Those polls you’re reading? They’re bullshit.

Photo: Daniel Lobo via Flickr

I’m sick of people treating presidential poll numbers as if they are facts. They’ve got no connection to the upcoming election. Drawing precise conclusions from a sample of 450 likely voters a year before the election is bullshit.

Look, I’m no newcomer to consumer surveys. I spent 18 years launching and working with Forrester’s broad and deep survey product, Technographics. My experience tells me you should ignore exactly 88.6% of what you read about presidential preference polls. (If that seems overly precise, you’re beginning to get the point.)

The venerable pollster Gallup has exited the horse-race polling business because they’re not sure what they’re measuring. If they’re giving up on polling, maybe you should, too. Here’s why:

  • All polls have sampling error. Put aside the obvious potential biases — too many men, or not enough black people. All polls have one huge bias you can’t eliminate: response bias. They undersample busy people, rich people, people who hate pollsters, shy people, young people out having a good time, and people who don’t answer when strange numbers show up on caller ID. They oversample lazy people, people hanging out at home, and opinionated people. And online polls oversample enthusiastic supportersYou can’t poll people who won’t take a poll. What bias does that introduce? It’s impossible to know — at least until the election.
  • Early polls inflate votes for wackos. People recognize Ben Carson, Carly Fiorina, and especially Donald Trump because they say incendiary things that spread fast on traditional and social media. Conversely, unless you’re diligent, you have no idea what sparsely-reported candidates like Martin O’Malley, John Kasich, or Lindsey Graham are saying. Telling a pollster “I like this guy” — especially if they’re for shaking up the status quo — doesn’t mean you’ll actually vote for them months later.

[tweetthis twitter_handles=”@jbernoff”]Early presidential polls have a bias toward wackos.[/tweetthis]

  • Polls are disconnected from election dynamics. Volunteer efforts, endorsements, and personal appearances make a huge difference in Iowa and New Hampshire — which is why Nate Silver’s gang at fivethirtyeight.com base their predictions on a lot more than poll numbers. When people in those states actually vote, they’ll have to make a serious decision about caucusing for Bernie Sanders or pulling the lever for Donald Trump. After Iowa and New Hampshire, many candidates will drop out and we’ll know a lot more about the ones that remain. That will reflect a completely different landscape of candidates, making today’s polls meaningless.
  • The media overestimate tiny variations. What does it mean if Trump goes up two points in a poll of 405 likely voters? It means that 8 more people said they would vote for him. A poll like that has a margin of error of +/- 5 percentage points. (And remember, out of 100 theoretically accurate polls, five will be off by more than the margin of error.) Even so, pollsters and media treat a variation of a couple of points as significant. Even a movement of more than 5 points means that among the dubiously representative sample of mostly ignorant people that the pollster reached, a few more of them said somebody’s name. That’s not data, that’s noise.
  • And don’t get me started on the expectations game. What difference does it make if somebody beats expectations? It says the expectations were wrong — in other words, the poll, and the pundits’ analysis of it, was bullshit. That’s a fact about the media, not a truth about the candidate. Only in the 24-hour news cycle world can the media spin “we were wrong” into “the candidate did well (or poorly).” Don’t be fooled.

Polls are a blunt instrument. They tell you the broadest possible truths only. Here’s the complete truth that recent presidential polls have “revealed.”

  • Hillary Clinton and Bernie Sanders (and Joe Biden, if he enters) are the only Democrats that might win.
  • Donald Trump, Ben Carson, and Carly Fiorina have name recognition.
  • Marco Rubio, Jeb Bush, and Ted Cruz have a better chance than the rest of the Republicans, but it’s too soon to count anybody out in a field that big.

Disappointing, I know. You want numbers. That’s why meaningless polls exist — because you want numbers. Just recognize that at this point polls are not data, they’re bullshit.

[tweetthis twitter_handles=”@jbernoff”]Polls exist because the media want numbers, not because they predict anything meaningful.[/tweetthis]

morpheus data 2

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.


  1. Carlo Rubio or Marco Rubio? “Carlo Rubio, Jeb Bush, and Ted Cruz have a better chance than the rest of the Republicans, but it’s too soon to count anybody out in a field that big.”

  2. I have lived in 10 states over the past 50 years, including New York and California. I have never meet or even heard of one person who was included in any survey or poll. So where does this information come from? Do these polling companies pull the data out of their ass?

  3. I’m reading the comments from four years ago what arrogant to think that Hillary one of the most revered females in politics got her clock cleaned by a businessman from New York go fucking figure