Thursday, April 22, 2021

Newsbusters and fact checks not published

We at PolitiFact Bias have long noted that fact checkers have a bias against fact-checking things that seem true, or at least publishing fact checks that confirm somebody said something true.

How does that affect an entity's "Truth-O-Meter" report card? Well, depending on how many favorable ratings were spiked, it skews the report card unfavorably. It's a form of selection bias.

And that brings us to NewsBusters (hat tip to Tim Graham).

In brief, Newsbusters published a chart to Facebook. PolitiFact initiated an investigation of the information in the chart, asking NewsBusters to show the specific source of the information. Newsbusters did that, apparently to PolitiFact's satisfaction, and PolitiFact never published any fact check about it.

How often does that happen and to whom? Nobody outside of PolitiFact really knows.

It's just one more symptom of a corrupt fact-checking system. PolitiFact buttonholes Newsbusters with the threat of social media sanction hanging over the latter. If Newsbusters doesn't respond, PolitiFact may issue an unfavorable rating based on its "burden of proof" criterion.

Meanwhile, PolitiFact can preposterously report that real wages are increasing but failing to keep pace with inflation with no plausible threat of sanction to its work.

Tuesday, March 30, 2021

Former PolitiFact expert source Brian Riedl on PolitiFact

 The tweet:

Click the link for the other two tweets in the thread.

Add Riedl's name to a fairly impressive list of similar experiences from people like Michael F. Cannon and Judith Curry.

Sunday, March 7, 2021

Layers of Editors: How fast is PolitiFact's stupidity growing?

Uh-oh! PolitiFact's incompetence unfairly harmed a Democrat again! This time it was hapless Joe Biden who ended up with the short straw by PolitiFact's blinkered judgment.

PolitiFact explained that over the past 10 years the number of Hispanics increased by about 10 million, while the number of Asian Americans went up by 5.2 million.

Why is an increase, on average, of 520,000 per year a faster increase than about 1 million per year?

PolitiFact explains, sort of:

Biden said "the fastest-growing population in the United States is Hispanic." That’s incorrect: The fastest-growing group is Asian Americans, with Hispanics ranking second. Hispanics did record the largest numerical increase in population of any group between 2010 and 2019, but that’s a different measure than "fastest growing."

Instead of recognizing more than one measure of "fastest-growing," PolitiFact arbitrarily accepts one measure while rejecting the other.

But an increase of 1 million per year on average is a rate of growth, and arguably more useful than measuring rate of growth as a percentage of an existing population.

We pointed out on Twitter that PolitiFact's reasoning would suggest that a one foot tall tree that doubles in size is growing faster than a 50 foot tall tree that grows two feet during the same span of time.

Sure, the first tree may surpass the second tree in size if it continues to double in size year-by-year. But it will never happen unless the first tree starts to surpass the second tree in the number of inches of growth per year.


And the math works similarly for population growth. Unless Asian Americans start adding more population in absolute numbers than do Hispanics, the number of Hispanics will forever be greater than the number of Asian Americans. Forever. In fact, Asian Americans will not start closing the gap between the two populations until they start adding more people in raw numbers rather than merely in terms of percentage.

So who do these fact checkers think they are?

Update March 8, 2021: Added the link to the PolitiFact "fact check" in the second paragraph.

Thursday, March 4, 2021

A big problem with PolitiFact's updated website

If it was a snake, it would have bit me.

When "Unwoke Narrative" used a Twitter thread to go off on PolitiFact for the "Mostly False" rating it gave to an Unwoke Narrative Instagram post, one of the issues was a charge PolitiFact used a misquotation.

We questioned Unwoke Narrative about that charge, noting that the supposed misquotation looked like a paraphrase or summary.

Unwoke Narrative made a great point in response. PolitiFact signals to readers that its quotations/summaries/paraphrases of the claims it is checking are quotations.

The deck material of PolitiFact fact checks gets formatted the way many blog templates show quotations. There's a solid vertical bar to the left of the quoted material. And a lone quotation mark in the margin.

Here's an example:

The image of Biden to the upper left occurs in the shape of a word balloon. Just below the Biden balloon and a little to the right we find the lone quotation mark. And to the right of that a vertical yellow line. That's three cues to the reader that what is not a quotation of Biden is a quotation of Biden.

For comparison, have a look at this example from a page designed to help bloggers blog more stylishly:

It's a similar scheme. It features the vertical line, albeit with a skinnier vertical line bracketing the quotation on the right margin. And there's the lone quotation mark just to the left of the quotation.

How did I overlook this for a year?

