Consider a hypothetical. Imagine that a bright-eyed pollster, working for Gallup or Harris or Quinnipiac, calls you on the phone. This assumes that you still pick up the phone when you do not know who is calling. Let’s say that this pollster, acting in a thoroughly professional manner, asks you your opinion of certain politicians-- and even asks you who you would vote for.
Would you respond?
If you supported Donald Trump, you would be seriously disinclined to respond. Why would you risk opprobrium?
While you are feeling all sorry for the polling industry and are preparing the plot where you are going to bury it, keep in mind that the current intellectual atmosphere has proclaimed that supporting Donald Trump, or any other Republican, for that matter, is akin to shouting Heil Hitler at Nuremberg. Would you put your life on the line to offer an opinion to a stranger?
I trust that you would not. In the first place, how do you know for whom the stranger is really working? And how do you know that your opinion will not be broadcast to the world-- causing you to lose your job, your friends and your family.
The political environment is so hostile to anything that smacks of conservatism that you can easily be forgiven for keeping your conservative opinions to yourself.
The polling failures are a symptom of a culture where there is only one acceptable opinion. Dissent, especially conservative dissent, is now considered to be a thought crime. Surely, you have noticed that the gods of Silicon Valley only censor conservative opinions, and have no problem with anti-Semitic eliminationist propaganda from the Ayatollah in Iran.
So, the polls underestimated Trump support, because-- why would you risk being denounced as a Nazi.
Consider the case of Sen. Susan Collins from Maine. You will recall that she offered the most sensible, most reasonable and most intelligent defense of the confirmation of Justice Brett Kavanaugh. For which she was targeted by those who hate intelligent, accomplished women.
In truth, all the polls had her losing her election to one Sarah Gideon, by more or less healthy margins. In the final analysis, Collins won by around 9%. How did the pollsters get it that wrong? One suspects that her name had become anathema and that admitting to a stranger that one supported Collins risked being denounced as an enemy of women, a fascist reactionary, a racist and whatever.
Extremist rhetoric has shut down public debate in America. It is not a good sign. It does not advance the cause of democracy.
Nonetheless, we feel for Frank Luntz. These remarks are from The Lid.
Veteran pollster Frank Luntz, contemplating his industry’s major misses in a second consecutive presidential election, declared his profession “done.” political polling dead.
The results of a highly contentious race between President Trump and Democrat nominee Joe Biden have Luntz feeling a little dismayed, especially when polls seemed to be indicating an easy win for the challenger for months.
“The political polling profession is done,” Luntz told Axios earlier today. “It is devastating for my industry.”
Luntz offers a different interpretation, namely that Trump supporters prefer to lie to the pollsters-- whether on purpose or because they do not feel that they have a choice, I do not know. Luntz knows better than I do, but still, why put in the effort to lie to a pollster when you can simply ignore the call?
Mainly, that supporters of the President love to mess with those conducting the surveys.
“Trump voters are hostile to pollsters because they think the pollsters want to make them look bad. They wear it as a badge of honor that they refuse to cooperate and refuse to be interviewed,” he asserted.
“As a result, measuring the breadth of support for Donald Trump is tricky,” Luntz claimed. “But to get it wrong twice in a row by huge margins is unforgivable.”
The Lid offers this analysis of the polling.
Trump clearly outperformed their estimates in nearly every crucial state.
And it wasn’t only the presidential race. The political pollsters suggested that the Democrats will increase their lead in the House of Representatives and take the Senate. Both predictions were wrong. If Joe Biden wins the election, a GOP Senate and a shrinking Democratic margin in the House will enable the Republicans to block the most radical parts of the Democratic agenda.
A saving grace.
Pre-retirement, I was frequently asked to consult on the statistical analysis of various polls and surveys. In addition to the above, there are two largely unremarked problems with polls and surveys that are, IMO, the Bismarckian sausage.
ReplyDelete(1) The most important thing is to identify the actual customer of the pollster/surveyor. In fact, I published an article in a professional journal on this very subject. My article was stimulated by a poll I personally observed. The poll, in early stages, clearly showed respondent rejection. As results began to trickle in, the early data were duly reported to the managing director. She demanded that the polling agency modify the poll to yield more positive results. The poll was modified, and her project, which ultimately failed miserably, received glowing poll results. This was but one case among many. The managing director was the paying customer, and the customer is always right. When you read "WSJ/NBC News poll", the WSJ and NBC News organizations are the customer. Not the public. And pollsters who want more contracts will see customers are satisfied; it's a business, not a calling.
(2) It is important to understand precisely who is doing the polling. Since the earliest days of psychometrics, it has been understood that questions must be asked in precisely the same way to every respondent. A mathematician colleague and I were retained by a large telecommunications company to assess the performance of a national polling agency hired to administer surveys of customers who had, in some way, contacted the company. The surveys themselves were unremarkable. But survey are administered by people. And the front-line people who actually administer the survey are poorly-paid, rigidly managed telephone drones working from call centers. My colleague and I visited the call center and surreptitiously listened in on calls. Abused by irate cold call recipients and engaged in a staggeringly boring job reading from scripts (you can see a modified WSJ/NBC script here: https://on.wsj.com/38ozDi2) on a computer display, call center employees tend to get sloppy and creative. We noted all this to the polling agency executives, an act which became known at telecommunications company headquarters as "The Showdown in Indianapolis". The company's contract was not renewed.
