People ditching landlines causing problems for pollsters?
Nearly seventy years ago Literary Digest conducted a poll of millions of voters that predicted that Republican Alf Landon would trounce Franklin Roosevelt in the 1936 presidential election by a whopping 57 percent to 43 percent. Roosevelt went on to win the election in a landslide, grabbing 61 percent of the vote, outperforming what the Literary Digest poll had predicted by a full 18 points. So what went wrong? Literary Digest drew their unrepresentative sample from automobile registration lists and telephone directories, which was all fine and good except for the fact that during the Depression days of the 1930s people who owned cars and had phones tended to be wealthier than the average American—and Republican (George Gallup's more scientific polling style more accurately predicted the winner, even though it drew on a smaller sample size).
Could something similar be happening today? Most presidential tracking polls these days only call landlines for their surveys, which wouldn't be such a big deal were it not for the fact that more and more Americans are ditching their landlines entirely and going completely wireless. They're not being surveyed because pollsters generally don't call cellphone numbers (there's no directory of numbers to draw on, for starters), but while the landline-less are still a relatively small proportion of the population (about 3 percent), the numbers are growing. The result is that they're basically ignoring three percent of the population that is generally made up of tech savvy types who don't see the need to own a landline; with polls as tight as they are even that small number could make a difference.