I thought of one way PolitiFact's format might not mislead people: Maybe it's a pull quote? If it was a pull quote, then PolitiFact could justify putting partial quotations inside a bigger quotation using the standard doubled quotation marks (" instead of ').

But these aren't pull quotes. They're summaries, paraphrases, quotations or sometimes combinations of those formatted as quotations.

We find it unimaginable that professional journalists could find this presentation acceptable. We imagine the experienced journalists at PolitiFact gave the design team too much free rein and then failed to see the problem when it came time to approve the revised format.

It's a deceptive practice and needs to go.

I still can't believe I didn't notice it without having it pointed out to me.

Editor who won't do corrections makes correction request (Updated)

Here and at Zebra Fact Check we have documented obvious errors pointed out to PolitiFact that PolitiFact simply declines to fix.

How delicious, then, when we heard that PolitiFact Editor-in-Chief Angie Drobnic Holan requested corrections of the conservative publication the Federalist. And royally botched at least one of them.

The Federalist noted PolitiFact had done little to fact check Democratic Party politicians in the new Biden administration, and PolitiFact apparently objected.

The Federalist's Tristan Justice told the story on Feb. 20, 2021:

“Your report is wrong in multiple ways,” Holan wrote in a Friday email. “For one, we published three fact-checks on Feb. 16, so our last check on Biden was the day before your report was published. The link you use in the story shows that. Please issue a correction ASAP.”

Holan followed up two hours later, demanding an update.

The link embedded in the initial post to Biden’s fact-checks on the PolitiFact website, however, which remains unchanged, shows PolitiFact published three fact-checks after the story was published in The Federalist, of statements made on Feb. 16, including Biden’s remarks on vaccines. Contrary to Holan’s assertion, these fact-checks were published on Feb. 17 and 18, following the Federalist article.

The Federalist story included an image backing up Justice's description, which we independently verified. It looks like Holan probably mistook the date the claims were made as the publish date. When PolitiFact revamped its website about a year ago, it moved the claim date up to the top of the article and the publish date down and to the left to accompany the writer's byline.

Old way:

The old layout had the date of the claim in the box up top and the publish date next to the author's byline, down at the bottom of the header section.

New way:

With the new layout the claim date moved all the way up to the top, under the source of the claim. The publish date moved, along with the writer's byline, to the left margin. And shrank.

It looks like Holan believed the claim dates were the publish dates. It's not the kind of mistake an organization wants to see from its editor-in-chief.

We asked Holan for an explanation but received no reply.

It's worth pointing out that Holan claimed there were a number of errors in the Federalist article. We also asked her about those, after we asked Justice, the Federalist writer, about them. If Holan found legitimate problems with the Federalist story then Justice should not have buried the fact.

But as for Justice's allegation that Holan was mistaken with the first part of her correction request, the Internet Archive record from Feb. 17, 2021 shows no fact checks featuring Joe Biden published on Feb. 16, 2021.

Update March 5, 2021: A commenter has correctly noted that this post could benefit from added context. We inserted a short new paragraph in the three slot and added the link to the Federalist story that was always intended to be there. We appreciate comments that help us improve our work.

Monday, February 22, 2021

PolitiFact's "In Context" deception (Updated)

In (a) perfect world, fact checkers would publish "In Context" features that simply offer surrounding context with objective explanatory notes.

This ain't no perfect world.

The PolitiFact "In Context" articles tend to serve as editorials, just like its fact checks. Two "In Context" articles from the past year (actually one from 2021 and one from 2019) will serve as our illustrative examples.

The Vaccine Supply

President Biden said "It’s one thing to have the vaccine, which we didn’t have when we came into office, but a vaccinator; how do you get the vaccine into someone’s arm?"

Instead of using context to figure out what Mr. Biden meant or perhaps intended to say, PolitiFact offered that he was not saying there was no vaccine when he took office because elsewhere in the speech he said there were 50 million vaccine doses when he took office ("we came into office, there (were) only 50 million doses that were available"):

You can judge his meaning for yourself, but it’s clear to us that Biden didn’t mean there were no vaccines available before he took office.
So Mr. Biden could have meant anything except for there were no vaccines available when he took office? Oh thank you, Pulitzer Prize-winning fact checkers!

The fact checkers at CNN at least made a game attempt to make heads or tails out of Mr. Biden's words:

Biden made a series of claims about the Covid-19 vaccine situation upon his January inauguration. He said early at the town hall that when "we came into office, there was only 50 million doses that were available." Moments later, he said, "We got into office and found out the supply -- there was no backlog. I mean, there was nothing in the refrigerator, figuratively and literally speaking, and there were 10 million doses a day that were available." Soon after that, he told Cooper, "But when you and I talked last, we talked about -- it's one thing to have the vaccine, which we didn't have when we came into office, but a vaccinator -- how do you get the vaccine into someone's arm?"

