All errors should be reported to DonSurber@gmail.com

Monday, October 29, 2018

Hey pollsters, don't call us, we'll call you

A reader pointed out this line from a Power Line post: "I find it interesting that the Times/Siena had to call nearly 38,000 people to get 737 respondents."

The post was about the latest Senate polls in Arizona, Florida and Indiana.

But the number that stood out was the number of calls they had to make to reach 737 people.

I can see why 98% of the people called either didn't answer or turned down the offer to answer a bunch of personal questions.

Polls have slightly more credibility than New York Times but then again, so does Joe Isuzu. And this was a Times poll done by Siena. That would be a deal breaker for me.

What exactly does the person called get out of the phone call?

I'm retired and I have all the time in the world. But I don't have time for this nonsense. I receive plenty of poll calls. I screen them out along with the Medicare calls. The last poll call I answered turned out to be a push poll by PPP, which I thought was a legitimate pollster.

Also, I already voted, strictly Republican.

Obviously, there is plenty of money in it for the pollster. If reaching 737 people is worth placing 38,000 calls then there is a lot of money riding on.

Ariel Edwards-Levy and Natalie Jackson wrote about this problem in HuffPost two years ago, under the headline, "Do Polls Still Work If People Don’t Answer Their Phones?"

David Dutwin, the executive vice president and chief methodologist for the survey firm SSRS, told them, "If four out of five people hang up on you, it doesn’t make any difference compared to the olden days when only one out of five hung up on you, insofar as the people who hang up are fairly random to the people who don’t hang up.

"And as shocking as it may seem, the research really shows that that’s more true than false."

Other polling experts said the same thing. Response rate does not matter.

But asking a pollster today if polls are any good could be like asking a tobacco company official in the 1940s if cigarettes are safe.

Robert Wuthnow wrote in First Things in August 2015, "Public confidence in the polling industry has dropped dramatically over the past decade. People have grown tired of its ceaseless rollouts, especially during election season. Fewer and fewer of us answer the phone when pollsters call. Response rates have fallen so low that it is impossible to know what exactly polls represent. Even in predicting elections, polls have been missing the mark so badly that poll watchers voice growing suspicion about erroneous methods and potential biases."

He wrote that well before Election Day 2016.

Gallup meanwhile has given up on election polling. The grand-daddy of polling admitted it cannot figure out how to get an accurate way to find out how people are voting.

The cliche is that the only poll that counts is on Election Day.

On behalf of apparently 98% of America, I say to pollsters: don't call us, we'll call you.

Sing it, Sugarloaf.

###

Please enjoy my books in paperback and on Kindle.

Trump the Press covers the nomination.

Trump the Establishment covers the election.

Fake News Follies of 2017 covers his first year as president.

For autographed copies, write me at DonSurber@gmail.com

17 comments:

  1. My favorite poll questionis:Race? "which one? I'm all three..?"

    ReplyDelete
  2. The answering machine is on 24/7 in this house, and I get a lot of callers who hang up after three or four rings. Some of them would be pollsters, no doubt.
    Funny thing is, they only want my opinion until they get my personal data.
    When I have spoken directly to a pollster they always ask a few questions 'for background'. As soon as they learn my age - 55+ - they politely withdraw. 'Thank you sir, but we've already gathered our quota of responses from your age group.'

    ReplyDelete
  3. Never got a poll call in 50 years, nor anyone I know.

    One analyst I know found a Poll service that would ask to speak with "the youngest voter in the house". In 2016 they constantly published Hillary with a large lead.

    ReplyDelete
  4. And how accurate and truthful would that 2 per cent be? That 2 per cent was too dim-witted to realize who was calling and answered the phone for whatever reason.

    That being said, there are three types of lies: Lies, Damned Lies, and Statistics (Polls). And the Left has proven many times over of their propensity to lie. - Elric

    ReplyDelete
  5. I refuse to respond to anything sponsored by the media I hate.

    ReplyDelete
  6. Answered a poll once, a long time ago. Didn’t give them the true answer. Now, I don’t answer the phone unless I know the number. If they want me, leave a message. Will not ever answer a poll again.

    ReplyDelete
  7. And yet if a Rasmussen polls feeds your confirmation bias, we're all about polls. SMH. We've always been at war with Eastasia. Is this what Trumpsterism does to your mind?

    ReplyDelete
    Replies
    1. Rasmussen and Investors Daily were the 2 most accurate polls in 2016.

