Pages Menu
TwitterRssFacebook
Categories Menu

Posted by on Feb 25, 2016 in 2016 Elections, 2016 Presidential Election, Featured, Politics, Science & Technology | 14 comments

Professor’s nearly infallible predictive model: Trump wins general election

Donald_Trump_in_Reno,_Nevada_January_2016

A political science professor’s model of predicting elections that has a 96.1 accuracy forecasts Republican Donald Trump will be elected President of the United States in the 2016 elections.

Stony Brook University’s Professor Helmut Norpoth’s system for forecasting elections has only been wrong once — in 1960. And this time it’s saying Trump has a 97 percent chance of winning the White House if he’s the GOP nominee and faces Hillary Clinton, and a 99 percent chance if he faces Vermont Sen. Bernie Sanders gets the nod.

Norpoth created a statistical model of presidential elections that uses a candidate’s performance in their party’s primary and patterns in the electoral cycle as predictors of the presidential vote in the general election.

Donald Trump has a 97 percent chance of defeating Hillary Clinton and a 99 percent chance of defeating Bernie Sanders in the general election, according to Norpoth’s formula.

“The bottom line is that the primary model, using also the cyclical movement, makes it almost certain that Donald Trump will be the next president,” Norpoth said, “if he’s a nominee of the [Republican] party.”

Norpoth’s primary model works for every presidential election since 1912, with the notable exception of the 1960 election. These results give the model an accuracy of 96.1 percent.

I’m actually not surprised one bit.

From the day Trump announced the bulk of analysts said he didn’t have a chance of getting the nomination and certainly not of winning. I’ve pointed out that if you read history and biographies history — and this includes entertainment history — is filled with examples of people who were underrated, ignored, laughed at, considered fringe, predicted to go no where, and they in fact came out on top. Then all the former conventional wisdom so smugly written and broadcast by people giving their take through their own political filters and assumptions were swept under the rug.

In fact, if someone is running for office for a major political party no matter what logic, trends and facts may say at the time, saying they will never get a nomination or will never be elected is just that. An assertion that could be proven wrong.

More details:

Norpoth began the presentation with an introduction of the potential matchups in the general election, including a hypothetical Sanders vs. Trump general election.

“When I started out with this kind of display a few months ago, I thought it was sort of a joke.” Norpoth said referring to Trump and Sanders, as many alumni in the audience laughed. “Well, I’ll tell you right now, it ain’t a joke anymore.”

As the presentation continued, laughter turned to silence as Norpoth forecasted a 61 percent chance of a Republican win in the general election.

This forecast was made using the electoral cycle model, which studies a pattern of voting in the presidential election that makes it less likely for an incumbent party to hold the presidency after two terms in office. The model does not assume who would be the party nominees or the conditions of the country at the time.

“You think ‘This is crazy. How can anything come up with something like that?’ ” Norpoth said “But that’s exactly the kind of equation I used to predict Bill Clinton winning in ‘96, that I used to predict that George Bush would win in 2004, and, as you remember four years ago, that Obama would win in 2012.”

Norpoth then added data from the New Hampshire and South Carolina primaries to narrow down the forecast to specific candidates. As he brought up the first slide with matchup results, the silence was broken by muttering from the audience.
And it gets more definitive, according to the model:

“Trump beats Hillary 54.7 percent to 45.3 percent [of the popular vote]. This is almost too much to believe.” Norpoth said, with a few members of the audience laughing nervously. “The probability of that [outcome] is almost complete certainty, 97 percent. It’s almost ‘Take it to the bank.’ ”

The primary model predicts a Trump victory with such certainty due to Trump’s relatively high success in the Republican primaries, Norpoth said. Clinton, in comparison, is in an essential tie with Sanders in the Democratic primaries. As a result, Sanders would also lose to Trump in a similar landslide if Sanders were to be the Democratic nominee, Norpoth said.

