Reuters asked 1432 ONLINE persons who claim they are over 18 and live in America. Of that 1432 there would be 64.9% registered (per the census) or 930. Of that 930 registered voters, But Reuters says that 463 CLAIMED to be registered Republicans! It is no wonder that this poll and prior Reuters online polls looked like they were skewed with too many liberals lying about being Republican. There was another online poll taken almost the same dates, that is also inaccurate, but apparently for another reason. Zogby did separate their online 2064 likely voters into likely GOP voters, vs likely Dem voters, and seems to have done a good job with little crossover. I estimate based on the Obama 55% disapproval of these participants that they found about 1135 likely GOP voters. But of their likely GOP voters, it seems obvious they are polling too many hard-core conservatives, compared to the GOP-leaning population as a whole. Here are those two online polls side by side:
|
It was obvious to me that Reuters 'Republican data' included too many liberals lying to answer the GOP question.
It was also obvious to me that Zogby polled too many hard-core conservatives.
-Certainly Cain must be somewhere between 12 and 26 (probably closer to Reuter's 12), and
-Certainly Mitt must be somewhere between 14 and 22 (probably closer to Reuter's 22), and
-Certainly Newt must be somewhere between 24 and 32 (also likely closer to Reuter's 24).
I thought there should be some way to merge the two online polls to get something that balanced the two, and I also wanted to see how my end result might compare to the more legitimate national polls.
So I merged the Reuters and Zogby results with a patent-pending process of my own design. It took me a lot of fiddling to find a recipe that looked 'ok'. And I finally found one that seemed good enough for government work...
(3 parts Reuter result for every 1 part Zogby)
So here is my 'merged' result I call ReuZog, along with the Reuters and Zogby results:
I didn't want my merged 'poll' to swamp out other results so I give this a weighting of 283 (212 Reuters results plus 71 Zogby), as follows,
Poll | Date | Sample | Gingrich | Romney | Cain | Perry | Paul |
Reuters/Ipsos online | 11/18 - 11/19 | 423 RRV | 24 | 22 | 12 | 10 | est. 8 |
Zogby online | 11/16 - 11/17 | 1135 LVgop | 32 | 14 | 26 | 6 | 8 |
ReuZog (Reuters & Zogby) | 11/16 - 11/19 | 283 Lvgop | 26.0 | 20.0 | 15.5 | 9.0 | 8.0 |
SO THEN, here below, is my 'ReuZog' result along with other national polls, in a 4-poll weighted running averages chart...
I also weighted the polls by their size. For example, a poll with a survey size of 1000 was weighted proportionately heavier than an 324 survey size.
Weighted 4-Poll Rolling Averages of the latest 20+ Polls |
26.0 20.0 15.5
These numbers seem strangely like the results of many polls in the past. IOW, regardless of which candidate polled which number, there seems to have been a strange sort of consistency in the top-tier numbers so far.
Am I right on that?
Has there been a fairly consistent range for the top three?
Seems to me we’re not getting close to coalescing around a candidate until the frontrunner breaks 30% and retains it and the third place candidate drops into maybe single digits. Just a thought.
Also it’s still pretty early into Newt’s surge as well as into Cain’s troubles.
Seems to me it will be a while longer before the polls reflect the settled views of both these candidates.