538's 2016 Primary Forecasts

[Trapsy]

The odds and polls for presidential primaries and caucuses, updated daily. What do you guys think?

http://projects.fivethirtyeight.com/election-2016/primary-forecast/iowa-republican/

[Asian Nazi]

IT BEGINS

[Penelope]

"polls-plus"

No need to argue in the polls forum, guys, I see Nate Silver has decided to unskew the polls for us! 

[Ljube]

Trump only 39% to win Hew Hampshire? And Hillary the favorite? What a hack!

[This account no longer in use.]

> "Polls-plus"
Hillary: 73% chance of Iowa win
> Polls-only
Hillary: 55% chance of Iowa win

[Senator Cris]

I appreciate that they released both models. Everyone can believe his preferred model.

[The Free North]

Isnt Silver the guy who said Trump would fade by September?

[/]

omg go away nate silver

[Orthogonian Society Treasurer]

Trash analysis.

[#TheShadowyAbyss]

LOL the Hack is Back!

[Ebsy]

I guess we'll see.

[Enduro]

I guess we'll see.

[Statilius the Epicurean]

Rubio 18% chance to win Iowa?

As they say in America: trash it.

[Lief 🗽]

Hahahaha what total garbage.

[An American Tail: Fubart Goes West]

I guess we'll see.

[Trapsy]

I think you guys should be a bit happier that they released two different models.

[This account no longer in use.]

No. The fact that Silver is trying to unskew the polls is enough to make me put this in the traaaaaaaaaaaaaaaaaaaaaaaaaaaaaysh.

[Wells]

What is a "polls-plus"?

You must be logged in to read this quote.

State Polls? Obviously.
National Polls? Um, what do national polls have todo with individual primaries?
Endorsements? Are you kidding? Who cares about that? I mean, I don't. (If you do, tell me.) And I won't unless Obama or somebody like that endorses someone.
Let's compare results:
Republicans
Iowa (Polls Plus): Cruz (49% Chance), Trump (28%), Rubio (18%)
Iowa (Polls Only): Cruz (42%), Trump (42%), Rubio (9%)

New Hampshire (Polls Plus): Trump (39%), Rubio (23%), Cruz (13%)
New Hampshire (Polls Only): Trump (55%), Rubio (12%), Cruz (9%)

Democrats
Iowa (Polls Plus): Clinton (73%), Sanders (27%)
Iowa (Polls Only): Clinton (55%), Sanders (45%)

New Hampshire (Polls Plus): Clinton (53%), Sanders (47%)
New Hampshire (Polls Only): Sanders (73%), Clinton (27%)

Overview: So, the "Polls Plus" model appears to underestimate Trump and Sanders. I haven't read this yet, but I will to see if I can make sense of their formula for "Polls Plus".

[cinyc]

Rubio 18% chance to win Iowa?

As they say in America: trash it.

As the only establishment candidate really trying in Iowa, Rubio has a small chance of pulling off an upset and winning the state.  Problem is, Chris Christie is also starting to advertise there, so Rubio's small window may be closing.

[Eraserhead]

Oh wow, more anti-Trump/anti-Sanders crap from Nate. Shocker.

[Wells]

     Okay, I read the article, and they make their polls plus sound very reasonable. They use national polls differently than I expected, candidates with lower numbers get a greater chance to win than higher numbered candidates in close races. What?
     And which do they think is better? Polls Plus or Polls Only?

You must be logged in to read this quote.

     So both are close in accuracy? Why the big discrepancy in NH (amongst Dems)? I still don't get why endorsements are there (they think endorsements are important).

No. The fact that Silver is trying to unskew the polls is enough to make me put this in the traaaaaaaaaaaaaaaaaaaaaaaaaaaaaysh.

     I don't see him unskewing polls. I don't know where you get that from, he's not -

You must be logged in to read this quote.

     Oh. Why, Nate Silver, why?
     Also, their Iowa models factor into their NH models, and their NH models factor into their other models. Cause the first primaries affect the race, I guess.

[Figueira]

"polls-plus"

No need to argue in the polls forum, guys, I see Nate Silver has decided to unskew the polls for us! 

This is AFAIK the same method he's always used, that let him predict 2012 so well.

That said, we'll never really know if he's "right," since he's just listing probabilities, not definite outcomes. I find both the mainstream worship of Nate Silver and the disdain for him on Atlas to be equally annoying.

[Mr. Morden]

     Okay, I read the article, and they make their polls plus sound very reasonable. They use national polls differently than I expected, candidates with lower numbers get a greater chance to win than higher numbered candidates in close races. What?
     And which do they think is better? Polls Plus or Polls Only?

You must be logged in to read this quote.

     So both are close in accuracy? Why the big discrepancy in NH (amongst Dems)? I still don't get why endorsements are there (they think endorsements are important).

Endorsements are there because they looked back at past presidential primary races, and found that endorsements are positively correlated with candidates doing well in the primaries, even after you control for a candidates' poll numbers.  Same with the national polls.  Lower national poll numbers correlate positively with doing well in the primaries, once you control for the state poll numbers.  (That weird factoid is discussed here: http://fivethirtyeight.com/features/to-win-in-iowa-or-new-hampshire-it-may-be-better-to-poll-worse-nationally/ )

Basically, the two models are 1) the simple model, "polls only", which just looks at one thing (the state polls), and uses it to the model the primary result, and 2) the more complicated model "polls plus", which adds in the extra variables, namely endorsements and national polls.  The latter model is by definition more accurate in past primary races, because adding in extra variables you can always get a better fit.  But by adding in the extra variables you also run the risk of overinterpreting weak correlations from years past, and assuming that they're going to continue into the future.  So it's kind of a matter of taste as to which one you regard as "better".

[Mr. Morden]

Rubio 18% chance to win Iowa?

As they say in America: trash it.

Not sure why that's unreasonable, given the late surges of candidates like Santorum, Kerry, Edwards, etc.

"But those were good candidates, and Marco Rubio is never going to match their success."

Well, OK, but this model isn't trying to do a subjective "candidate quality" analysis.  It's just looking at these variables, and telling you what to expect in this race given what happened to previous candidates who were in a similar position.

[Leinad]

we'll never really know if he's "right," since he's just listing probabilities, not definite outcomes. I find both the mainstream worship of Nate Silver and the disdain for him on Atlas to be equally annoying.

^I agree with this a LOT but don't have much else to add and annoy myself when I empty-quote stuff.