With the primaries drawing closer each week, we’ve recently been seeing a lot of political opinion polls being aired. These are the bar graphs that flash up on news stations’ screens, showing what percentage of Americans are backing each political candidate… Supposedly.
When visiting underdeveloped countries, researchers are often frustrated by the difficulty of acquiring accurate statistical data. This is often because it’s simply too hard to distribute enough surveys to definitively measure public opinion. In developed countries like the United States, we often run into the opposite problem: So much data that’s it’s almost impossible to figure out which is the most accurate true. Now technically we could figure out a precise measurement, as we do with the census every ten years. The problem with political polls is that they simply don’t. And so we’re left with something called selection bias: a statistical measurement that’s not representative of the population being analyzed.
Right now polling methods are considerably outdated, because they rely on a rather antiquated form of communication: landline telephone calls. Twenty years ago, only about 6% of households used cell phones only. Today that number has increased to 60%. And remember, cell phones can’t be used for polling because most providers will protect their users from “spam” phone calls, including surveyors and telemarketers (“With Primaries Around the Corner, and Look at How Much We Trust Polls”).
So that leaves only about 40% of the population available to be polled. And let’s consider what types of people who use landline phones typically are like. First of all, they’re generally on the older side (do you know anyone our age who owns a landline, even if they own their own home or apartment?) (Lepore). This bias can often be problematic, especially for Republican candidates, whose constituents tend to be older Americans. If you remember back in 2012, many polls predicted a win for Mitt Romney before Obama somewhat unexpectedly won–and by a decent margin. The polls had failed to take into account the millions of young people who planned to vote–the majority of which were democratic and did not own a landline phone.
Second, since the 1920’s, the response rate (i.e. percent of people who pick up the phone and say “yes, I’m willing to participate in this survey”) has dwindled from 90% to less than 10%. And nowadays the people most likely to agree to answer surveys attend church, volunteer, and/or are particularly politically active, adding another layer of bias into the mix.
In short, rather than extend an equal opportunity to all Americans who plan to vote in the elections, the use of landline phones has created a selection bias that favors older Americans, churchgoers, volunteers and political activists. The creates a sample size that’s only about 4% of the population, rather than 100%.
While these polls can generally place a candidate in the race, they certainly can afford to be much more accurate. For example, polls could start appearing online instead of on the phone, allowing more of the population to reach and respond to them. However, the demographic most likely to vote (older people) is the least likely to spend enough time on the internet for that strategy to be effective. I’m wondering if any readers out there have any other ideas for alternative poll measures. Can you think of something I haven’t yet?
Until then, try to take the graphs flashing on your TV screen with a grain of salt.
Lepore, Jill. “Politics and the New Machine.” The New Yorker. n.p. 16 Nov. 2015. Web. 1 Feb. 2016.
“With Primaries Around the Corner, a Look at How Much We Trust Polls.” Here and Now. Trustees of Boston University, 26 Jan. 2016. Web. 1 Feb. 2016.
Zukin, Cliff. “What’s the Matter with Polling?” The New York Times. The New York Times Company, 20 June 2015. Web. 2 Feb. 2016.