NY professor predicts Trump for President

[JohnClay]

Get ready for President Trump, says election whiz who’s scary accurate

A New York professor whose formula has proven accurate in every presidential election but one since 1912 says Donald Trump has a 97.6 percent chance — or better — of taking the White House if he’s the Republican nominee…

…One major assumption is that a party that has held the Oval Office for two consecutive terms is less likely to win another term.

Norpoth applied his model, which he created in 1996, to every presidential election going back to 1912, and it worked for every one — with the exception of 1960, when John F. Kennedy narrowly defeated Richard Nixon…

[GIGObuster]

Its deja vu all over again…

Univ. of Colorado predicts Romney win by a landslide!!

Univ. of Colorado study predicts Romney wins by a landslide!!!

Forcasting model developed by to CU Boulder professors has accurately predicted the winner of every POTUS election back to 1980.

"Well they were wrong then weren’t they!"

[John_Mace]

Except most of the primaries haven’t been held yet. Was there a “Bigfoot Discovered!!” article right next to this one?

This is the NYPost, remember.

[JohnClay]

@John_Mace

Except most of the primaries haven’t been held yet. Was there a “Bigfoot Discovered!!” article right next to this one?

This is the NYPost, remember.

It says “if he’s the Republican nominee”

[CarnalK]

@JohnClay

It says “if he’s the Republican nominee”

I will totally vote for BigFoot regardless of his platform.

[JohnClay]

Professor gives Trump as a 97 percent chance of beating Clinton in Nov

Professor Helmut Norpoth’s statistical modeling gives Trump a 97 percent chance of beating Clinton in November - if he’s the GOP nominee

He’d have an even better shot against Bernie Sanders - 99 percent

Norpoth’s model has accurately predicted the winner of every national election since 1912, The Statesman reports, except the election of 1960

If either Rubio or Cruz were the GOP nominee, the model shows them losing to Clinton but winning against Sanders

[Measure_for_Measure]

A New York professor whose formula has proven accurate in every presidential election but one since 1912 …

There’s a red flag. Presumably the predictions are “Within-sample” predictions. Meaning you use the results to… predict the results. It’s a valid technique, but the probabilities that pop out tend to overstate matters.

More red flags:

No discussion of underlying methodology.

Election prediction suffers from small sample problems. One way of getting around that is to predict the margin of victory. There’s no evidence the professor did that and no discussion of the same.

Morale: Don’t go to the tabloid press for discussions of science. Especially when there are better sources like 538, Sam Wang and Ray Fair.

[Steve_MB]

These predictions have a 100% accurate track record!

[John_Mace]

@JohnClay

It says “if he’s the Republican nominee”

But his algorithm involves knowing the results of the primaries, which he can’t know at this point. If he’s claiming to predict who wins based on who gets the nomination, then he’s crazy.

[GIGObuster]

@Measure_for_Measure

There’s a red flag. Presumably the predictions are “Within-sample” predictions. Meaning you use the results to… predict the results. It’s a valid technique, but the probabilities that pop out tend to overstate matters.

More red flags:

No discussion of underlying methodology.

Election prediction suffers from small sample problems. One way of getting around that is to predict the margin of victory. There’s no evidence the professor did that and no discussion of the same.

Morale: Don’t go to the tabloid press for discussions of science. Especially when there are better sources like 538, Sam Wang and Ray Fair.

QFT and I have to add another red flag as I pointed in another thread about this (It seems to be very popular among Trump’s followers, but it is not good as I explain):

I have to say that besides that bit of news being close to nut picking, the point I make here is that like with Romney in 2012 I have the strong feeling that “news” like this one become viral among supporters of what in reality is the underdog; because many other more serious sources don’t have such rosy scenarios of victory for Trump.

In reality, based on current polls and projections, this bit of “news” is like going to the bottom of the barrel; and articles like the one quoted by the OP actually tell us that many of the sources from the right are having virtually nothing good to offer about Trump’s chances in the general election so this will have to do to keep the minions happy in the meantime.

[NeonMadman]

@JohnClay

…One major assumption is that a party that has held the Oval Office for two consecutive terms is less likely to win another term.

Norpoth applied his model, which he created in 1996, to every presidential election going back to 1912, and it worked for every one — with the exception of 1960, when John F. Kennedy narrowly defeated Richard Nixon…

Maybe I’m reading this wrong, but I don’t see how Kennedy was an exception to the principle stated in the quotation. Kennedy, a Democrat, won the White House after Eisenhower, a Republican, had held it for 8 years.

It seems to me that the exception would be George H. W. Bush (R) being elected in 1988 after two consecutive terms of Reagan (also R).

Am I missing something?

[Chronos]

Presumably that’s not the only factor the model uses, and the other factors (whatever they are) favored Nixon enough to overcome that.

