Polls Are Only Accurate In The Immediate Months Before An Election.
It seems people love hearing about poll results, because it feels like we’ve been getting inundated with Gallup this and Rasmussen that in regards to the upcoming presidential election since … well … the last presidential election. And those results sure do seem to change a lot, don’t they? Historically, from week to week and month to month, you might see a number of candidates in the lead who would go on to drop out of the race entirely due to lack of interest. One notable exception is Martin O’Malley, who never quite left the “Who?” list before dropping out.
The truth is that when you see a poll that doesn’t take place immediately before the election itself, the results mean practically fuck-all where accuracy is concerned. And the reason should be obvious: Circumstances change. But that doesn’t stop the incessant polling from taking place, nor the talking heads on the news from pretending what the information they’re giving you isn’t a steaming load of jibba-jabba.
People Who Take Polls Don’t Represent Normal People
There was a time, believe it or not, when Americans were A-OK with strangers knocking on their doors on days other than Halloween. Back in the old days, there were door-to-door salespeople, religious folks who weren’t Jehovah’s Witnesses or Mormons, and weirdos who left milk on your porch before running away. Pollsters, in other words, weren’t immediately treated with shifty eyes every time they knocked on a door.

But now we’re not that trusting, way more apathetic, or maybe we simply can’t be bothered because we have a quadrillion better things to do. Whatever the reason, the response rate for polls is way down from what it used to be. We’re talking a “Vanilla Ice from 1990 to 1991” kind of decline in popularity. It’s become so bad that the only people participating in polls, or at least a disproportionate number of them, are the same sort of loons who call in regularly to AM talk shows.
Deriving a truly random sample of the larger population has become a daunting task for pollsters, and without a decently large number of respondents, the risk of a failed prediction becomes more like a foregone conclusion. Especially when so many of the people who are willing to take part can best be described as “oddballs,” or those who are looking to screw around for laughs.