed go bust, but the final exit polls saw a Bush win and got very close to capturing the final margin. Gallup’s failure four years ago is well known, especially to conservative bloggers like me who clung to their numbers as evidence that Romney had a real chance. But that’s why data nerds are constantly nudging people not to focus on individual polls but rather on poll averages. You can’t put too much stock in any one pollster, even one as esteemed as Gallup, since their sample may produce an outlier by chance or its turnout model may be flawed. The poll average, which incorporates many samples and many different turnout models, is likely to be closer to the eventual result. The nerds will also tell you to follow state polls more closely than national ones since, after all, Election Day is 50 separate state elections, not one national election. And in fact, the state polling averages were almost perfect four years ago in calling the outcome, pointing to the correct winner in every state except Florida. Saying that all polling is unreliable because of Gallup four years ago seems to me a bit like saying that you can’t trust the stock market because Enron went bust. The answer to that problem is diversification, not hoarding money in the mattress.
Incidentally, thanks to Marco Rubio’s poor performance in the New Hampshire and South Carolina primaries, the “Primary Model” gave Hillary Clinton an 86 percent chance of defeating him in the general election. Show of hands — who thinks Trump will finish with a higher share of the vote next month than Rubio would have?