Politi-Score: Elections 2010 Part One

Data Averages Overall

PolitiFact (PF) recently published its own evaluation of the recent mid-term election campaign, calling it, in Truth-o-Meter terms, a "Barely True."  So, was it?

Well, a bit of time was spent was going through the ratings one by one from September 1 through November 1 (the day before the election) to see if I could validate Bill Adair's calculations.  My numbers came out a bit different than his.  Since he published his evaluation on October 27, I probably counted additional rulings that slightly changed the percentages that he calculated.  I counted 414 rulings in total.  This was a bit different from the way I had initially surveyed, which was by selecting only those individuals who had over five (5) rulings by going through the list as shown under PF's "People" tab.  This time it was a total cross section of two months of all rulings, including organizations, bloggers and chain e-mails.

The result was an agreement with PolitiFact on "Barely True" in total.  But further dissection of the numbers to divide them between Democrats and Republicans revealed who was, as judged by PolitiFact, more than Barely True.

First 2 columns are OVERALL percentages of rulings (mine compared to PF).

There was *barely* a change in making adjustments to my initial selection of rulings in August; in fact, this time the Democrats went down a bit in the Politi-Score truth-o-meter and the Republicans  went up by a fraction (.16), even though the measure above shows that in total, during the campaign, the average was well below the overall score I had calculated at the time (August 22, 2010) which was 80.88 (80 is equivalent of "Half True") based on 950 rulings.  Based on 1,117 aggregated rulings through November 1, it was now 80.38.

The difference between Democrats and Republicans in this new cross-section complete sample, however, is quite stark here, especially when done using what I am calling the "Wide Measure" (as suggested by Facebook commenter Joanne Libby)--which instead of school-grade-like points assigned, uses more intuitive 25-point increments with 100 as True, 50 as Half True, zero for False and -25 for Pants on Fire.  With that measure, the Democrats are over 40 percent "truthier" than Republicans.  So do people like being lied to (since the Republicans won despite being almost "Barely True") or as Bryan White (and Jeff Dyberg) contend, PolitiFact is really biased left?

My next measure (Elections 2010 Part Two!) will be to re-calculate the numbers excluding organizations, bloggers and chain e-mails, since they make up roughly 20 percent of the 414 records (and are definitely more skewed toward *untrue* rulings), to see what effect this has on the Democrats and Republicans' Politi-Score averages.   I'm also planning an analysis by state, since some of the newspapers partnering with PolitiFact may have more mixed political leanings (particularly, PolitiFact Virginia), and I want to see if there's any evidence of this in the rulings.  Stay tuned.