Presidential polls: What to know about margin of error, methods and more

Dr. Andy Smith from the University of New Hampshire Survey Center joins LiveNOW from FOX on Election Day with tips for deciphering the math behind the polls and how to better understand them.
As Americans await election results, many are looking to the polls for answers. 
Dr. Andy Smith from the University of New Hampshire Survey Center joined LiveNOW from FOX on Election Day with tips for deciphering the polls and how to better understand them. 
Margin of error describes how close a survey result can reasonably be, relative to the true representative of the entire population. In other words, even the purest random sampling of surveyors won’t be an exact match of the entire American population. 
Calculating the margin of error works through a complicated mathematical theory known as the central limit theorem, Smith tells us, but gave us some tips for understanding it more easily. 
He said the biggest confusion about sampling error is people think it applies to the percentage gap between the candidates, which isn’t the case. 
“What you have to do is apply a margin of sampling error to both estimates,” he said. “So Donald Trump at 46%, in this example (with a 3% error of margin), he might be up to 49%; he might be as low as 43%. Harris at 50%, could be as low as 47% or as high as 53%.”
“You really have to about double the margin of sampling error if you want to understand if the gap between those two candidates is statistically significant,” he added.
RELATED: Presidential election polls: Who is currently ahead?
FILE – A close up photograph of a man holding a cell phone.
“The polling industry is going through what I would call a paradigm shift, where we’re moving from the types of methods that worked 30, 40 years ago and don’t really work anymore,” Smith said. 
He said response rates for phone interviews have significantly declined in recent years, to the point where he said only about 5% of people who get contacted for telephone surveys actually complete them. 
“This results in very, very expensive telephone surveys and, (as we saw) in 2016, inaccurate telephone surveys,” he said. 
He said more surveys are being conducted digitally this election, but that several different methodologies for digital methods also exist.
“We’re going to see after this election quite a bit of research to understand which one of the methodologies that were used actually performed best, and use that knowledge that we got going forward,” he said. 
“I’m very cautious in saying one poll is better than another poll right now because frankly, we just don’t know. We are in the middle of this development of our best practices and we’re not at the end of that yet.”
RELATED: When will we know the results in the swing states?
Every election cycle, a handful of polls publish shocking results that drum up significant attention. 
Most recently was the highly anticipated poll from J. Ann Selzer, the “gold standard” pollster in Iowa, showing Kamala Harris beating Donald Trump by 3 points in the Hawkeye State.
READ MORE: New Iowa poll shows Harris ahead by 3 points in the Hawkeye State
So, how do some poll results vary so differently? 
Smith says it could have to do with how the survey was conducted, or that an unusual sampling group was used that may have excluded a certain demographic.