In contrast, Norpoth forecasted that a hypothetical presidential race with Ted Cruz or Marco Rubio on the Republican ticket would be a much closer race. The results showed Clinton with a 55 percent chance of winning the race against Cruz or Rubio with a 0.3 percent lead in the popular vote.

So the jokes and dimissive new and old media columns will continue about Donald Trump, now edging towards considering him a serious candidate who could get the nomination.

The Republican estabalishment and Democrats underestimate at his peril.

And it might be wise for many pundits to add a few journalistic hedge words about how he’ll never be elected or this or that group will never support him.

Could this model be wrong?

Yes.

What are the odds of it being wrong given its track record?

Low.

Photo: by Darron Birgenheier from Reno, NV, USA (Donald Trump in Reno, Nevada) [CC BY-SA 2.0 (http://creativecommons.org/licenses/by-sa/2.0)], via Wikimedia Commons

Click here for reuse options!
Copyright 2016 The Moderate Voice
  • Slamfu

    Question. When did he start applying this formula publicly? I know he says it predicts most races back to 1912, but obviously he wasn’t using it BEFORE the 1912 race but after, and a 20/20 hindsight predictor is obviously going to be right more often than not 🙂

    But if he’s been making public estimates since the 90’s BEFORE the chips fall that have proven accurate, then it sounds like he’s onto something.

    • Well, he claimed that he predicted Clinton’s reelection, so I would guess he’s been at it since at least 1996.

  • My wife and I had this discussion last night…not about the science and predictive modelling, but about whether Trump could win the general. She can’t imagine it; I can; she can’t believe that I can; and from that our discussion proceeded. My points included that Trump presents himself as a strong charismatic leader who commands the camera, and that he crosses over to appeal to independents and Democrats, much like Reagan.

    The one caveat to both my thinking and the professor’s modelling is Trump’s likely inability too attract any meaningful support from minorities. The professor’s model may take this into account; I don’t know.

    • Slamfu

      Not to go all Godwin on this, but I often marveled at fact that Hitler was elected to office. He wasn’t born into his leadership role, he got there from very humble beginnings. And he wasn’t shy about his policies either, he told the German people what he was all about and still managed to get popular support and get into power. That always scared me because I wondered if it could happen here.

      Now we have Trump, a man no where near as bad as Hitler, but one who nonetheless is being very up front about wanting to do some very authoritarian, racist, and heavy handed policies that are clearly bad ideas, but yet appeal to many millions of Americans. I realize that he is largely only winning 1/3rd or so of GOP voters over right now, which themselves make up just over a quarter of registered voters, but still the fact he’s gotten even this far is pretty scary considering what he’s been saying. Add to that fact we have been skipping down the path these last 15 years of doing nothing while our most important Rights are negated, makes me wonder just how far an authoritarian demagogue can go in turning this nation into a Fascist state. By law, we are largely there already, our leaders just haven’t been abusing their power to the full extent that they could. But history has shown it’s only a matter of time until we have leaders that will.

      • Hitler was not elected to power. He lost the only election he ran in. He was appointed by Hindenburg and then gobbled up the late president’s (Hindenburg’s) power.

        The Nazi Party held about 37% (?) of the seats in the Reichstag but Hitler would not accept a “coalition” government. Thus President Hindenburg, due to the power vacuum, asked Hitler to create a government w/o the normal 50% support of the Reichstag.

    • JIM SATTERFIELD

      I have to agree with your caveat. While some people point out that Trump has some supporters in minority communities it’s not much at all. In fact I feel it’s safe to say that he will have less Hispanic support than any recent candidate. Support in the black community isn’t going to be very high either. Also I wonder if his model can account for factors like the one pointed out in this article from Vox.

      http://www.vox.com/2016/2/23/11099644/trump-support-authoritarianism

  • His model is largely based upon the difficulties for one party to hold the White House three times in a row.

    There are two reasons to think that maybe this won’t apply this year. There is the extremism of the Republican Party which might prevent them from winning when they might have historically–especially if they have a candidate like Trump. Plus if the Democrats nominate Sanders he can be portrayed as a real change and not just a third term for the same parties and policies.

