Thursday, February 29, 2024

PolitiFact's "no spin" lie

 Apparently PolitiFact's hot new method for getting people to trust their work is to lie to them.

I see it at the top of every fact check these days. PolitiFact posts a summary of its fact check near the top of the story, and at the bottom of that summary lets readers know "No spin, just facts you can trust."

"No spin."

Really?

That howler accompanied a PolitiFact fact check I ran across today, published on Feb. 23, 2024.


PolitiFact's "no spin" approach added spin to the article deck, even before the assurance that PolitiFact has a "no spin" approach: "Do immigrants crossing the US southern border take union jobs? Fact-checking Donald Trump."

The deck claim doesn't match the headline quotation.

So, why the spin?

The second paragraph counts as the key to the fact check:

"The biggest threat to your unions is millions of people coming across the border, because you're not gonna have your jobs anymore," Trump said at the Feb. 17 rally, later adding "The truth is, though, when you have millions of people coming in, they're going to take your jobs."

PolitiFact cited a CSPAN video that clocks in at over an hour. PolitiFact is also the organization that claims it constructs its fact checks to make them replicable. Clue to PolitiFact: If you're trying to allow people to fact-check your work, you tell them where to find key quotations taken from a long video.


 

It's even possible at CSPAN to create a snippet of limited length to include both quotations. We did that.

Trump's making the common sense point that importing millions of low-skilled laborers makes it easy for employers to hire low-wage workers instead of high-wage union workers. PolitiFact turns that point into a straw man, visible at the top of PolitiFact's summary section "If Your Time is Short": "Economy and labor experts told PolitiFact immigrants who recently crossed the U.S. border likely aren't taking Michigan's union jobs."

Did Trump say immigrants who recently crossed the U.S. border were taking Michigan's union jobs? As though an immigrant can run up to Michigan, accept a union job and thereby displace the former holder of that union job? No, that's no what Trump's talking about. He's talking about the general depression of wages that undercuts the stability of an established high-wage union job. Low wages in Kentucky, for example, can eliminate union jobs in Michigan if the employer relocates to Kentucky and hires non-union workers.

If somebody thinks PolitiFact was actually treating Trump's claim exactly the way I suggested it should be taken, the corrective is no further away than the next bullet point in PolitiFact's summary: "(N)ewly arrived migrants are likelier to work in jobs Americans don’t want to do, such as day laborer positions. These aren’t union jobs."

PolitiFact missed Trump's point, whether intentionally or otherwise. The point is immigrants taking jobs Americans don't want to do depresses the value of labor. Cheap labor works its way through the economy, affecting jobs Americans do want to do by making labor cheaper for those jobs as well.

We see a hint of that point in PolitiFact's third bullet point: "There is a correlation between an increase in immigration and a drop in unionization. However, experts said that’s not evidence that immigrants are taking union jobs."

So, what do we do with this claim that an increase in immigration correlates with a drop in unionization in conjunction with the claim that it's not evidence immigrants are taking union jobs?

As noted above, there's ambiguity here. If immigration lowers unionization, that's certainly evidence, albeit not definitive proof, that lower wages from immigration cost the economy union jobs.

It looks like the fact check hinges on an equivocal phrase, "taking union jobs."

But taking Trump's point as we suggested, the fact check affirms Trump's accuracy. PolitiFact included this in its story summary:
(E)xperts agree immigration and union membership numbers move in concert: as immigration rises, unionization drops.

"As immigration rises, unionization drops." That's what earns Trump a "Mostly False" rating instead of "Pants on Fire," I suppose. It arguably makes Trump's claim "Mostly True." 

We consider it unforgivable for a fact checker to leave ambiguity around what is meant by "taking union jobs," and the problem is magnified when the fact checker opts for the interpretation most damaging to the person it is fact-checking.

It's yet another case of uncharitable interpretation, violating the basic interpretive principle of charitable interpretation.

Thursday, February 22, 2024

PolitiFact's how-to primer on improperly fact-checking an analogy

There's so much wrong with this Feb. 22, 2024 PolitiFact fact check that I'm bound to go way beyond the scope of the title.


