Face The State Staff Report
Voters be warned: while campaigns heavily promote polling data to suggest public support, political analysts allege that such data is highly subject to manipulation by pollsters, spinsters, and campaign consultants.
Over the last few months, polling on Colorado congressional races, state legislative contests, and ballot initiative campaigns has become intense. With the state now identified by both presidential campaigns as one of a handful of toss ups that could determine control of the White House, voters have become even more likely to be surveyed for their opinions.
An Oct. 6 Mason-Dixon poll funded by The Denver Post raised eyebrows among political insiders because its conclusions sharply contradicted those seen in other polls. Of the five ballot questions surveyed by Mason-Dixon, two saw dramatically different results from previous polls. Amendment 58 would eliminate a tax credit for energy companies and is backed by Gov. Bill Ritter. Amendment 46 would prohibit race and gender preferences in public education, hiring and contracting. It is modeled after a similar measure approved by California voters in 1996.
The Mason-Dixon poll conducted from Sept. 29-Oct. 1 found that just 28 percent said they would vote yes on Amendment 58, while 41 percent said they would vote no and 31 percent said they remained undecided. A poll sponsored by the Economic Development Council of Colorado and conducted by Denver-based Ciruli Associates from Sept. 19-23 demonstrated double-digit differences, with 51 percent of registered voters indicating a yes vote, 25 percent saying no, and 19 percent remaining undecided.
Mason-Dixon surveyed 625 likely voters and weighted in an additional 200 undecideds while Ciruli surveyed 501 registered voters. Mason-Dixon had a margin of error of plus or minus four percentage points, whereas Ciruli had a margin of error of 4.4 percentage points.
Multiple national polls on Amendment 46 show similar disparities in results. In June, a poll conducted jointly by Quinnipiac University, the Wall Street Journal, and WashingtonPost.com, found that 66 percent of likely voters would vote to approve the proposal, while just 15 percent said they would vote no, and 18 percent said they were undecided. In August, however, a Rasmussen poll found that 55 percent of voters said they would approve the amendment, 23 percent said they would vote against it and 22 percent said they are in the undecided camp. The most recent Mason-Dixon/Denver Post poll, however, shows that support for Amendment 46 stands at 42 percent, with 29 percent of respondents saying they will vote no and another 29 percent remaining undecided. Of the five measures polled by the Post, Amendment 46 is still the only one predicted by the paper to prevail.
"The devil is in the details," said Amendment 46 spokeswoman Jessica Peck Corry. "When voters are asked specifically whether they support the actual ballot language, they say yes it at nearly 70 percent. When pollsters add or cut from this language, however, support shrinks and people get confused so we see higher numbers of undecided voters."
The polls differed in how questions were framed to voters. The Quinnipiac/WSJ poll used the actual ballot language, the Rasmussen poll used a summary of the ballot language, and the Mason-Dixon uses part of the ballot language, but adds that "the terms 'discrimination' and 'preferential treatment' are not defined in the measure."
"When you add language like that, it suggests to voters that ambiguity exists where none does," Corry added. "Discrimination is clearly understood in a legal context and bans on preferential treatment have been upheld by our courts."
Pollsters face challenges when it comes to keeping voters interested enough to finish a poll. They say that they must summarize ballot language because it is often bogged down with legalese and political jargon. Trimming down the questions to make them more digestible.
Larry Harris, principal at Mason-Dixon, says his firm tries to stay as close to the ballot text as possible when conducting public opinion surveys.
“Our aim is to make it as reflective of the actual ballot language as possible,” said Harris. “We’d rather read it verbatim, but sometimes it requires not as much and sometimes it requires more.”
But Golden-based pollster Lori Weigel of Public Opinion Strategies says when pollsters truncate ballot language they run the risk of skewing results. The more closely a poll question resembles the actual language of the ballot initiative, she believes, the more closely the poll will predict what will happen on Election Day.
“The fact is the actual ballot language, actually matters,” Weigel said. “That’s why a well-run initiative campaign will spend a lot of time and effort to make sure its ballot language is clear, concise, and compelling.”
Weigel said initiative proposals with shorter language tend to fair better not only with polls, but also at the ballot box.
Although both the Mason-Dixon poll and the Ciruli poll used the official ballot language as a starting point, the eventual wording of the questions ended differently.
When Ciruli asked registered voters how they would vote on 58, the question was posed as: “Amendment 58 is a statutory amendment to increase the state severance tax paid by gas and oil companies $321 million annually by eliminating an existing state tax credit for local property taxes paid by gas and oil companies. The increased tax revenue would go to college scholarships for state residents, wildlife habitat, renewable energy projects, and local transportation and water projects."
When Mason-Dixon polled the measure, it also told voters that the measure would "exempt all oil and gas severence tax revenue from state and local spending limits."