Lil White Lies: Double Bubble Bias Trouble

In his latest PolitiFact critique, Bryan White brews up what appears to be a pretty open and shut case about “selection bias” at PolitiFact. In a televised debate, Delaware Democrat Senatorial candidate Chris Coons disagrees with a statement by his opposing Republican candidate Chris O’Donnell that unemployment in New Castle County (one of Delaware’s three counties) has doubled in the last two years. PolitiFact finds that Chris Coons' assertion that it has not doubled is True (despite Bryan’s claims that wasn’t exactly what he said).

But Bryan's words are poisoned with the essence of snark and pinch of paranoia (too many pinches, which almost spoils his brew) towards PolitiFact. “PolitiFact has selection bias on hand to guide us through this difficult process…” “There's nothing wrong with a little garbling, is there?” “Precisely why was the comparison kept to August in the first place? Other than to explicitly benefit Coons, of course.” “Whether by accident or not, PolitiFact cherry-picked to Coons' benefit.”

When the unemployment rate in New Castle County began to seriously climb in May of 2008, it served to lessen the two year increase percentage between 2010 and 2008 up to the point where writer Jacobson cited the statistic. As this spreadsheet of unemployment since 2005 shows, the real increases began year over year in May of 2008, with a culmination of doubling (102.7%) in April of 2009. Thereupon it began to decrease year over year since the high unemployment was continuing, which makes the increases appear less because both years are high.

Click for closer viewing.

The percentages circled in green in the 2009 column represent the year over year numbers of unemployment “doubling” and the percentages circled in red is the column of two-year comparisons that equal Bryan’s table. A few other items stand out in this table, however. Notice that in 2006 and 2007, unemployment dropped under Coons, almost 15 percent total from 2005 to 2007. The numbers aren’t all bad. One could assume similar numbers throughout the country as the economy deteriorated during the financial crisis. I’m not defending Coons, but I think if O’Donnell was senator two years ago, these numbers would be the same.

But what measure is the right one? When O’Donnell said the LAST two years, wouldn’t that mean the two years up until now (for which we have statistics)? If PF used Coons’ one year mistaken interpretation as LAST year (August 2010 to August 2009), it would favor him even more, because the increase is only 2.4%; a better measure would be 2009 to 2008, or 2008 to 2007, when it was just over 50 percent both years. I favor using the annual percentage: even if you held the rate the same through the remainder of 2010 (or even decrease it slightly), it would still be an average increase of close to 100% over two years. But isn’t this cherry-picking too?

And flying monkeys to the rescue with additional commentary by “JD” (pray tell who could that be????) with even more pronounced paranoia that PolitiFact is certainly “beating up” on O’Donnell and deliberately avoiding giving any Republicans too many Pants on Fire by *reversing* the ruling to a True by the Democrat. But in the Politi-Score, either way makes “the spread” in averages the same, so it makes no difference.

Bryan also mentions that the August rate is “preliminary” and not “seasonally adjusted” and doesn’t take into account “real” unemployment. This all reminds me of an argument I’ve seen him use: “[Karen] employs a kind of MXC/wall buggers argument: Have Japanese people covered in velcro rope-swing at a velcro-covered wall and hope one of them sticks.”

Bryan way over-played his hand, and didn’t have to bother. While I can understand how PolitiFact determined the calculation, he did have a credible argument and I think PolitiFact needs to take another look at this, as they did with the Jim DeMint ruling noted in the table.

Title and Link	In debate, Chris Coons says Christine O'Donnell is wrong on county unemployment rate
Who? Affiliation	Chris Coons (Democrat)
Ruling	True
Checkers	Louis Jacobson, Writer/Researcher; Martha Hamilton, Editor
# of words	752
#Sources Cited	Five (5)
Argument Summary	In a debate between *Chrises* Coons (D) and O’Donnell (R) who are running for Senator from Delaware, O’Donnell stated the unemployment rate has doubled in Delaware’s New Castle County “in the last 2 years.” Coon disagreed. PolitiFact compared the most recent month available, August,2010, to the rate of unemployment two years previously (August, 2008), and found it had not doubled (had increased from 5.5% to 8.7%, or 58%) and so ruled Coons claiming it had not doubled as “True.”
Bryan’s Argument	Coons did not really say it wasn’t true, just questioned where O’Donnell got her facts, and also erred to say it hadn’t doubled in one year, not two as O’Donnell alleged. But more important, PF used *selection bias* in comparing the most recent month available because comparing January through May of 2010 to the same months in 2008, the unemployment rate had more than doubled.
Quick Interpretation	This reminds me of another ruling on Jim DeMint…another chaos of numbers in a bedlam of bills.
My criticism:	So, where does PF *select* then?
Guidelines	Both, although PF should have done a little more research and gone with an annual rate.
Rhetorical Devices/Logic fallacies	None
DOES IT CHANGE THE RULING?	Yes, to False or Barely True.
My view	Now Bryan is going on a little “mission creep.”
Comments	O’Donnell made things worse when a spokesperson for her campaign “clarified” by saying “O'Donnell was actually referring to the rise in unemployment starting at the beginning of Coons' tenure as county executive” which was January 2005--because the unemployment rate from January, 2005 until June of 2008—three and a half years—held steady at around 4.0%, a pretty admirable record for Coons.
Consequences of Bryan’s interpretation (What I have to believe if he is correct)	Christine O’Donnell stated the unemployment rate has doubled in the last two years. If one compares the first eight months of 2010 against 2008, it is clear that the rate of increase slightly exceeds double.
Bryan’s Grade	Two Pepés: He wants to cherry pick too; and his snark permeated the write up.
PF Grade	Five Pepés. Jacobson should have bothered to look back on the unemployment stats even though the most recent *three months* available makes the most sense in terms of “last.”