In Depth Look at Polling
  View Video
A few days until election day you are probably seeing a lot of polls predicting outcomes. How reliable are they? Are they biased? I crunched the numbers and talk to a local professor to get the answers. A lot of people have lost confidence in the polling. But not everyone has lost faith. "I think they do a good job of predicting the election." Local officials said that those political polls have an extra challenge. "The trick is trying to figure out who was going to show up. That's the challenge." They said national polls were pretty accurate by historical standards. Finding that 2016 national polls were slightly more accurate on average "my feeling was that the state polls were not in the field late enough to pick up what was going on." Saying that as many people have ditched land mines it is difficult to get people to answer polling questions. "It's the young ones that are impossible. They cannot answer the phone. That has led to more internet polls that happened inherent bias. The website keeps track of election polls." You can see how far off they were or if they were consistently off one way or another. In 2016 polls predicted that Hillary Clinton would take the popular vote by 2 points. Trump ended up winning by .7%. In other words they were 2.6% pro-democrat. But in that 2012 and 2008. The polling average predicted that Barack Obama would win. But they predicted that the race would be closer than it was. Three points pro GOP. This year there is no presidential election. So we are crunched historical numbers on state governor and senate races, you might be surprised by the results. State governor races the polling average has predicted the outcome. Missing by one an 1.2%. And the state senate the poll incorrectly had Katie McGintey beating Pat Toomey in a 2016. And the three previous senate races ended up being pro gop. Between 2.5 an 5.8 %, when Bob Casey beat Rick Santorum in 2006. So how accurate are polls? We crunched every national poll for the last 20 years. The average error was a 5.9%. The average was a bit closer. As for a slant or bias. They found that they tended to miss one way or another by 2 to 3 points and in total since 1998 polls tended to favor democrats by 0.3%.