Embiggening Embarrasses: "Font Fiasco Tanks Climactic Iowa Poll"

Or should I say, "uncromulent embiggening"?

From Axios (Rebecca Falconer & Mike Allen):

In a Saturday night stunner, the Des Moines Register and CNN scrapped the final Iowa Poll before Monday's caucuses because of fears it was tainted….

"It appears a candidate's name was omitted in at least one interview in which the respondent was asked to name their preferred candidate" [wrote the Register]….

A CNN source told Axios the amazing backstory: An interviewer at the poll's call center increased the font size of the questionnaire on their screen so much that the bottom choice (which rotated between calls) wasn't visible.

Comments

[Krayt]

Assuming they record who the interviewer was, they could re-normalize his answers to the state average from the other interviewers, or just throw them out. It shouldn't cause much difference.

Unless a particular interviewer doesn't interview a random selection but a region, where such choice would necessarily differ vs. state average, but that's a stat invalidity all by itself because one interviewer does one region and another, another, is an area bias can creep in via the interviewer.

[Brett Bellmore]

Unless the issue is that they have no idea which interviewers might have replicated this mistake, and so aren't in any position to do anything about it.

Still, seems like an over-reaction: I used to be a registered Libertarian, and back in the 90's I got polled plenty of times. Virtually every single time the options given me omitted my favorite candidate. I don't recall any of THOSE polls being canceled!

I guess it's different when you falsify the results by accident, instead of deliberately.

[santamonica811]

Brett,
From what you wrote; it seems like your polls were not being conducted with integrity, while the Des Moines one was. It is such a well-respected poll, and it was being looked forward to so much (as the final one before the caucus) . . . we all agree that the people who did the poll definitely *wanted* to release it. It is to their credit (and embarrassment) that, after--what I have to assume--long deliberation, those people decided that there was no way to manipulate the data (to correct for this error) that would (a) end up with a valid result, and (b) give the public the sense that the poll had been conducted fairly.

I always respect when a company voluntarily and publicly "recalls" a product before word has hit the street about that product's defect. Maybe because it's my impression that it happens so rarely. Good for the Des Moines Register. Good for CNN. Their short-term loss is offset by their long-term gain in credibility.

[Brett Bellmore]

I would agree about the "without integrity" bit, but the practice of omitting third party candidates from the polling was (And still is?) routine enough that you have to conclude the industry thought it wasn't an ethical violation.

In this case I agree: The canceled the release because the data wasn't just compromised, it was compromised in an unquantified manner. It was a respectable decision.

[Commenter_XY]

I'm pretty sure this poll company also told us Crooked Hillary was going to win.

[Jeb Kerman]

I read the original story and I think this comes down to; Windows 10 sucks balls.

The number one complaint about Windows 10 since 2015 has been fonts. Windows 10 is made for cell phones and tablets. The fonts in Windows 10 are designed to be able to pivot from landscape to portrait like a cell phone and be read on tiny little cell phone screens.

If like me, you can't read this stuff without glasses, you bump the Windows 10 screen to 125% and use a PC with a 20" screen or larger. But that makes text blurry and for some programs without scroll bars, hides text. Microsoft tried to fix that in 2018 with limited success but for most programs you have to run the compatibility wizard and fix the high DPI scaling.

Per CNN the issue wasn't that one person did this. The problem was that multiple people couldn't read the script and they don't know how many people had to adjust the screen size to read it.

[anorlunda]

Nope. I use the Chrome browser on a Chromebook. No MS Windows involved. I have the same problem.

It's not just scrolling. "Smart browsers" start throwing away whole features such as sidebars, and menu choices, as you zoom in. Unless you sometimes zoom out to check, you'll never know what features of that page you're missing.

To me, the volunteer's error is completely plausible and understandable. Human machine interfaces will never be perfect. When machines take over, one of their arguments will be that humans make too many errors.

[Caphon]

I think similar skullduggery happened in the last presidential election cycle where debate thresholds were established based on polling numbers, a candidates name was left off the poll and he was excluded (Mike Gravel or Gavel). The debate commission had zero sympathy for the “error”.
It’s simultaneously sad and humorous to watch the machinations or DNC primary candidate orchestration.

[LawTalkingGuy]

I move for a bad-poll thingy.

[Bob from Ohio]

Now we will never know who won Iowa.

[Longtobefree]

Whoever Hillary & Obama said should win - - - - - -

[David Bremer]

The only people hurt by this are the pundits, since they can't do the predictable and cliche "Candidate X finished [ahead/behind] what the most recent poll said. How do you explain Candidate X's [collapse/surge]," followed by weeks of stories and narratives this creates.

[Longtobefree]

1. Since 2016, polls are irrelevant
2. This just shows the "reasonable accommodation" of visual disabilities should be scrapped
3. Who cares anyway? It is just a democratic primary, and without secret ballots!

[RobinGoodfellow]

While “embiggening” may be a perfectly cromulent word, uncromulent is not a word.

[Doug Huffman]

All polls are corrupt #FakeNews click-bait. Our Founding Fathers feared mobocracy. The founders of democracy failed at every turn taken to protect it. Read Mogens Herman Hanson Democracy in the Age of Demosthenes, translated from Danish lest one think it corrupted by American politics.

[Brett Bellmore]

When you combine this with the mess the actual caucuses turned out to be, the conspiracy theories almost write themselves.

[tkamenick]

Wait, if the bottom choice was unseen (or less likely to be seen), but that choice rotated or was random, shouldn't that not effect the final results (if the sample size was large enough)?