How To Wrongly Fact Check an Analogy

PolitiFact's summary/quotation of Trump's statement counts as reasonably accurate. He drew an analogy between the fine imposed in the New York fraud case to the political persecution of Alexei A. Navalny, who notably opposed Vladimir Putin in Russian.

At its most basic level, the analogy says Navalny and Trump were treated unfairly in court over politics. But PolitiFact affords nearly zero attention to the basic comparison. Instead, PolitiFact focused on differences as though differences can erase similarities.

Karl Malden's nose remains Karl Malden's nose even if you put it on Emma Stone. And Emma Stone with Karl Malden's nose is Emma Stone having a point of similarity with Karl Malden.

PolitiFact classed Trump's statement (wrongly, we think) as hyperbole, but then justified revoking Trump's license for hyperbole because "we determined there were enough factual elements at play to rate his statement on the Truth-O-Meter."

We're not sure how that's supposed to work. As we noted on X, PolitiFact could use a similar approach to find a vegetarian "Pants on Fire" for comparing vegetarian bacon to regular bacon. The bacon example came straight from a dictionary definition of "analog."

Here We Go Again: "Experts"

Part of PolitiFact's schtick comes from its interviews of experts. Typically the pool of experts leans left, and often has a record of giving politically to Democrats. For some reason PolitiFact doesn't see that as a mark against its own credibility.

Let's take a look, shall we?

Harley Balzer
Highly partisan political giving. One of the most partisan records we've encountered, and that's really saying something.

Erik Herron
We found no political giving under Herron's name. But we did find an X post by Herron that appears to acknowledge the legitimacy of analogies where the comparison may seem strained.


Ric Simmons 
Simmons (employed at The Ohio State University) has two political donations listed. One was to Democrat Joe Biden and the other to the anti-Trump group "The Lincoln Project."

Scott Gehlbach
Gehlbach's partisan political giving fails to challenge that of Balzer, but it's solidly behind Democrats with the exception of one nonpartisan figure, now a (liberal) judge in the Wisconsin court system.

Stephen Sestanovich
Sestanovich has six donations, minimum $250, all going to Democrats.

Kathryn Hendley
Hendley has only one political donation listed, with a Democrat listed as the recipient of the $200 gift.

Mark Osler
Osler has given exclusively to Democrats, with six donations in the range of $50-$250.

What are the chances a fact checker can find seven expert sources and six out of seven have given exclusively to Democrats? It's as though PolitiFact intentionally seeks out Democrats to serve as its experts.

Of course, the mere fact that the experts give to Democrats should not discredit their expertise. But PolitiFact simply uses the experts to underscore that the Navalny case is different than the Trump case. We don't need experts to prove that, and as we pointed out above, differences are irrelevant to the similarities, The former cannot erase the latter.

PolitiFingers on the Scale

As if distracting from the point of Trump's argument and using partisan experts wasn't enough, we have PolitiFingers on the scale of this fact check.

PolitiFact omits all mention of two significant aspects of the fraud case against Trump. Both aspects tend to support the Navalny analogy.

First, the trial judge found that Trump's fraud did not damage anyone financially. That makes the prosecution and the judgment unusual. The fine represents higher conjectured interest charges from lower valuations of Trump properties. We doubt such a basis has ever before been used in the United States to support a fraud penalty.

USAToday:

(Gregory) Germain, the Syracuse professor, said the government did a good job of showing Trump inflated the value of his properties, but noted that sophisticated financial institutions didn't require a third-party appraisal like they do for a typical mortgage on a home.

"There are no cases like it," Germain said.


An Associated Press story makes a related point:

And though the bank offered Trump lower interest rates because he had agreed to personally guarantee the loans with his own money, it’s not clear how much better the rates were because of the inflated figures. The bank never complained, and it’s unclear how much it lost, if anything. Bank officials called to testify couldn’t say for sure if Trump’s personal statement of worth had any impact on the rates.

“This sets a horrible precedent,” said Adam Leitman Bailey, a New York real estate lawyer who once successfully sued a Trump condo building for misrepresenting sales to lure buyers.

Second, PolitiFact's fact check misrepresents the ease of appealing the ruling. 

CNBC:

Former President Donald Trump is gearing up to fight a massive fine in the New York business fraud case that threatens to erase most of the cash he says he has on hand.

