There’s been a lot of discussion about polling since the election, mostly ridiculing polls, wondering why we pay attention to them at all and questioning their role in future presidential campaigns.
Polls pretty much hit Joe Biden’s percentage of the popular vote on the nose and very few states went in an unexpected direction. But Donald Trump’s percentage was significantly higher than expected and several states were either much closer or much more lopsided than polling predicted. Additionally, polls seemingly got it wrong in a number of Senate and House of Representatives races, usually with Democrats underperforming expectations.
So what happened? Certainly a number of “unlikely voters” decided to vote after all and the vast majority of those who claimed to be undecided turned red. But polling itself, in the 21st century, is also to blame.
Reputable polling groups attempt to make their polls scientific. The key is randomness. Everyone in all population groups need to have an equal chance of being polled and it is critical to reach all demographics in proper percentages, accounting for variables and weighting results if necessary.
But what if respondents — those actually willing to participate — skew the data so badly there’s no way to scientifically account for it? This is critical because polling data is so frequently cited in campaign coverage and can have an effect on who votes.
David Hill, the director of a consulting group and a fellow at the University of Southern California’s Center for the Political Future, recently wrote an Op-Ed piece for the Washington Post detailing how much things have changed during his long career as a pollster.
He wrote that 40 years ago he could expect one in five of the people called to answer and respond. Today? Thanks to caller ID, ubiquitous cell phone use and people’s almost universal aversion to accepting calls from numbers they don’t recognize, Hill can only get about one in 100 called to participate.
Who does that leave actually participating in these polls? It’s sort of like the adage about trials: Who wants their fate decided by the only 12 people who can’t figure out a way to avoid jury duty? In this case, do we want the direction of our elections being influenced by people with nothing better to do than answer phone calls from strangers? And only those politically motivated enough to then remain on the line for the time it takes to complete the questioning?
Hill wrote that 1,510 interviews were needed for a representative poll of Florida. To accomplish this, those conducting the poll had to make 136,688 calls. And it got more and more difficult as Election Day neared. From Oct. 25-27, he wrote, only four-tenths of 1 percent of those they attempted to reach responded. Based on election results from the Sunshine State compared to pre-election polls, “likely Trump supporters” were unlikely to pick up the phone and participate in the poll, and that fact wasn’t properly accounted for.
This is all bad news for the future of polling. If certain demographic groups won’t respond, the poll won’t have a random sampling that’s anywhere near representative of the electorate.
By the next election, we’re likely to see far more polling done via text, as people are more willing to reply to texts than answer a phone call. Heck, maybe going forward Facebook or Twitter will be able to facilitate polling as accurately as Gallup or Quinnipiac or a poll designed to reach younger people will be conducted via TikTok.
There’s some evidence that surveys on topics that have become politicized are skewed as well. So it’s difficult to know what percentage of people really plan to receive the coronavirus vaccine or believe schools should be open as answers tend to change as the direction of political winds change.
The good news?
It seems that non-political surveying remains accurate. So, at least when Steve Harvey tells you on “Family Feud” that, out of 100 people surveyed, 68 of them think the worst kind of shoes to wear in a marathon are high heels, you can trust him.
Bob Blubaugh is the editor of the Carroll County Times. His column appears Sundays. Email him at firstname.lastname@example.org.