Facts First: Biden got at least one of these statistics wrong -- in a way that made Trump look better, not worse, so Biden's inaccuracy appeared accidental, but we're noting it anyway. A White House official said that Biden's claim about "10 million doses a day" being available when he took office was meant to be a reference to the 10 million doses a week that were being sent to states as of the second week of Biden's term, up from 8.6 million a week when they took over.

CNN's "Facts First" went on to explain that the Trump administration released all vaccine reserves to the states instead of holding back the second doses recommended by the manufacturers. CNN also pointed out that the Biden administration continued that same policy.

The CNN account makes it appear Mr. Biden uttered an incoherent mixture of statistics. PolitiFact didn't even make an attempt in its article to figure out what Biden was talking about. PolitiFact simply discounted the statement Biden made that seemed to contradict his dubious claim about the availability of 50 million vaccine doses when he took office.

PolitiFact's "In Context" article looks like pro-Biden spin next to the CNN account. And we thought of another "In Context" article where PolitiFact used an entirely different approach.

Very Fine People

PolitiFact used Mr. Biden's statement about "50 million doses" to excuse any inaccuracy Biden may have communicated by later saying the vaccine cupboard was bare when he took office.

But PolitiFact's "In Context" article about the circumstances of President Trump's reference to "very fine people," published April 26, 2019, made no similar use of Mr. Trump's same-speech clarification "and I’m not talking about the neo-Nazis and the white nationalists -- because they should be condemned totally."

With Biden, readers got PolitiFact's assurance that he wasn't saying there were no vaccine doses when he took office, even though he used words to that effect.

With Trump, readers were left with PolitiFact's curiosity as to what the context might show (bold emphasis added):

We wanted to look at Trump’s comments in their original context. Here is a transcript of the questions Trump answered that addressed the Charlottesville controversy in the days after it happened. (His specific remarks about "very fine people, on both sides" come in the final third of the transcript.)

Not only did PolitiFact fail to use the context to defend Trump from the charge that he was calling neo-Nazis "fine people," about a year later (July 27, 2020) PolitiFact made that charge itself, citing its own "In Context" article in support:

• As president in 2017, Trump said there were "very fine people, on both sides," in reference to neo-Nazis and counterprotesters in Charlottesville, Va.
Making the situation that much more outrageous, PolitiFact declined to correct the latter article when we send a correction request. PolitiFact remained unmoved after we informed the International Fact-Checking Network about its behavior.

Is PolitiFact lucky or what that its owner, the Poynter Institute, also owns the International Fact-Checking Network?

This is how PolitiFact rolls. PolitiFact uses its "In Context" articles to editorially strengthen or weaken narratives, as it chooses.

It's not all about the facts.

Correction: We left out an "a" in the first sentence and also misstated the timing of the two articles our post talks about. Both errors are fixed using parenthetical comments (like this).

Thursday, February 18, 2021

Republican uses wrong embedded URL, receives "False" rating as a reward

Have we mentioned lately that PolitiFact is biased and leans left? Ready for another example?

Buckle up.

Before looking at anything PolitiFact California had to say about House Minority Leader Kevin McCarthy's claim, we had reason to suspect something weird was going on. We knew the CBO estimated potential job losses in the millions for a minimum wage hike to $15 an hour. And PolitiFact often gives partial credit for a valid underlying point. So a politician can give a figure that's substantially off and still get a rating in the "Mostly False" to "Mostly True" range.

In McCarthy's case, it seems that his office put together a press release and used the wrong URL. The Congressional Budget Office put together a report in 2019 estimating the effects of a minimum wage hike to $15 per hour and its high-end estimate for job losses was 3.7 million.

On the other hand, the CBO recently gave a new assessment of raising the minimum wage to $15 per hour and revised its job loss estimates down. The high-end estimate dropped to 2.7 million.

Here's the relevant part of the Feb. 9, 2021 press release:

At this critical point, the Democrats’ big, creative response is to raise the federal national wage to $15 an hour — a move the nonpartisan Congressional Budget Office previously found could cost nearly 4 million workers their jobs. Plus, these job losses would disproportionately hit already economically disadvantaged populations the most.

The new version of the CBO's report came out on Feb. 8, 2021, the day before McCarthy's office published the press release. It's plausible and perhaps likely the press release was largely composed before the CBO released the new report. The press release cited the known report, and staffers hunted up the embedded URL leading to the brand new report.

