The True Losers of the Michigan Democratic Primary: The Pollsters
The Clinton campaign team wasn’t the only organization embarrassed on Tuesday.
As you can see in the image below, every single poll reported figures far different than the ultimate vote count.
Let’s try to figure out how it all went wrong.
The Best of the Worst
CBS News/YouGov was the least wrong of the March 2016 polls, reporting a Clinton lead of 11%.
The CBS News/YouGov polled 597 self-described likely voters and estimated a 7.7 margin of error. Respondents “were selected from YouGov’s and two other online panels. These are ‘opt-in’ panels are open for anyone to join.”
The Worst of the Worst
The methodology of the least accurate poll was conducted by Mitchel Research & Communication on March 6th and reported a Clinton lead of 37%.
The poll used an “IVR (Interactive Voice Response) automated phone survey of 475 likely March 2016 Michigan Democratic Presidential Primary voters.” Mitchell estimated its survey had a 4.5% margin of error.
The first words of the automated message were, “Hi, we are conducting a survey of Michigan voters for TV and radio stations on the Presidential Race and other issues.”
(You can find links to all the Michigan democratic polls here.)
Possible Conclusions
Online Surveys are Superior to Automated Phone Calls
The use of opt-in online surveys seem to be more likely to accurately predict results since the participants actively want to tell a pollster who they’re voting for, instead of receiving an unsolicited phone call from a robot.
Mentioning that a survey is being conducted “for TV and radio stations” may put pressure on respondents to say whatever they think they’re supposed to say, instead of the actual truth.
A higher number of respondents is linked to more accurate results, and reaching more respondents is easier to accomplish online.
The “Margin of Error” May Have More to Do with Humility than Mathematics
The YouGov survey claimed a 7.7% margin of error and was ultimately off by 12.5%. However, the Fox News/Mitchell survey reported that it had a 4.5% margin of error, yet ended up being 38.5% inaccurate.
A Little Known Problem with Polls
You may be surprised to learn that polls do not report the exact numbers gathered. Pollsters “weight” their findings based on previous voter turnout based on age, gender, race, and political party. Essentially, this means that you might say you’re going to vote, but if you belong to a demographic which hasn’t voted much in the past, they won’t include your answer as part of the final poll results. Its unfair, and somewhat insulting, but that’s a discussion for another time.
Coming Up Next…
The next piece I’ll be posting will discuss the revolutionary online political futures market which forecasts election results better than any poll.
Stay tuned.
Alex Marshi is an American writer, comedian, and marketer currently residing in California. Follow him on Twitter (@Alexandermarshi) and visit his satirical news site, Too True News.
