A good poll reflects opinions, as long as you take the margin of error into account.
But, a good poll is extremely difficult to achieve. You can no longer just dial phone numbers at random. If someone has a landline at all, they don't answer it unless they recognize a caller ID. And, you can't call cellphones without prior permission.
And there's also the "shy voter". Trump supporters are abused on a regular basis, both physically and verbally. Many will not volunteer that information to a stranger.
Rasmussen does better than most. They use an interactive voice response system, rather than a live human. And, they also attempt to survey people that have abandoned land lines.
They called the 2016 popular vote almost exactly, predicting Clinton +1.7, while actual results were Clinton +2.0. Of course, that doesn't account for the Electoral College.
That's one of the complicating factors for polling: in order to forecast the Presidential election, you have to conduct 51 different polls. Actually, it's 54 polls, as Maine and Nebraska can split their electoral votes.
Nate Silver tried to do this, but he didn't have control of the per-state polls and how they were conducted. And, the results were so close that he couldn't predict Trump would sweep the Rust Belt states.
A good poll reflects opinions, as long as you take the margin of error into account.
But, a good poll is extremely difficult to achieve. You can no longer just dial phone numbers at random. If someone has a landline at all, they don't answer it unless they recognize a caller ID. And, you can't call cellphones without prior permission.
And there's also the "shy voter". Trump supporters are abused on a regular basis, both physically and verbally. Many will not volunteer that information to a stranger.
Rasmussen does better than most. They use an interactive voice response system, rather than a live human. And, they also attempt to survey people that have abandoned land lines.
http://www.rasmussenreports.com/public_content/about_us/methodology
They called the 2016 popular vote almost exactly, predicting Clinton +1.7, while actual results were Clinton +2.0. Of course, that doesn't account for the Electoral College.
That's one of the complicating factors for polling: in order to forecast the Presidential election, you have to conduct 51 different polls. Actually, it's 54 polls, as Maine and Nebraska can split their electoral votes.
Nate Silver tried to do this, but he didn't have control of the per-state polls and how they were conducted. And, the results were so close that he couldn't predict Trump would sweep the Rust Belt states.