How accurate is polling in terms of predicting election outcome?

[ssunlimited]

Right now Trump is winning NH and overall. It seems if its true that Trump will win the nomination if it was now(Though he is doing terribly against Ms. Clinton). But REALLY... IF it was election day today or around this time would people TRULY vote for Trump and have Trump win or is it not over until its over(even if election happened today)? If so, then what is the point of exit polls? They do those in order to know how people have voted so I guess its not over until its over(until people vote)? Can many voters(like millions) TRULY change their mind by the time they actually cast their vote?

[CPUd]

When these polls get released, you have to look at the entire poll to get a better picture, not just the summaries. One of the polls this week showed 52% of the respondents do not even have a #1 or #2 preferred candidate. You can also look at the favorables, sometimes they are broken down like "favorable", "unfavorable", "never heard of", "no opinion". In those numbers, you will see Hillary and Trump get around 1 or 2% "never heard of" and Jeb around 10%. The other candidates are 20% or higher. If you sort by that column it is relatively close to the rankings they give on page 1. As other candidates get more name recognition, their favorables tend to polarize in one direction or the other.

Undecideds are still in the 15% range in the national polls, but are going down. Some people (maybe a lot) don't make up their mind until they are in the voting booth and look at the ballot. The week or so leading up to election day, the campaigns have a really good idea what they will get, but any election can have its surprises.

[ssunlimited]

Hmm I wonder... This got me thinking!

[Ronin Truth]

Prediction is very difficult, especially if it's about the future.
Niels Bohr

[acptulsa]

How good a poll is depends on the methodology they use. They have been using decent methodology close to election day to make themselves look good. At other times, they'll ask things like, 'Who would you trust in the job whether you want him to have it or not,' and pretend like that's a measure of ho people actually want. If it suits their purpose and gets them the results they want.

The question is, does all this horse race rhetoric affect the outcome? Ask yourself that. Maybe you like Martin O'Malley. He could win the Democratic nomination. But no one mentions his name, then they take a poll of people who have never heard his name, and then they tell you he's too far back at the quarter mile post to ever catch up with the frontrunners so don't bother voting for him. And the person who might be the best candidate for the job, and who most people might prefer if they checked him out and stuck to their guns, gets kicked to the curb.

If Americans had an ounce of principle and intestinal fortitude, they could hire the best person for the job. And polls are no small part of they reason why they never, ever do.

There are two kinds of polls: There are the ones they use to push you into supporting the candidate they picked for you, and there are the ones they use to see how good a job they did of pushing you into supporting the candidate they picked for you. And the latter are actually reasonably accurate.

[cindy25]

it only predicts the first few. when candidates start pulling out everything is realigned. starting March 1 they become winner take all. IA and NH have few delegates.