Monday, February 24, 2020

Nothing To See Here: Sanders blasts health insurance "profiteering"

While researching PolitiFact's false accusation that Democratic presidential candidate used "bad math" to criticize the budget gap created by fellow candidate Bernie Sanders' spending proposals, we stumbled over a claim from Sen. Sanders that was ripe for fact-checking.

Sanders said his proposed health care plan would end profiteering practices from insurance and drug companies that result in $100 billion or so in annual profits (bold emphasis added):
Just the other day, a major study came out from Yale epidemiologist in Lancet, one of the leading medical publications in the world. What they said, my friends, is Medicare for all will save $450 billion a year, because we are eliminating the absurdity of thousands of separate plans that require hundreds of billions of dollars of administration and, by the way, ending the $100 billion a year in profiteering from the drug companies and the insurance companies.
PolitiFact claims to use an "Is that true?" standard as one of its main criteria for choosing which claims to check.

We have to wonder if that's true, or else how could a fact checker pass over the claim that profiteering netted $100 billion in profits for those companies? Do fact checkers think "profit" and "profiteering" are the same thing?

Is a fact checker who thinks that worthy of the name?

Sanders' claim directly implies that the Affordable Care Act passed by Democrats in 2010 was ineffective with its efforts to circumscribe insurance company profits. The ACA set limits on profits and overhead ("medical loss ratios"). Excess profits, by law, get refunded to the insured.

Sanders said it's not working. And the fact checkers don't care enough to do a fact check?

Of course PolitiFact went through the motions of checking a similar claim, as we pointed out. But using "profiteering" in the claim changes things.

Or should.

Ultimately, it depends on whether PolitiFact has the same interest in finding falsehoods from Democrats as it does for Republicans.

Sunday, February 23, 2020

PolitiFact absurdly charges Pete Buttigieg with "bad math"

PolitiFact gave some goofy treatment to a claim from Democratic presidential candidate Pete Buttigieg.

Buttigieg compared the 10-year unpaid cost of fellow candidate Bernie Sanders' new spending proposals to the current U.S. GDP.

PolitiFact cried foul. Or, more precisely, PolitiFact cried "bad math."


Note that PolitiFact says Buttigieg did "bad math."

PolitiFact's fact check never backs that claim.

If Buttigieg is guilty of bad anything, it was a poor job of providing thorough context for the measure he used to illustrate the size of Sanders "budget hole." Buttigieg was comparing a cumulative 10-year budget hole with one year of U.S. GDP.

PolitiFact notwithstanding, there's nothing particularly wrong with doing that. Maybe Buttigieg should have provided more context, but there's a counterargument to that point: Buttigieg was on a debate stage with a sharply limited amount of time to make his point. In addition, the debate audience and contestants may be expected to have some familiarity with cost estimates and GDP. In other words, it's likely many or most in the audience knew what Buttigieg was saying.

Let's watch PolitiFact try to justify its reasoning:
But there’s an issue with Buttigieg’s basic comparison of Sanders’ proposals to the U.S. economy. He might have been using a rhetorical flourish to give a sense of scale, but his words muddled the math.

The flaw is that he used 10-year cost and revenue estimates for the Sanders plans and stacked them against one year of the nation’s GDP.
PolitiFact tried to justify the "muddled math" charge by noting Buttigieg compared a 10-year cost estimate to a one-year figure for GDP.

But it's not muddled math. The 10-year estimates are the 10-year estimates, mathematically speaking. And the GDP figure is the GDP figure. Noting that the larger figure is larger than the smaller figure is solid math.

PolitiFact goes on to say that the Buttigieg comparison does not compare apples to apples, but so what? Saying an airplane is the size of a football field is also an apples-to-oranges comparison. Airplanes, after all, are not football fields. But the math remains solid: 100 yards equals 100 yards.

Ambiguity differs from error

In fact-checking the correct response to ambiguity is charitable interpretation. After applying charitable interpretation, the fact checker may then consider ways the message could mislead the audience.

If Buttigieg had run a campaign ad using the same words, it would make more sense to grade his claim harshly. Such a message in an ad is likely to reach people without the knowledge base to understand the comparison. But many or most in a debate audience would understand Buttgieg's comparison without additional explanation.

It's an issue of ambiguous context, not "bad math."

Friday, February 21, 2020

PolitiFact's dishonest dedication to the "Trump is a liar" narrative

It's quite true that President Trump makes far more than his share of false statements, albeit many of those represent hyperbole. Indeed, it may be argued that Trump blurs the line between the concepts of hyperbole and deceit.

But Trump's reputation for inaccuracy also serves as a confirmation bias trap for journalists.