But first, he has to secure a bond — and that might not be so easy.

Why doesn't PolitiFact tell you any of that?

Because they're biased.

They make sure there are no observations from a conservative such as Andrew C. McCarthy:

Afters:

PolitiFact is on a real tear against Trump early in 2024. It's almost like they're trying to retroactively make true their false claims about Trump's "Truth-O-Meter" record.

In fact it was Louis Jacobson, listed first on the byline of PolitiFact's fact check, who recently endured two corrections from Slate after it published an interview with him. Jacobson made two flatly false claims about Trump's record on the "Truth-O-Meter."


One wonders whether publicly making false claims about Trump should disqualify Jacobson from working on fact checks involving Trump.

Note: Huh--Looks like Slate botched its editor's note: "It has also been updated to clarify that among major politicians frequently fact-checked by PolitiFact, Trump has the highest percentage of Pants on Fire ratings." I gave them the example of Michele Bachmann, who has had 72 "Truth-O-Meter" ratings.

Hmm. Looks like it's time for another correction request, if there's no clear justification for that claim.

Thursday, February 8, 2024

The "Pants on Fire" bias study updated through 2023

 We have updated our "Pants on Fire" bias study with data from 2023.

What is it? We use a spreadsheet to track all "False" and "Pants on Fire" ratings given to partisan Republicans or Democrats whether candidate, officeholder or appointed administration official plus party officials or organizations. We then calculate the percentage of false ("False" plus "Pants on Fire") ratings given the "Pants on Fire" rating.

Why do we do it? Because PolitiFact has never offered an objective means of distinguishing its "False" rating from its "Pants on Fire" rating, we infer that the difference is either substantially or wholly subjective. Assuming the substantial subjectivity of the ratings, we expect that differences in the percentages will help identify PolitiFact's partisan bias, if any.

Here's the updated chart:


What have we learned so far?

We've learned that national PolitiFact after 2007 shows a consistent bias for Democrats/against Republicans. That trend shows poorly on the graph above because this graph includes ratings from PolitiFact's various state operations. Before PolitiFact changed its website making it far less clear which franchise was responsible for what, we kept track of each part of the organization separately. The years from 2010 through 2015 show a moderation of bias thanks to state operations that sometimes were legitimately tough on Democrats. PolitiFact Wisconsin was notably tough on Democrats during that period, for example.

By looking at the total number of various ratings given to the political parties, we've also noted that Republicans (after 2007) receive far more of PolitiFact's bottom two ratings. That effect may stem from Republicans lying more or simply because of bias in story selection and ratings. We've documented enough of the latter two factors to reasonably prefer the second option. That's where the evidence leads.

If, as the available evidence suggests, PolitiFact's "Pants on Fire" rating has no objective basis, "Republicans lie more" carries no objective explanatory value respecting the percentages on our graph.

We've also learned that harsh ratings for both parties are on the decline, in terms of raw numbers. The most obvious explanation for that trend stems from PolitiFact's social media partnerships. If PolitiFact fact checks a politician, revenue consists of donations, grants and ad revenue. But if PolitiFact fact checks something for its social media partners, there's a payday for that. PolitiFact discloses that more than 5 percent of its revenue comes from the social media company Meta. The Chinese social media company TikTok likewise accounts for over 5 percent of PolitiFact's revenue.  

Why doesn't PolitiFact offer more transparency than that regarding its income? Good question, but we don't have an answer free of conjecture.

As for our study of PolitiFact's numbers in 2023, the Republican average fell well below its historic norm, establishing an all-time low for the GOP. PolitiFact's ratings of Democrats pulled their historic average down for the eighth straight year.

A potential weird Trump effect?

The percentages for Republicans haven't really changed much over the years, defying the existence of any Trump effect in terms of increasing Republican dishonesty (in PolitiFact's data, anyway). But the percentages for Democrats have declined noticeably since around 2016 as Trump ascended politically.

Could Trump help explain an increase in Democratic Party honesty?

More likely those changes happen because the makeup of PolitiFact's franchises has shifted over time. State franchises no longer take the edge off the pro-Democrat bias of national PolitiFact.