  • JSpencer

    I am baffled by Trump’s appeal, but clearly he is charismatic to some, and as much as it pains me to admit it, charisma probably is probably a greater factor in presidential races than substance. I sure hope Norpoth is wrong, because (like many) I believe a Trump presidency would be catastrophic in a multiplicity of ways – including the rest of the civilized world having their worst fears about Americans confirmed. I definitely need another cup of coffee now – in part to prep for digging out after this snow storm. I’d post a pic of my view out the back door, but there seems to be no provision for that anymore.

  • rudi

    He called Obama in 2008, but his margin of victory wasn’t even close.
    2008

    — In the earliest completed forecast made in January 2008, Helmut Norpoth’s (Stony Brook University) “Primary Model” uses candidate support in presidential primaries to predict the general election two-party popular vote outcome. Norpoth’s forecast makes Senator Obama the favorite by a razor-thin margin, predicting a 50.1% to 49.9% Obama victory, but also indicates only a 50% chance that Obama will gain a majority.

    2008 Results
    Never heard of Helmut Norpoth. A Google search of TMV only shows two hits for HN.

  • Markus1

    I have a model that 100% predicts every one of the previous fifty Super Bowls. It’s predictions about the future that are hard. I think that the professor should put up some real money at two to one odds; being 96+% sure makes it a lock. I think that all prognosticators especially professors of economics should be required by law to put up 50% of their annual salary whenever they make a prediction; for journalists make it 75%.

    • KP

      Most professional bettors are average handicappers but decent shoppers for the best odds. Markus1 gets it. You may give your 2cents about who you think will win an election or super bowl, but to make that bet you may have to give 3 to 1 odds or give up six points. Whatever make sit a 50/50 proposition.

      A 55 percent winning history under those conditions is impressive. That would put the person among the very best professional handicappers in the world; that means showing a profit over time.

      This guy may or may not be on to something, bit it’s not gambling.

  • The Ohioan

    From what I can tell, he is predicting the Republican candidate will win, not Trump per se. Any name could be plugged in and be the predicted winner. So if a comet hits the Republican debate tonight, Christie or Romney could win the election. On the other hand, I would expect Bush to break the model’s winning streak, such as it. His model predicted Gore in 2000…

    I could be wrong.

  • StockboyLA

    I think his model uses the complete primary results from previous elections. But what if he were to go back and use only the first couple of primary results in all the previous elections. Does he still have the same degree of accuracy? In other words we don’t have all the primary results for this election so instead of saying his model is 96.1% accurate, he should use data at the same point in time before each election to determine the accuracy this early in the season.

    Did his Feb 2000 model predict Bush? Or did his prediction of a Bush win Coke after all the primary results. In this article there is mention of who won the popular vote in his model. There is no mention of who won the electoral vote. Did his 2000 model predict Bush would win the popular vote and the presidency? Because Bush LOST the popular vote by 500,000 or 600,000 votes. He only became president because of the electoral vote. So if his 2000 model predicted a Bush popular vote win, it was wrong (even though Bush became president, so he can sorta claim to be right). If his model is about popular vote, then the model’s accuracy is lower. (Though still impressive). But again – back to my first point – I think his model predicts the winners after all primary / caucuses. I don’t believe his model is 96.1 % accurate with just a couple states’ primaries wrapped up.

  • It is very easy to create a model that overfits with any data and enough variables.
    Lets say you want to predict house prices and have information on 10 recently sold houses (actual price the house is sold, size of lot, room count, region etc).. It is very unlikely that these variables would be exactly same for any of the houses. So that you will have a unique correspondence of these 6-7 variables to your outcome (the actual price the house is sold). Therefore you can come up with a formula that is 100% accurate for those 10 houses you have. However, this doesn’t mean when you want to predict the 11th house you would be accurate. What you have is an overfitting model.. and in real life overfitting models never work..

Twitter Auto Publish Powered By : XYZScripts.com