I have two thoughts about responding to pollsters, if one should ever call me, and none has: 1) I'm voting straight ticket on the Full-Tilt-Boogie Looney Party; and 2) Who are you trying to reach? I'm Detective Brown, Homicide; how do you know the victim? Just a minute; OK, guys, take out the body.
ReplyDeleteSoooooooooooome day, maybe, I'll get the chance.
During the vietnam war I was called by a pollster and asked various questions about the war. At one point in response to a question I replied that we should bomb Hanoi. The pollster had been trying to lead me to an antiwar point of view but once I said this he just hung up. I assume he completed this entry into the poll the way he wanted to.
ReplyDeleteLuntz is Duntz. Virtually every polling organization will tell you that the fundamental problem is almost no one answers polls since about 2000. The response rate is about 2% and half of them fall into one demographic, white female Millennials. Contrary to his claim, pollsters aren't even reaching Trump supporters to be lied to.
ReplyDeleteThis also ignores polls like Baris and Trafalgar that largely did get it right but were ignored because Trump coming from behind was not the story Big Media wanted to tell.
Probably near 80 million votes for Biden when its done, and near 75 million votes for Trump when its done. That's 14 million more for the Democrats, and 12 million more votes for the Republicans. Looks like a job well done. Trump mobilized voters on all sides, that and perhaps blue cities rioting and burning all summer helped the anti-Democrat vote.
ReplyDeleteI can't imagine how polling can be expected to be very close when there are 40 million people who may vote or not vote, and depending on which set show up. Without 12 million more votes for Trump, surely this could have been massive landslide against Trump. And similarly, if Democrats were sure Trump must lose, perhaps they wouldn't show up and Trump could have won a landslide. Imagine Trump will get 5-6 million more than Obama in 2008. Trump is loved. He is also the biggest loser unfortunately.
No one should have trusted the poll either way, or if anything assume your vote will make the difference even if it won't. And be happy when your side shows up, and you rivals stay asleep. But maybe Republicans will stop convincing themselves that voter suppression is the only way they can win. Now they have 70+ million to play with, thanks to Trump.
Boy, what a pile of cowlot by-product
ReplyDeleteAO - I can't imagine how polling can be expected to be very close when there are 40 million people who may vote or not vote, and depending on which set show up
First, not every poll gets it wrong as shown below. (results from Decision Desk HQ, polls from Trafalgar or RCP as noted).
The Des Moines Register still has a little Midwest decency and concern for its reputation to not put out a biased poll at the end of the race.
Iowa
DDHQ
CANDIDATE VOTES PCT%
Donald J. Trump 896,445 53.23%
Joe Biden 757,758 44.99%
Des Moines Register 48 41 Trump +7
RCP average Trump +2
DDHQ
CANDIDATE VOTES PCT%
Joni Ernst 864,249 51.80%
Theresa Greenfield 753,543 45.16%
Des Moines Register 46 42 Ernst +4
RCP Average Ernst +1.4
Here's the results from the Trafalgar Group. Note that they don't just forecast Trump victories. They also got the Wisconsin result right.
Florida
DDHQ
CANDIDATE VOTES PCT%
Donald J. Trump 5,665,748 51.19%
Joe Biden 5,293,092 47.83%
Trafalgar
Donald Trump 49.4%
Joe Biden 47.3%
RCP Florida November 3rd 47.9 47.0 Biden +0.9
Ohio
DDHQ
CANDIDATE VOTES PCT%
Donald J. Trump 3,074,418 53.36%
Joe Biden 2,603,780 45.19%
Trafalgar
Donald Trump 49.2%
Joe Biden 44.4%
RCP Average 47.3 46.3 Trump +1.0
Wisconsin
DDHQ
CANDIDATE VOTES PCT%
Joe Biden 1,630,579 49.57%
Donald J. Trump 1,610,021 48.94%
Trafalgar
WI Oct 25, 2020 (with leans)
Joe Biden 47.5%
Donald Trump 47.1%
RCP Wisconsin November 3rd 51.0 44.3 Biden +6.7
If you note something about all those results, the RCP average is wildly biased to all the other polls in favor of Biden, and other Democrats. If polling is simply wrong, why do the errors always seem to go in the same direction? How about showing me one or two polls that forecast a *Republican* blow out that turned into a Democrat one. Are there any at all?
and more feedlot runoff
But maybe Republicans will stop convincing themselves that voter suppression is the only way they can win
Republicans, and all Americans, want to have open and transparent elections where the ballots are presented by legitimate voters and counted by honest means. It's not impossible. Look at the states of Florida (after 2000), Iowa, Colorado, Kentucky, Tennessee, Iowa. Now guess what all these states have in common. We all know where the trouble spots are. California, Oregon, Washington, Chicago, Milwaukee, Detroit, Philadelphia. Anywhere a big Democrat machine runs things you can almost guarantee elections will be chaotic because a chaotic system is one that enables fraud to be disguised.
ALSO, as I live in a very rural county, I don't expect to be called by a pollster, and so far, no one has. Not to mention, the 33 years before I moved from previous address, and all other previous addresses before that.
ReplyDelete