I’m curious, though, what his model would “predict” for 2000. Any model that predicted that Bush would win in 2000 was wrong. Any model that predicted that Gore would win in 2000 is also wrong. The correct prediction for any model for 2000 would be “too close to call”.

[Martin_Hyde]

I had read into this when it was mentioned in another thread, I think people are being a little unfair to the professor here because of the low quality of reporting in the Daily Mail. He’s been running his model for a long time, and if you follow these models most of these guys aren’t claiming that they know who will be the next President, they’re just claiming they can develop a model, that when applied to historical data, generates results similar to what actually happened.

His model is called the “Primary Model” and isn’t based on the entirety of the primaries, it’s based on modeling the results of the early primaries and a few other factors to see if there’s a way you can run that and get somewhat accurate predictions for the general.

And his model has not accurately predicted every election since 1912. For example in 2008 his model predicted that Hillary Clinton would beat John McCain in the general election, but that did not actually happen.

Here’s some information from his site:

candidate as two separate predictors. For elections since 1952, the primary-support measure relies solely on the New Hampshire primary, while using all primaries prior to that date. The model is estimated with data from presidential elections going back as far as 1912, the first year of presidential primaries, with an adjustment applied to partisanship for pre-New Deal elections. In addition to primaries, the forecast model takes advantage of a cycle in presidential elections that appears related to the two-term limit of the American presidency. The out-of-sample forecasts of the PRIMARY MODEL pick the popular-vote winner in all but one of those 24 elections.

With the New Hampshire Primary decided, the model is capable of making a final forecast for any match-up in November between Democratic and Republican candidates. In the race between the two primary winners, Democrat Hillary Clinton would edge Republican John McCain by a narrow margin: 50.5 to 49.5 percent of the two-party vote. Barack Obama would win by the narrowest of margins over McCain: 50.1 to 49.9 percent. Needless to say, both of these forecasts are well within one unit of the forecast standard error.

You can read more here.

FWIW the model predicts % of popular vote, so the idea that it just attempts to predict winner/loser is not accurate, either. He also counts his prediction in 2000 as being correct, because it was only .1 off of being accurate. His model predicted very close to the actual vote share Gore would receive–it’s not actually a model that predict the electoral college at all (so doesn’t really predict the election outcome per se), but instead is designed to try and predict the popular vote. So his model correctly could predict to very close accuracy that Gore would win the popular vote and by what percentage, but it didn’t predict (or attempt to predict) the outcome of the electoral college.

[Measure_for_Measure]

@Martin_Hyde

I had read into this when it was mentioned in another thread, I think people are being a little unfair to the professor here because of the low quality of reporting in the Daily Mail.

Thank you for digging up the original source material. Overall, I would say that the Daily Mail provided a near-crackpot presentation of the underlying science. I think some of the author’s work is problematic, but I don’t think it’s nutty. It was just presented that way by the tabloids.

Here’s the first paragraph of the paper, from your link:

[INDENT][INDENT]American elections in November are typically preceded by “primary” elections earlier in the year. So is the voting in presidential primaries a leading indicator of the vote in November? Remarkably so, as it turns out. How well presidential candidates do in primary elections foretells their prospects in the November election with great accuracy. [/INDENT][/INDENT] I’ve skimmed the paper. I have not read it.

1. That’s a plausible story.

2. The primary system only really began in 1972, so proper treatment of data before that is tricky. There were primaries, but they were considered beauty pageants at the time, since they didn’t select any delegates to the convention.

3. I’m not sure whether Table 2 represents the complete regression. But if it does, I’m a little concerned that it doesn’t control for the economy. One of the problems with small sample sizes is that outcomes tend to be overdetermined. Which means you want to control for lots of things. But that tends to destroy your significance.

3a. In effect you need to study a regression table with lots of columns, which is not provided in the link.

4. The author presents out-of-sample results, which shows his seriousness. He does this by including data both before and after the election being forecast though, which tends to overestimate his accuracy. For example, we’d get better predictions of the 2016 results if we had data from 2020 and 2024. But we don’t.

5. Again, I have only skimmed the paper: I haven’t read it. I opine that there’s some real science there, but it would take a while to uncover. As in more than the 5 minutes I spent with the paper.

[scudsucker]

I learnt about this predictive method today, of course from a crowing right winger - it - back on March 2nd - predicted the chances at 90% Trump / 10% Biden.

The Primary Model

FWIW the website does, despite Martin_Hyde saying

@Martin_Hyde

it didn’t predict (or attempt to predict) the outcome of the electoral college.

does actually predict the electoral college 362 / 176.

I think Mr. Norpoth might like to revisit the criteria for his model after this disastrous year.

[Hari_Seldon]

Just remember that without the Comey surprise, this prediction would almost certainly have been wrong.

[bobot]

Agreed. Also, Hi, Zombie! Also, also, I am encouraged by the immense early voting turnout. When the people speak (and their votes are counted) Republicans lose elections.