Grading PolitiFact: Chris Coons and his phantom fact

The issue:

The fact checkers:

Louis Jacobson:  writer, researcher
Martha Hamilton:  editor

Analysis:

This story by PolitiFact counts as one of the more unusual ones I've evaluated.  Review the graphic PolitiFact uses to summarize the story.  "New Castle's unemployment rate has not almost doubled in the last two years."  "In debate, Chris Coons says Christine O'Donnell is wrong on county unemployment rate."

What did Coons actually say?  In response to opponent O'Donnell's claim that unemployment had doubled in the past two years for New Castle County, he said this (according to PolitiFact, bold emphasis added):

A moment later, Coons countered, "I also frankly can't imagine where she found the numbers that unemployment doubled in just the past year under my watch. I suspect we're going to need to keep a close eye this evening on the numbers that go flying back and forth."

Is "New Castle's unemployment rate has not almost doubled in the last two years" in there anywhere?

Did Coons claim that O'Donnell was wrong about the county unemployment rate?

The answers are no and no.  Coons made an ambiguous reply that implies skepticism about O'Donnell's claim but falls short of contradicting it.  Given that Coons did not directly contradict O'Donnell, why not fact check O'Donnell and deal with the claim directly?  A good question, but not one we're likely to see answered.

Sounds like a dispute tailor-made for PolitiFact: How much did unemployment rise in New Castle County, the most populous of Delaware's three counties?

In one year or two, we might wonder.  But not to worry.  PolitiFact has selection bias on hand to guide us through this difficult process:

Lets first clarify a few matters. In Coons' reply, he misstated the time frame O'Donnell had used, saying it was one year when in fact she had said she was looking at the past two years. We'll analyze the question using O'Donnell's criteria -- two years -- while noting that Coons garbled O'Donnell's time frame.

There's nothing wrong with a little garbling, is there?

More:

We also won't get into the question of whether a county executive deserves direct blame for rising unemployment, even though one can argue that unemployment is more sensitive to national and international economic factors than local ones.

Coons' detrimental influence on unemployment, if any, would have been O'Donnell's underlying argument beneath the unemployment numbers she cited.  I'd count that as a bad argument without mention of some policy from Coons that reasonably might affect unemployment.  Otherwise it smacks of the cum hoc ergo propter hoc fallacy.  But this isn't about how I would have checked the facts.  This is about how I second guess how other people check facts.

O'Donnell said she got her stats from the Department of Labor, so PolitiFact sensibly paid a visit to the website of the Bureau of Labor Statistics, an arm of the Department of Labor.

In the most recent available month, August 2010, New Castle County, Del., registered an 8.7 percent unemployment rate. Two years before, in August 2008, the rate was 5.5 percent. (The figures are not seasonally adjusted, but since we're comparing August of one year to August of another, making seasonal adjustments should be irrelevant in this comparison.)

PolitiFact found that the 3.3 percentage point difference represents a 58 percent increase in unemployment.  And either for reasons of selective reading or selection bias, PolitiFact ignored comparisons for other months.  It may be that the PolitiFact researcher (Louis Jacobson) accepted a narrow set of search parameters at the bls.gov site.  I selected a search method providing a broader view.

Here's what I found (click image for enlarged view):

The information has a couple of significant implications in terms of fact checking.

1)  PolitiFact used a preliminary figure (from 2010) for the August comparison and did not inform the reader.
2)  For five out of the eight months represented in the data, O'Donnell would have been understating the increase in unemployment.  An increase over 100 percent represents a more than doubled rate of unemployment.

By completely overlooking the numbers for the rest of the year, caused by the decision to focus on the August 2008-August 2010 comparison, PolitiFact made O'Donnell's claim seem far off.  It was primarily the increase in unemployment from May through August of 2008 that pushed unemployment below "close to double" the rate from two years ago.  For July, the difference of 66 percent might qualify as "almost doubled" and the June figure of 82 percent ought to reasonably qualify.

Precisely why was the comparison kept to August in the first place?  Other than to explicitly benefit Coons, of course.

The historical record shows that data might easily justify a claim of doubled unemployment between 2008 and 2010.

PolitiFact kept its focus on August, of course:

That's a difference of 3.3 percentage points, or an increase of 58 percent over the August 2008 unemployment rate of 5.5 percent. If the rate had doubled as O'Donnell said, it would have needed to increase by 100 percent, which would have brought the unemployment rate to 11 percent today. And it clearly isn't that high.

"If the rate had doubled as O'Donnell said ..."  I thought O'Donnell said "almost doubled"?  I guess PolitiFact decided to go with the garbled Coons version where he expressed doubt that the rate had doubled rather than "nearly doubled."

Ready for the incredible conclusion?

While O'Donnell is correct that unemployment has risen in New Castle County, as it has everywhere, the increase wasn't almost double over two years -- the period she said during the debate. By saying that she meant to measure it from the beginning of Coons' term as county executive, her campaign team is essentially conceding that she got the fact wrong. In our role as debate referee, we give Coons' stance that O'Donnell erred a rating of True.

As shown above, the unemployment rate more than doubled over the past two years, depending on when the measurement begins and ends.  Whether by accident or not, PolitiFact cherry-picked to Coons' benefit.  And PolitiFact is simply wrong that O'Donnell concedes that she got the fact wrong by stipulating what she meant to say.  She could, without logical contradiction, be correct both in terms of what she said and what she meant to say.  What she said is only necessarily wrong in terms of not being what she intended to say.

Neither candidate covered themselves with glory in this exchange.  O'Donnell should have said what she meant, and either backed it up on her website as promised or provided an explanation.  Coons defended himself with skepticism but received the benefit of very generous fact-checking from PolitiFact.  PolitiFact invented a claim on his behalf and then made the invented claim look much truer than it was.  As it turned out, Coons' skepticism could only receive support with cherry-picked numbers.

The grades:

Louis Jacobson:  F
Martha Hamilton:  F

Coons gets a "True" after changing O'Donnell's statement by a year and rounding up from "almost doubled" to "double"?  Must be nice.

The failing grades are dispensed for the tunnel-vision treatment of the data, failure to deal forthrightly with what Coons' actually said and for faulty logic in supposing that O'Donnell's admission of stating something other than what she meant to say means that what she said was therefore not factual.  And I could throw in the fact that the underlying argument is actually the important thing in cases like this.  O'Donnell made a statement that, without considering the underlying argument, should have been ruled at least "Barely True."  But even if it had been 100 percent true it doesn't necessarily reflect poor performance by Coons in office.

Oct. 18, 2010:  Corrected inaccurate quotations of O'Donnell--they were OK as paraphrases but not as quotations.

Afters:

I met with some skepticism the PolitiFact claim that seasonal adjustment on the unemployment numbers would make no difference.  Seasonal adjustments were not available specific to New Castle County.  The 2008 and 2010 numbers for August bumped up by .1 for the state of Delaware overall.  That results in a very slight decrease by percentage, perhaps an illusory decrease.

What about "real unemployment," what the Bureau of Labor Statistics calls "special unemployment"?  Those numbers are considerably more difficult to pin down, especially at the state and local level.  I get the impression that real unemployment changes largely mirror the changes in employment as measured by the BLS unemployment number.  In other words, the percentage change probably would not vary greatly using the real employment measure.