They laughed at PostScript when she built the bunker. It might have been because her fly was down, but even so, they’re not laughing now, Day One of the promised 2012 apocalypse. Actually they might be laughing, but PostScript is in the bunker and so can’t hear it.
Anyhow, in other news about our entire doomed way of life, Robert Samuelson argues today that our political polls are unrepresentative in ways we can’t even determine. For example, young people only have cell phones and smart alecks lie to pollsters. Plus, the last two national election cycles had wildly different voting populations, such that a “representative sample” is an opinion, not a fact.
Nate Silver, mad devourer of numbers over at his New York Times FiveThirtyEight blog, gives President Obama nearly a 75 percent chance of taking this thing home with him; UnSkewed Polls man, who was a big touted deal a month ago, now predicts 359 electoral college votes for Mitt Romney, based (this is true) mostly on questionable polling data that Nate Silver is a girly man.
So while we suspect that we can’t trust the polls we have, we also won’t know in what WAYS they are wrong until after the election. Yet obsess we must!
SchaulRonny delights PostScript with his or her use of “data” as plural and assertion that the overwhelming winner of all polls is [click] dial toooooonnnnnnnnnnne:
Response rates of 9% render poll estimates meaningless because the sample is almost certainly unrepresentative (sampling data are guilty until proven correct). Response rates of 91% might give meaningful estimates. The 1948 “Dewey Beats Truman” error was due to the fact that it was a telephone poll, in 1948 most people who had phones were wealthy, and wealthy people tended (and still tend) to be Republican.
Jumper says having preset answers in polls means nuanced opinions are needlessly simplified:
If someone believed the President has practically nothing to do with the economy — a debatable proposition, but one I assume many people have — here will be a question: “Who will be better for the economy, Obama or Romney?” and there will not be an “it doesn’t matter” option, only “undecided.”
lohengrin prefers to let the free market predict:
It will just prove once again that Intrade is better than all these guys. It was right in every state in 2004 and right in 48 out of 50 states in 2008. Even in 2008 it was only one electoral vote off.
MNUSA admits to being one of the poll-hostile who never gets counted:
You make a good point, Mr. Samuelson. I’ve often wondered how polling groups get people to answer their phones. This time of the year, our phone rings constantly with everything from charities seeking money, politicians seeking money and time, pollsters, salesmen trying to sell windows and doors and other seasonal products. I don’t answer any of them.
And three commenters have fun strategies for deliberate poll sabotage, which is its own way of making one’s opinions known:
I have gotten several polls, before I stopped answering the landline. If they appear slanted I hang up.
You should answer them. If it’s a political survey, tell them you’re a Libertarian. Or that you’re going to write in Ross Perot. Or that you’re a Moonie. Tell them you already voted. Lie about your race. Be creative. Pacific Rim Islander is good. Tell them you don’t recognize any of the names they mention, Obama, the other guy, whoever. It’s fun.
Another good line when a person actually reaches you. “How much are you going to pay me for my time?”
That’s all very well, but PostScript would feel much better about life in general if Intrade and Nate Silver would apply themselves to predicting power outages, flooding and downed tree branches right about now.