Gay Stats Whiz Nate Silver Says Political Polls Are B.S.
Considering that the Presidential election is an entire year away, all the media coverage and polls this early might seem kind of silly. But recently, statistician Nate Silver of FiveThirtyEight.com sat down with On The Media‘s Brooke Gladstone for an interview… where he said, yeah, polls this early are kind of silly.
He didn’t say it in those specific words, but Silver did say that he’s “never seen this much attention paid to polls so early in the campaign.” Given that it’s so early, the polls will swing wildly as we learn more about candidates and they get winnowed out, until sometime in 2016 when things settle down a bit.
Silver also points to the media as feeding polls, which the media then reports on. He calls this process “Discovery, Scrutiny, Decline”: First, a candidate surges in popularity (for example, Ben Carson, who recently overtook Trump as the frontrunner in the Republican race). This results in increased media coverage — Ben Carson’s surge in popularity became a news story, and the media start focusing on his policies and beliefs. That added scrutiny results in a decline in Carson’s poll numbers — both as people find out what he stands for, and also just because he’s no longer an appealing cypher.
“Discovery, Scrutiny, Decline” happens especially when, in Silver’s words, “a candidate’s near term polling is out of line with what we call the fundamentals… Where you’re kind of scratching your head and saying ‘I can’t figure out why so many people think that candidate is so great.’” People might say they like Ben Carson, and that’s probably who they’d vote for today but they still haven’t really made their mind up. As people get closer to making up their mind, they may abandon a candidate for another. Silver says, that if polls were honest, “It would say 70 percent undecided, 6 percent Trump, 5 percent Rubio, 4 percent Carson — that’s probably a more accurate representation of how many people have really made up their mind so far.” But since people don’t want to look foolish, they’ll give pollsters an answer of whichever candidate they’re most interested at the time. In other words, even though Carson’s popular, he’s not pulling the support he needs for the long haul.
There’s also the issue of how data is collected. Silver says most polling is still done via telephone, which means landlines — which are mostly owned by older Americans, and the most common demographic to answer a phone poll is older, white women. To help account for the demographic spread, pollsters might weight answers from a young Latino man more heavily, and answers from an older white woman might only count for half. While this means the polls are a little bit more representative of America’s demographic makeup than they would be, this type of massaging of data means the data isn’t nearly as accurate or reliable as it should be.
To avoid this sort of bias in the data, Silver says that he has a partnership with Google, since they keep track of both search queries and how much news coverage a candidate is getting. And, for FiveThirtyEight.com, he compiles polls from around the country, weighting them based on their accuracy, and uses that to come up with final numbers. Silver says his method has been successful so far — he called nearly every state right in the most recent election — but he’s worried about the reliability of polls. He says “here have been massive polling errors in the UK and in Israel and Greece, in Scotland and many other Westernized democracies. They’re using the same basic techniques, so the fact that we’ve had these errors elsewhere is a sign that we could have trouble in the US sooner or later too.”
All of these issues add up to one thing: Political polls aren’t worth the pixels they’re printed on, especially at this point in the race.
(Featured Image Via DonkeyHotey)