A November 3 poll for Colorado released by YouGov showed the Colorado presidential race to be in favor of Obama 48-47. So what's to kvetch about?
Well, some folks were dismayed because this apparently represents a 2% improvement for Romney since YouGov's October 11 poll. On the other side, as in the case of Pennsylvania (see yesterday's diary) this kept the fuse sputtering under Karl Rove's ample bum, filling the CO airwaves with their foulest emanations, and the R&R boys scurrying around Colorado like Energizer rats. Once again, good. Waste of Republican energy, $$ and focus. For, while marginally better than the Susquehanna garbage lowlighted yesterday, the YouGov polls are also trash.
As I wrote then, in order to examine a poll, we try to get behind the headlines to look at the underlying questionnaire and, more importantly, the demographics of the respondents. These are not always easy to get to (see Note 1 below), but, in this case they were, to YouGov's credit (links provided earlier in the text). We then check the demographics against two gold standards: (a) the 2008 CNN Exit Polls (b) the 2010 U.S. Census Bureau data. If there are significant discrepancies between the poll demographics and the standards, we use various methods to adjust and "normalize" the results for a more accurate projection.
The YouGov Polls
The three categories in these polls which varied most egregiously from the corresponding categories for Colorado in the CNN 2008 Presidential Election Exit Poll and the 2010 US Census are shown below, with the biggest offenders highlighted.
(Data in cells are percentages. Row total in category may not equal 100 because of rounding)
Source: YouGov Colorado Polls, CNN 2008 President Colorado Exit Poll, US Census Colorado 2010
Analysis
Two problems just jump out at you from the chart.
(1) The "thumb on the scale" problem is most pronounced in the October poll - the typical Republican-favoring skew: fewer young, more oldies and of course, whiter and less brown.
(2) The "apples vs. oranges" problem: they apparently tried to compensate for the October farce by gathering a more representative sample for November. But now the composition of the two samples is so wildly different that no comparisons can legitimately be made. Thus, the claim of a two percent improvement for Romney is pure imagination..
The October poll has so many other problems and is out-dated anyway, it can be tossed. As for the November 3 poll, we used the same techniques described yestreday for "unskewing" or unbiasing the sample to resemble more normal voting demographics -- we used the 2008 CNN numbers for our model -- and recalculating. The results could not be more heartening. GO COLORADO! But don't rest on your bums -- Karl Rove isn't on his. GOTV.
YouGov Nov. Poll: Obama 48.20 Romney 46.60
Remove Age Bias: Obama -0.10 Romney -0.26
Remove Race Bias: Obama +1.00 Romney -0.70
2008 Vote Bias : Obama +0.75 Romney -0.75
which all adds up to:
Barack Obama: 50% Mitt Romney: 45%
which actually represents a 2% improvement for Obama and not the other way round, compared to YouGov's own October polls.
This 5 point lead for Obama is virtually unassailable even in the unlikely event that the entire 5% "Other/Undecided" broke the other way.
The likeliest outcome: Obama: 52% Romney: 47% in Colorado 2012
Conclusion
It is not my intention to keep skewering pollsters who skew. But when they clearly screw up (as in YouGov's case) or screw with the public (as with Trib/Susquehanna yesterday), we'll keep nailing 'em.
Note 1: Most reputable polling organizations will provide the backup questionnaires, demographics and cross-tabs for published polls. Most University affiliates (Suffolk, Marist, Franklin & Marshall etc.) are especially good in this regard, the commercial pollsters less so. Most are signatories to the American Association for Public Opinion Research Transparency Initiative, but significantly
NOT YouGov, nor, as mentioned yesterday,
Susquehanna and Rasmussen (or its parent: Pulse Opinion), notorious for its Republican tilt. Be warned.