Commentary
Poll Misfires: Maybe TV Show Questions Would Elicit Truer Results
- by Wayne Friedman , Staff Writer, November 23, 2020
But don’t stop there. How about revealing what TV shows and theatrical movies your friends and family consume?
In addition to its usual “probability” polls -- where respondents are asked who they want as President -- the USC/Dornsife Daybreak presidential poll also asked a social-circle question: Who will your family/friends be voting for?
The probability question is what many pollsters typically do -- concerning one’s overall choice for president. The problem this year -- as in 2016 -- many polls were fairly off from actual results.
Just before the election, many polls had Joe Biden up by 10 percentage points over Donald Trump, versus current actual popular results, that currently have Biden 3.8% higher over Trump via the popular vote.
advertisement
advertisement
When factoring in “social-circle” results -- in a post-poll effort -- USC says data was more in-line with actual results. Turned out, there was seemingly more support for Trump than anticipated. (Biden is still the winner.)
Why the question? Jill Darling, survey director of the USC Daybreak poll, speaking to NPR’s “Marketplace” said:
“People may not want to admit they’re voting for a very polarizing candidate, particularly if they’re in a situation where their family and friends aren’t voting for them.”
This reminds me of some entertainment surveys -- “soft research” offering data on TV shows, “popular” TV networks, theatrical movies, results that may be skewed.
Will some consumers admit to watching hours worth of guilty-pleasure cable TV reality shows? How many adult moviegoers own up to seeing hard-core horror tributes? Conversely, can we find TV consumers who admit to watching serious historical, but perhaps dry, nonfiction shows on genre-specifc cable TV networks.
Putting aside panel-based and/or census-based measurers, one can imagine entertainment consumers may want to tell pollsters -- and, possibly, their public -- what looks good.
Perhaps surveys should ask: What kinds of media are your family/friends consuming media tonight? “Grey’s Anatomy,” “Hannity”, “The Crown,” “Real Housewives” or “Anderson Cooper 360”? Step right up and cast your vote.
Wayne Friedman is West Coast Editor of MediaPost. You can reach Wayne at wayne@mediapost.com.
The idea that people lie in polls because they don't want to admit they are voting for a polarizing candidate is gibberish and the type of nonsense pollsters like to use as an excuse for their inaccuracies. The way most polls are designed and structured is fundamentally flawed as are ridiculously small sample sizes that are no longer representative of a much more splintered universe.
Steve, you say that small samle sizes are a major problem as the universe is so slintered---which makes sense for TV channels, radio stations, magazines, individual websites, etc. in the media world. But in the case of national as well as most local or statewide elections, we are talking about three groups---those favoring one of the two parties and independents. Typically the split is 35-40% Democrat Party, 35% GOP and about 20% independent. This is hardly a splintered situation as these are big slices of the pie---like TV during its Golden Age in the 1950s when the two top TV networks split 75-80% of the primetime audience---that's almost the same degree of dominance as the two political parties do now in our elections. Then, Nielsen used only a thousand metered homes to rate national TV and there was no problem for most of the shows. Indeed compliations of local market rating surveys, with combined samples 100-150 times thaf of Nielsen's national panel ,produced identical ratings when their findings were aggregated.
What some of us are saying is that aside from how the questions are posed---certainly a key point---- that subsets of mindsets within the three basic divides---pro-Democrat, pro-GOP and independent---may be at work in political surveys which generate far more heat than TV shows when people are asked about them. Hence, we suspect that lying as well as distrust of such surveys which causes certain segments within the larger groupings to opt out can't be dismissed as causative factors.
Make that "sample sizes" in the first sentence of my reply not "samle sizes".
Just some observations from afar.
Yes, some polls had Biden +10 early on. But at the same time some polls had Trump +3. That differential says more about some of the indiviual pollsters rather than the populus.
But all polls are a 'point-in-time'.
Here in Australia we get a lot of coverage of US news, and have access to many US news programmes. The sense I (and others) was getting was that Trump really ramped up his rallies, while Biden remained being COVID-19 aware with his public appearances. To put it into sailing parlance ... Trump was coming home with a wet sail.
So those early polls MAY have been within the margin of error.
But rather than forensically dissect the polls, maybe a question or two should be posed about the electoral system. As one colleague said to me ... win by 3 million you lose, but win by six million you win.
And make that "splintered" as opposed to "slintered" also in my first sentence. Sigh!
I wouldn't ever say who I'm voting for to a pollster ever and I only believe one poll that is when the results come in that night. Since the polls are wrong and didn't learn anything from 2016 since they were way off again in 2020.
You may be too young to remember, but nobody voted for Nixon.