Monday, July 25, 2011

Piquing PolitiFact: What definition of "Half True" is used at PolitiFact Ohio?

It's going to appear shortly.

Allegedly.

(clipped from cleveland.com)

I should have saved the image from the first time I tried to post the nearly identical message.  It's possible that I did something that kept it from posting, but as far as I can tell I followed the exact same procedure this time and received the message that my reply was received and would appear shortly.

Perhaps the idea is to delay the appearance of my comment until the discrepancy can be fixed.

Good luck with that.  I took note of it quite some time ago and I've been posting about it occasionally on the 'Net for weeks.

A number of comments do a nice job of pressing editor Robert Higgs for needed explanations.  One even brought up Eric Ostermeier's study suggesting selection bias at PolitiFact.  Higgs' reply to that one is hilarious:
I don't know enough about how the Minnesota study was done to talk about it. It's (sic) focus was PolitiFact.com, our national counterpart. Feel free to contact them via email. If you send them questions, you'll likely get a response.

What I can tell you about PolitiFact Ohio is that in our first year we did 81 Truth-O-Meter rulings on statements from Democrats and 103 from Republicans (more on the GOP side principally because Kasich is governor and there's twice as many Republicans in power as Democrats).

The average grade for Democrats: Half True

The average grade for Republicans: Half True
Some say the Texas Rangers have a better team batting average than the Seattle Mariners.

What I can tell you about it is that right now both teams are in the American League.

The Seattle Mariners are batting roughly .250.

The Texas Rangers are batting roughly .250

Higgs isn't saying anything.  The "average" when there are only six positions on the scale is a very rough approximation, just like rounding the Rangers' batting average down from .272 and the Mariners' battering average up from .226.  Is a difference of .46 statistically significant?  You betcha.  It's huge.

The question is, does Higgs buy his own explanation or does he really not know any better?  Either option is a bit scary.


***
The reply has appeared.  Now we'll see if it lasts.


Correction 7/25/11:  Used strikethroughs to improve the sense of the final paragraph.

3 comments:

  1. He isn't, Bryan. I was looking at that PF Ohio piece yesterday and went over to your blog to see if you had something to say about it. In terms of the PF's own Truth Index, Democrats are a positive 18.48 and Republicans are a negative 13.79. Yes right around Half True (zero) but 32 points is "statistically significant."

    ReplyDelete
  2. Hi, Karen.

    Thanks for confirming my suspicion. Do you agree that Higgs' response is a bit disturbing or what?

    ReplyDelete
  3. Bryan, he appears to be downplaying any significance of the Democrat Truth Index average being higher than the Republicans. It's almost like a willful ignorance...because PF is well aware the Dems average higher. Also, incidentally, PF Virginia (the one that endorsed Republicans and has Cantor's wife on its board) is still averaging higher for Republicans for all rulings.

    ReplyDelete

Please remain on topic and keep coarse language to an absolute minimum. Comments in a language other than English will be assumed off topic.