Case in point, from PolitiFact's Twitter account:


The tweet does not tell us what Trump said about windmills and wildlife, though it links to a supposedly "similar claim" that it fact-checked in the past.

That fact check concerned something Trump said about the number of eagles killed by wind turbines:



The linked fact check had its own problems, which we noted at the time.

One of the things we noted was that PolitiFact gave short shrift to the facts to prefer advancing the narrative that Trump says false things:
PolitiFact's interpretation lacks clear justification in the context of Trump's remarks, but fits PolitiFact's narrative about Trump.

A politician's lack of clarity does not give fact checkers justification for interpreting statements as they wish. The neutral fact checker notes for readers the lack of clarity and then examines the possible interpretations that are at the same time plausible. The neutral fact checker applies the same standard of charitable interpretation to all, regardless of popular public narratives.
PolitiFact's tweet amplifies the distortion in its earlier fact check. Trump said wind turbines kill eagles by the hundreds. PolitiFact made a number of assumptions about what Trump meant (for example, assuming Mr. Trump's "by the hundreds" referred to an annual death toll) then produced its subjective "Mostly False" rating based on those assumptions.

Did Trump say something comparable in Colorado?

PolitiFact's tweet communicates to readers that Trump uttered another mostly falsehood in Colorado. But what did Trump say that PolitiFact found similar to saying wind turbines kill hundreds of eagles every year?

Here's what Trump said in Colorado, via Rev.com (bold emphasis added):
We are right now energy independent, can you believe it? They want to use wind, wind, wind. Blow wind, please. Please blow. Please keep the birds away from those windmills, please. Tell those beautiful bald eagles, oh, a bald eagle. You know, if you shoot a bald eagle, they put you in jail for a long time, but the windmills knock them out like crazy. It’s true. And I think they have a rule, after a certain number are killed you have to close down the windmill until the following year. Do you believe this? Do you believe this? And they’re all made in China and in Germany. Siemans.
Got it? "Knocking (Bald Eagles) out like crazy"="(killing eagles) by the hundreds"

How many is "like crazy"? Pity the fact checker who thinks that's a claim a fact checker ought to check. If wind turbines kill tens of Bald Eagles instead of hundreds, that can support the opinion that the turbines kill the eagles "like crazy," particularly given the context.

It's hard to argue that Trump said something false about Bald Eagles in his Colorado speech, yet PolitiFact did just that, relying largely on hiding from its audience what Trump actually said.

How many eagles do wind turbines kill? That's hard to say. But federal permits for wind farming potentially allow for dozens of eagle deaths per year:
(Reuters) - Wind farms will be granted 30-year U.S. government permits that could allow for thousands of accidental eagle deaths due to collisions with company turbines, towers and electrical wires, U.S. wildlife managers said on Wednesday.
Does it follow that Trump said something "Mostly False" in Colorado?

Or are the fact checkers at PolitiFact once again chasing their narrative facts-be-damned?


Thursday, February 20, 2020

PolitiFact weirdly unable to answer criticism

Our title plays off a PolitiFact critique Dave Weigel wrote back in 2011 (Slate). PolitiFact has a chronic difficulty responding effectively to criticism.

Most often PolitiFact doesn't bother responding to criticism. But if it makes its liberal base angry enough sometimes it will trot out some excuses.

This time PolitiFact outraged supporters of Democratic (Socialist) presidential candidate Bernie Sanders with a "Mostly False" rating of Sanders' claim that fellow Democratic presidential candidate Michael Bloomberg "opposed modest proposals during Barack Obama’s presidency to raise taxes on the wealthy, while advocating for cuts to Medicare and Social Security."

Reactions from left-leaning journalists Ryan Grim and Ryan Cooper were typical of the genre.



The problem isn't that Sanders wasn't misleading people. He was. The problem stems from PolitiFact's inability to reasonably explain what Sanders did wrong. PolitiFact offered a poor explanation in its fact check, appearing to reason that what Sanders said was true but misleading and therefore "Mostly False."

That type of description typically fits a "Half True" or a "Mostly True" rating--particularly if the subject isn't a Republican.

PolitiFact went to Twitter to try to explain its decision.

First, PolitiFact made a statement making it appear that Sanders was pretty much right:



Then PolitiFact (rhetorically) asked how the true statements could end up with a "Mostly False" rating. In reply to its own question, we got this:
Because Sanders failed to note the key role of deficit reduction for Bloomberg.
Seriously? Missing context tends to lead to the aforementioned "Mostly True" or "Half True" ratings, not "Mostly False" (unless it's a Republican). Sanders is no Republican, so of course there's outrage on the left.

Anyway, who cuts government programs without having deficit reduction in mind? That's pretty standard, isn't it?

