A Tale of Two Polls in the Colorado Governor’s Race: Advantage Magellan

Update, 10/25: Casting further doubt on the Denver Post / SurveyUSA poll is today’s release from the Democrat-leaning Public Policy Polling, which finds Hickenlooper up 47-44 over Tancredo, with only 5 percent in the Maes camp. This poll was conducted after both Magellan and SurveyUSA — from Thursday, October 21, to Saturday, October 23. (H/T Curtis Hubbard, The Spot)

On Friday conservatives were treated to the release of poll results from Magellan Strategies showing Tom Tancredo narrowly trailing John Hickenlooper 44-43 in the governor’s race. This morning readers opened up the Denver Post to see its SurveyUSA poll showing Hickenlooper with a 49-39 lead over Tancredo. In both polls Dan Maes is stuck at 9 percent.

The Magellan poll was conducted on Wednesday and has a margin of error plus/minus 3 points (1,067 likely voters). The SurveyUSA poll was conducted between Tuesday and Thursday with a margin of error plus/minus 4 points (621 likely voters). So we’re looking at the same time frame, but the margins of error don’t overlap. The rules of logic dictate that one of the two polls has to be wrong. In my estimation the Magellan results are closer to reality, not because they comport with my own wishes but for the following reasons:

  1. The internals of the Magellan show not only a larger voter sample but also one that reflects a reasonable estimate of partisan voter turnout (38 Republican, 35 Democrat, 27 Unaffiliated); so far I have not been able to find internals on the Denver Post / SurveyUSA poll
  2. SurveyUSA came out with polling the same number of days before the August 10 primary — they were within the margin of error on the Republican gubernatorial primary (3 points off) but were 6 points off on the Republican Senate primary and nearly 12 points off on the Democratic Senate primary
  3. One other recent poll of the Colorado governor’s race (Reuters/Ipsos) closely mirrors the wide advantage exhibited in SurveyUSA, but that was based on an even smaller likely voter sample of 405 and again did not reveal the internals of the poll; two other polls showing a closer race — Rasmussen (Hickenlooper +4) and FoxNews (Hickenlooper +5) had significantly larger samples of 750 and 1,000 likely voters, respectively

(Note: The first two cautions are also applicable to the Senate race polling that places Ken Buck and Michael Bennet in a tie.)

In the final analysis, while the Denver Post / SurveyUSA poll results shouldn’t be demoralizing to Tom Tancredo supporters, they should provide an extra dose of reality. Democrat John Hickenlooper’s lead is more likely 3 or 4 points than either 1 or 10. In other words, the Denver mayor is still the favorite in this race, but his victory is not assured.


Leave a Reply