Political polling is suspect, more now than ever. Accuracy – as in 2016 – is a real question. Biden appears to lead in national polls – but deeper analysis tells a different story. Without wishful thinking, unpack facts. And remember, elections not polls are what matter.
First, survey response rates are critical – and unknown. Statisticians say response rates affect results. If a phone surveyor contacts 10 people before one answers, “non-response bias” exists. That bias is important. It comes from how people feel, plus survey administration, question length, and circumstantial factors, such as the respondent’s time, interest, fear, and distrust.
Here is the kicker. Most political surveys, conducted on a short fuse by phone, do not reveal their response rates. They do not admit “non-response bias.” What could that mean? A lot.
Consider a good response rate is one-third of those contacted. Non-political experts say, “a low response rate can give rise to sampling bias if the nonresponse is unequal among the participants.” In other words, a survey that irks those called – who are busy, stressed, fearful, distrustful, increasingly private, or maybe just conservative – misses an entire demographic.
Experts say “non-response bias may make the measured value” useless, even misleading. What happens “behind the curtain” matters. For most political surveys, we have absolutely no idea.
Before leaving “response rates,” note Pew Research reports political polling saw a 36 percent response rate in 1997 – but today gets six percent. That means more than 15 people say “no” to a caller, before one says “okay.” See, https://www.pewresearch.org/methods/2019/11/19/a-field-guide-to-polling-election-2020-edition/.
Another issue is “survey design.” Short surveys increase response rate, but mask big issues. Knowing who won a debate does not reveal who someone will vote for, or if they will vote. Leading questions often produce desired answers. That makes survey data inherently dubious. Omitting options – like USA Today omitting Pence’s name while surveying post-Pence-Harris debate – tips responses. Net-net, survey design can also undermine results, and mislead.
Now note this: Phone response rates have been falling for decades. As people shift to cell phones, call blocking, avoiding spam, protecting privacy, and get tired of surveys, they get less accurate. Online options and social media reduce interest. Then, add exhaustion with politics.
What else affects accuracy? Curiously, in the United States more than elsewhere, COVID-19 stress is affecting mental health. Perhaps a third of the country is highly stressed. This reduces public engagement, even when emotions run high. Surveys are likely their last concern.
As Pew notes, online polling – while convenient – makes demographic identification hard, so “random sampling” is impossible. Today, 10 percent of Americans are “offline.” Pew coyly notes they are demographically unique. Yes, and another thing: They vote.
The last reason to question polls is counter- evidence. Perhaps Biden is leading, but that is hard to square with Trump rally data. On October 28, Trump’s rally in Bullhead, Arizona – with 23,000 in attendance – showed 24 percent not Republican, and 45 percent new voters since 2016.
Likewise, the Goodyear, Arizona rally produced 17,000 attendees, of whom 20 percent were not Republican and 35 percent new voters since 2016. Reflecting other rallies, the Omaha, Nebraska rally saw 43,000 attending, with 40 percent not Republican and 24 percent new voters.
What does all this mean? It could mean a repeat of Reagan’s 1984 realignment of Blue-Collar Democrats behind a Republican president. Reagan stunned Democrats – winning a second term with his 1984 walk-off homerun. Could it happen again? Hard to say, but polls do not take these hard numbers into account – in critical electoral states.
Of course, the real reason polls fail to predict elections goes beyond dismal response rates, absence of random sampling, and demographic inaccuracy. It goes beyond survey bias.
The real reason polls fail – is that they are not elections. Polls – even the best – are just snapshots. They are not real. They are here and gone, of no consequence – soap bubbles not a signature. They are a whim, venting of emotion, if people bother …which many do not.
Elections are very different. On voting day, early or late, absentee or in-person, we are exercising a sacred right – a solemn, constitutional duty. We know this affects our economic wellbeing, chances of work, return on investments, fate of family, right to speak and worship freely, travel unrestricted, protect ourselves, private insurance, stability on streets, security at borders, fidelity with Founders. Elections are real.
In 2020, our vote affects who will sit on federal courts, determining whether our Constitution remains “law of the land,” and whether this Republic – in Lincoln’s words – “will long endure.”
That is why voting – is different from polls. That is why the modern obsession with knowing the unknowable before an election – more vanity than validation, nonsense than horse-sense, is an unhealthy distraction, deception, defiance, not accurate refraction, reflection, or science.
Trump knows this. Whether Biden does is secondary. An electoral win is the key – and depends on our voting. Polls aside, the big day is here. The nation’s future depends on Americans of good heart, who love this country, voting. Tomorrow, polls will be dust in the wind. This election – this one precious moment – is what really counts. Use it well.