How can PolitiFact be this bad at explaining itself?

In its next explanatory tweet PolitiFact did much better by pointing out Bloomberg agreed the Obama deficit reduction plan should raise taxes, including taxes on wealthy Americans.

That's important not because it's on the topic of deficit reduction but because Sanders's made it sound like Bloomberg opposed tax hikes on the wealthy at the federal level. Recall Sanders' words (bold emphasis added): "modest proposals during Barack Obama’s presidency to raise taxes on the wealthy."

Mentioning the proposals occurred during the Obama presidency led the audience to think Bloomberg was talking about tax hikes at the federal level. But Sanders was talking about Bloomberg's opposition to tax hikes in New York City, not nationally.

PolitiFact mentioned that Bloomberg had opposed the tax hikes in New York, but completely failed to identify Sanders' misdirection.

PolitiFact's next tweet only created more confusion, saying "Sanders’ said Bloomberg wanted entitlement cuts and no tax hikes. That is not what Bloomberg said."

But that's not what Sanders said. 

It's what Sanders implied by juxtaposing mention of the city tax policy with Obama-era proposals for slowing the growth of Medicare and Social Security spending.

And speaking of those two programs, that's where PolitiFact really failed with this fact check. In the past PolitiFact has distinguished, albeit inconsistently, between cutting a government program and slowing its growth. It's common in Washington D.C. to call the slowing of growth a "cut," but such a cut from a higher growth projection differs from cutting a program by making its funding literally lower from one year to the next. Fact checkers should identify the baseline for the cut. PolitiFact neglected that step.

If PolitiFact had noted that Bloomberg's supposed cuts to Social Security and Medicare were cuts to future growth projections, it could have called out Sanders for the misleading imprecision.

PolitiFact could have said the Social Security/Medicare half of Sanders' claim was "Half True" and that taking the city tax policy out of context was likewise "Half True." And if PolitiFact did not want to credit Sanders with a "Half True" claim by averaging those ratings then it could have justified a "Mostly False" rating by invoking the misleading impression Sanders achieved by juxtaposing the two half truths.


 Instead, we got yet another case of PolitiFact weirdly unable to to answer criticism.

Monday, February 10, 2020

Nothing To See Here: Stephanopoulos Interviews Joe Biden

Democratic presidential candidate Joe Biden appeared on "This Week" with interviewer George Stephanopoulos on Feb. 9, 2020.

Biden made a number of questionable claims during the interview, particularly where he claimed President Donald Trump has never condemned white supremacy (Washington Examiner).

Biden also said the 2009 stimulus bill passed by Democrats and the Obama administration had no waste or fraud to it.

For our money, a left-leaning operation like PoliitFact is likely to ignore Biden's claims on "This Week" in favor of getting to the bottom of whether Biden was quoting actor John Wayne when he ("jokingly") called a woman a "lying dog-faced pony soldier."

I guess we'll see!


Sunday, February 9, 2020

PolitiFact's charity for the Democrats

PolitiFact is partial to Democrats.

Back in 2018 we published a post that lists the main points in our argument that PolitiFact leans left. But today's example doesn't quite fit any of the items on that list, so we're adding to it:

PolitiFact's treatment of ambiguity leans left

When politicians make statements that may mean more than one thing, PolitiFact tends to see the ambiguity in favor of Democrats and against Republicans.

That's the nature of this example, updating an observation from my old blog Sublime Bloviations back in 2011.

When politician say "taxes" and does not describe in context what taxes are they talking about, what do they mean?

PolitiFact decided the Republican, Michele Bachmann, was talking about all taxes.

PolitiFact decided the Democrat, Marcia Fudge, was talking about income taxes.

Based on the differing interpretations, Bachmann got a "False" rating from PolitiFact while Fudge received a "True" rating.

That brings us to the 2020 election campaign and PolitiFact's not-really-a-fact-check article "Fact-checking the Democratic claim that Amazon doesn't pay taxes."

The article isn't a fact check as such because PolitiFact skipped out on giving "Truth-O-Meter" ratings to Andrew Yang and Sen. Elizabeth Warren. Both could easily have scored Bachmannesque "False" ratings.


Yang and Warren both said about the same thing, that Amazon paid no taxes.

Various news agencies have reported that Amazon paid no federal corporate income taxes in 2017 and 2018. But news reports have also made clear that Amazon paid taxes other than federal corporate income taxes.


Of course neither Yang nor Warren will receive the "False" rating PolitiFact bestowed on Bachmann for a comparable error. PolitiFact treated both their statements as though they restricted their claims to federal corporate income tax.

Is it true that despite making billions of dollars, Amazon pays zero dollars in federal income tax?

