Coffman Over Bennet in Colorado Q-Poll, But are the Numbers Reliable?

April 17, 2015 — Quinnipiac University released a new Colorado statewide poll midweek (March 29-April 7; 894 Colorado registered voters) that surprisingly projects Rep. Mike Coffman (R-CO-6) with a three-point lead over Sen. Michael Bennet (D), 43-40 percent, in a hypothetical 2016 US Senate contest.

Rep. Coffman has survived two difficult re-election battles since a court-drawn redistricting plan left him with largely a Democratic suburban Denver district. Though he has won against significant odds in both 2012 and 2014, he failed to reach 50 percent in the presidential year election.

Coffman ran tough campaigns both times, and spent a combined $8.4 million in securing his last two House terms. Originally winning a safe Republican seat in 2008, he was easily re-elected two years later (66-31 percent). Redistricting radically changed the 6th District after the 2010 census gave the seat 42 percent new territory and transformed it from a Republican district to one that supported President Obama with 54 and 52 percent of the vote in 2008 and 2012, respectively.

Based upon his record as a campaigner and prodigious fundraiser, the congressman appears to be the top choice of the National Republican Senatorial Committee leadership to challenge Sen. Bennet next year. Coffman, however, is not providing much indication that he is eager to run statewide, but polls such as this might provide greater encouragement. The congressman’s wife, Cynthia Coffman, was elected state attorney general last year, and she is also mentioned as a potential senatorial candidate. But, she is evidently less inclined than her husband to make the race.

But, this poll is not without its flaws. First, while Bennet trails Coffman on the ballot test, the senator’s personal favorability rating is a relatively strong 40:27 percent with a good job approval score of 46:29 percent. Therefore, for an incumbent with such favorable ratings to be trailing an opponent who is less known is curious.

The Quinnipiac data becomes highly inconsistent when examining Independents. According to the crosstabs, Coffman leads the senator 44-32 percent among the polling respondents who self-identify as Independent. Yet his job approval score within this same group is 43:28 percent. Therefore, it is difficult to believe that a poll suggesting a certain ideological segment who judges an officeholder to be doing a good job would turn against him in almost the exact opposite proportion.

Another flaw is detected in polling sample selection, both in terms of ideology and the racial demographic.

The current Colorado voter registration breaks 33 percent Republican, 31 percent Democrat, and 35 percent Independent. This is typical of the state’s history, as the party segmentation here has divided almost evenly by thirds for more than two decades. But, the Q-Poll skews too heavily toward Independents, particularly those who fall into the minor party, or “others” category. The polling sample understates Republicans by seven points and Democrats by three, while over-sampling Independents by two percentage points and “others” by almost nine full points. The over-sampling, plus the inconsistency detected among Independents in their portrayal and support factor of Sen. Bennet could help explain the Coffman advantage.

The poll’s racial component badly understates Hispanics. According to the latest US Census Bureau numbers, Colorado has a white only population of 69 percent, yet the polling respondent pool reaches 80 percent. African Americans comprise both four percent of the actual population and four percent of the polling sample. But, among Hispanics, the numbers are also faulty. In the poll, 10 percent of the respondents are Hispanic, way below their actual population share that approaches 21 percent.

The Q-poll is weighted to a 3.3 percent error factor, but the response and demographic inconsistencies suggest that the accuracy factor may be of lesser reliability.

Leave a Reply

Your email address will not be published.