Of course we do not know that is what happened. Certainly if McCarthy knew of the new report and its updated estimates and stuck with the old estimates--even while using the qualifier "previously"--that counts as misleading.

But here's the thing: PolitiFact concluded, just as we did, that McCarthy appeared to cite the 2019 report. PolitiFact put that reveal front and center in its "If Your Time Is Short" bullet point summary:

  • McCarthy appears to be citing a Congressional Budget Office analysis from 2019 that said, at the high-end, 3.7 million jobs could be lost from the wage hike.

  • But in his statement, he cited the CBO’s new analysis published this week which found an average estimate of 1.4 million jobs would be lost.

Yes, PolitiFact, McCarthy appears to be citing the 2019 CBO analysis. And, yes, the press release "cited" the 2021 report insofar as it embedded the link to that report instead of the 2019 one.

Therefore, PolitiFact said, McCarthy's claim was "False" notwithstanding his valid point that the CBO gave a high-end estimate for job losses in the millions.

Anyone who thinks that's fair needs to learn more about PolitiFact's inconsistencies on rating numbers claims. McCarthy could easily have received a "Half True" and especially so if he was a Democrat making a claim Democrats would like to be true.

Wisconsin Democrat Steve Bullock said a quarter of gun sales avoid background checks. PolitiFact Wisconsin said his claim was "Mostly False" even though it said the true number was only 13 percent. Going by that number, Bullock exaggerated by at least 53 percent.

For comparison, McCarthy's use of the 2019 CBO figure exaggerated the truth by 37 percent. And PolitiFact ruled McCarthy's claim "False."

It's yet another reminder that PolitiFact's "Truth-O-Meter" ratings are subjective and not ultimately rooted in objective findings.

Thursday, February 11, 2021

Story Focus Shenanigans

We've pointed out for years that PolitiFact's story focus often determines the "Truth-O-Meter" rating at the end. Story focus shenanigans never go out of style at PolitiFact.

PolitiFact's method allows its fact checkers any number of ways to approach the same claim. A fact checker might focus on what the claimant said was meant. Or the fact checker could focus on how an audience might perceive the claim. One approach might lead to one "Truth-O-Meter" rating and another approach to a different "Truth-O-Meter" rating. There's no good evidence of any objective criteria guiding the process.

That brings us to two timely examples that help illustrate the phenomenon.

"Mostly False" for the Republican

President Biden set policy to allow essential workers who are undocumented to receive coronavirus vaccines. Why is Scalise's statement "Mostly False"? Apparently because American citizens who are not among the first groups eligible for the vaccine are not waiting to get the vaccine:
But, we wondered, does allowing this population access to the vaccine mean they are being invited to step in front of American citizens in the queue?

PolitiFact weasel-words "in the queue" so that Americans in low priority vaccine eligibility groups are not in the queue at all and are thus not skipped over when undocumented immigrants join those in the high priority groups.

You're not waiting for the vaccine if you're not in that narrowly-defined queue. PolitiFact quoted a Scalise spokesperson who explained his meaning. To no avail. Scalise received a "Mostly False" rating even though his statement was literally true taken in context, with "in the queue" encompassing all American citizens awaiting the vaccine.

"Half True" for the Democrat


Pointedly, PolitiFact does not look at all the ways Mr. Biden's claim fails the test of truth. It does mention some of them, but breezes past such technicalities to point out that IF the person making below $15 per hour is the sole breadwinner in a family of four AND/OR lives in an area with high living expenses THEN they would fall below the poverty level.

How many of those earning less than $15 per hour meet those conditions? Well, if that was important then PolitiFact would have given us a number. Obviously it's not important. What we need to know is that under some conditions Biden's statement could be true. Those missing conditions count as missing context and that matches PolitiFact's definition of "Half True"!

Marvel at PolitiFact's rationalization:

A spokesman for Biden said he was referring to a family of four with one full-time income using the federal government’s poverty guideline, an explanation Biden didn’t include in the interview. Using that measurement, that family with a paycheck of $13 an hour would live below the poverty line. At $15 an hour, the same family would clearly be above the poverty line. So Biden was off by about a dollar, using the existing standards.

But experts said wages alone don’t tell the full story about whether a household lives in poverty. Other factors include child care and housing costs, for example, which can vary by geography. Generalizing a "poverty wage" to a specific number ignores the different circumstances that families face. Other experts said the federal definition of a poverty level is out of date and needs changing.

We rate this claim Half True.

You wonder why similar reasoning couldn't justify a "Half True" for Scalise?

Why do you hate science?