Short answer: Amazon’s tax returns are private, so we don’t know for sure what Amazon pays in federal taxes. But Amazon’s estimates on its annual 10-K filings with the U.S. Securities and Exchange Commission are the closest information we have on this matter. They show mixed results for the past three years: no federal income tax payments for 2017 and 2018, but yes on payments for 2019.

That's the type of impartiality a Democrat can usually expect from PolitiFact. They do not need to specify what kind of taxes they are talking about. PolitiFact will interpret their statements charitably. 

Afters

It's worth noting that PolitiFact admitted not knowing whether Amazon paid federal income taxes in 2017 and 2018 ("we don’t know for sure what Amazon pays in federal taxes"). And PolitiFact suspends its "burden of proof" criterion yet again for Democrats.


Feb. 10, 2020: Edited to remove a few characters of feline keyboard interference.

Sunday, February 2, 2020

PolitiFact updates its website, makes "Corrections and Updates" even harder to find (Updated: Fixed!)

Over the years we've enjoyed poking fun at PolitiFact's haphazard observance of its policy on corrections. Aside from simply not doing quite a few needed corrections, PolitiFact does things like:
  • Correcting stories without a correction notice
  • Not tagging stories with "Corrections and Updates" as promised in its statement of principles
We've also needled PolitiFact over the way it hides its supposedly transparent page of corrected or updated claims. Looking up "corrections and updates" along with "PolitiFact" using a search engine would allow readers to easily find the page, but finding that page from PolitiFact's home page was so hilariously complicated that we posted instructions on how to do it.

Now in February 2020 PolitiFact has revamped its website and at long last fixed the problem succeeded in making the problem even worse.

Hopefully the worsening of the situation is only temporary, but PolitiFact's history marinates that hope in thick, gooey skepticism.

Our Feb. 1, 2020 survey of the PolitiFact website makes the "Corrections and Updates" page look like an orphan.

We tried to help. Seriously.

When I (Bryan) heard on Twitter that PolitiFact was updating its website, I tweeted out a reminder for PolitiFact to make its "Corrections and Updates" page more available to readers:


Instead of fixing it, the "Corrections and Updates" page is one of the very few (this is the only other one we found) that did not experience a facelift in keeping with the new look of the website.

For our money, the redesign looks pretty bad on the big screen. And it's not much better on mobile.

But one thing we did like, though perhaps that means it won't last.

What We Liked


In addition to PolitiFact's dodgy behavior on corrections, we've endlessly criticized PolitiFact for publishing sciencey-looking graphs of aggregated "Truth-O-Meter" ratings with no regular disclaimer attached. The ratings are subjective and PolitiFact does not attempt to choose a scientifically representative sample of claims. So the graphs are nonsense in terms of representing a politician's overall reliability.

PolitiFact still isn't attaching any disclaimer, but the new design largely neuters the visual impact of its graphs.

Let's look at Donald Trump's PolitiFact "scorecard" before and after the update.

Before


That has some visual impact. The graph groups the bars closely, emphasizing the visual difference between, say, 5 percent and 35 percent.

After


What a difference! Thirty-four percent looks visually smaller on the new graph than 5 percent looked on the old graph. Sure, there's an attempt to spice it up by adding colors, but the short graph scale and thin lines suck away almost all of the impact.

Do we think PolitiFact did this intentionally so that the graphs would do less to mislead readers? No, unless it's part of an effort to farm out the deception.* But if it stands, it doesn't really matter if it's an accident or a mistake. PolitiFact will probably deceive fewer of its readers as a result.

One Other Positive!

PolitiFact has always offered the total number of each rating for individual politicians. But now it is publishing the totals for PolitiFact as a whole, as well as the states.

That makes doing certain types of research on PolitiFact's numbers easier. Though researchers will still need to realize that the subjectivity of the ratings means they tell researchers about PolitiFact, not so much about the politicians making the claims.

It's a very simple matter now to document how many more ratings Donald Trump has received than did Barack Obama, and in a shorter span of time as candidate/president.

*The downside? Those who are motivated to use PolitiFact "data" to prove Republicans are liars and whatnot will have less work to do in collecting the numbers. People irresponsibly publishing such nonsense may end up misleading more people in spite of the positives we noted.


Afters

Centered text? Seriously?


Updated Feb. 3, 2020 with edits thought already complete: strikethrough and URL linking earlier PFB post about finding PolitiFact's "Corrections and updates" page.

Update Feb 4, 2020: Whether it was the plan all along or whether in response to our cajoling, PolitiFact has added "Corrections and Updates" to its menu, under the heading "About Us."  That fixes one of our major complaints about PolitiFact. Now will PolitiFact add corrected "articles" to its list of corrected stories?