      Larry Schweikert and Richard Baris had the 2016 election called pretty accurately, but they look at data and don't do polls.

      When you're not making snark comments, are you the one in the back of the class that's 3 years behind after failing, putting your hand in your armpit and making fart noises?

      Delete
    2. G-D that is a great description and I am at a loss to add anything Vote and ride the Red Wave.

      Delete
  8. If I had a nickel for every call that went unanswered, I'd have a fancy car like Don.

    ReplyDelete
  9. If you can get an accurate description of the 2% responders, then you can probably construct a meaningful poll out of a poor response rate. However, I don't think it is possible to get such a description.

    Think about this way- if everyone responded, the accuracy of the poll would be completely dependent on who you called and how truthful those respondents were. However, when 98% of the people don't respond, it means the 2% that do are an idiosyncratic bunch-they are out of the mainstream by definition- and I don't think you can just wave away that difference. It is almost certainly why polling increasingly fails to predict the election outcomes, not only nationally, but in individual state and district races.

    ReplyDelete
    Replies
    1. I used to have a statistics professor in college that told the entire class "If you allow me to word the questions and choose the sample, I can prove to you that Hitler was a humanitarian and a gentle soul.

      Delete
    2. And here's the canonical video clip to demonstrate that: https://www.youtube.com/watch?v=G0ZZJXw4MTA

      Delete
  10. We would hardly expect a polling firm executive to admit any problems with their methods would we?

    Bucky

    ReplyDelete
  11. That's probably $15,000 spent to carry out a poll with a nominal accuracy of around 4%, which is next to worthless when the battle lines are so tightly drawn. To be useful, the number of respondents would have to be 4-5X bigger and correspondingly more expensive. With a low 2% response rate, that means contacting 200 thousand people to get a sample of decent size.

    ReplyDelete
    Replies
    1. Iapetus,

      While you're correct on the necessity of dramatically increasing the sample size, in this type of case that's moot.

      Even with a hugely expanded sample size and even after the enormous effort of attempting to collect opinion information from 200k+ people — you still have the built-in sample bias of (...as Yancey Ward labels them in a 13:10 post...) “idiosyncratic people”.

      When there is so low a response rate anyone answering the questions must, by definition, be idiosyncratic and out-of-the-mainstream in nature. That is a base fault condition that essentially destroys accuracy of a poll.

      Also, consider: in present state our society that respondent most probably falls within one of two groups: pro-question/pro-opinion poll people who will answer honestly and con-opinion poll people who will attempt to give false data.

      The problen then becomes꧇ Just HOW do you adjust for that?

      Admittedly statistics isn't my concentration so I might be mistaken; but I don't see any way to do such an adjustion.

      I say that because imo any "method" used would introduce the statistician's opinion, blowing-up the whole exercise. The fellow might as well “just guess”.

      This whole situation reminds me of that most famous pollster blunder: i.e., the 1936 Dewey/Truman Literary Digest poll. Except the pollsters now supposedly know better.

      Delete
  12. Granting that not every phone number is connecting to a phone with caller ID displayed as it "rings", there are these possibilities:

    1. Name & number displayed.

    2. No name & number displayed.

    3. No name & no number displayed.


    #1 includes the recently appearing name of "scam likely", which is fucking hilariously appearing more and more often, as this person, "scam likely" seems to have quite a large quantity of phone numbers.

    My policy: If I do not recognize either a no name with number I know or a name & number I know, fuck it, I ain't answering. If voice mail kicks in, I !isten, then delete.

    Anyone who is calling me for the first time has been informed to leave a voice mail and I'll return call as soon as I can. That might be within minutes or hours or days.

    Those who answer calls from unknown callers are, to put in gently, fucking retarded.

    Thus, polls are the opinions of the fucking retards who not only answer calls from unknown callers but continue to cooperate with a total stranger who called for no purpose defined by the recepient of said call.

    These fucking scam artists polling firms would get a closer to accurate result if they constructed a framework to solicit calls which they could then do their screening for catagories they find so fascinating, plus a simultaneous on line polling which, while subject to the obvious possibilities of "russian" interference, could, with the ever irritatiing "I am not a robot" monkey dance routine, could result in some interesting data.

    Even with the inputs of those crafty whoevers who say the opposite of what they truly mean.

    And no cost to the volunteers.

    ReplyDelete