January 20, 2016
This podcast from the NPR politics unit offers a useful perspective on political polls:
To Nate Cohn, reporter with The New York Times‘ The Upshot, not all polls are created equal. In fact, very few are worth your time. He joined NPR’s Tamara Keith and Danielle Kurtzleben on the NPR Politics Podcast to talk about political polls, the methodology behind them, and which ones can actually tell us something about the race for the White House.
The program discusses various factors that can alter the results of political polls. Among other things, this includes:
How participants are selected — at the most basic, calling people with cell phones only vs. people who also have landlines tends to skew results towards different demographics.
How the poll is conducted — answers tend differ based on whether polls are conducted by phone with another person on the other end, versus more anonymous phone polls where participants select a candidate by pushing a button or online polls.
Additionally, how people say they’re going to vote doesn’t actually directly correlate with how they do vote, and people are more likely to say that they intend to vote than actually do vote.
Timing is also important when considering poll results. As we enter the primaries for the upcoming presidential candidate nominations state-based polls are more important than national polls. At this stage in the game, national polls tend to inaccurately predict results.
Given all this, are any polls reliable?
The podcast experts suggest looking at aggregates of polls (can be found, for example, on The Huffington Post) as well as at pollsters who have a good track record and use deeper, bigger data sets than mainstream media polls typically do.
Methods Madness, Policy
© 2017 deeptech