AR-Sen: Rasmussen, Arkansas Voters, & Cognitive Dissonance

By now, I assume most people have seen the most recent Rasmussen poll showing John Boozman with a sizable lead in hypothetical matchups against both Blanche Lincoln and Bill Halter.  On the surface, the numbers were pretty damning.

A new Rasmussen Reports telephone survey of Likely Voters in Arkansas, taken Wednesday night, shows Boozman, the winner of Tuesday’s state GOP Primary, with 66% support in a match-up with Senator Blanche Lincoln. The Democratic incumbent picks up just 28% of the vote. Two percent (2%) prefer some other candidate in the race, and four percent (4%) are undecided.
If Lieutenant Governor Bill Halter is his opponent, Boozman, currently the state’s only GOP congressman, earns 60% of the vote to the Democrat’s 33%. Four percent (4%) like someone else in the race, and three percent (3%) remain undecided.

It’s numbers like this — Rasmussen has been consistent in showing both Halter and Lincoln losing to Boozman by 25-30 points — that have many people, both normal and pundit alike, saying that the June 8 runoff is a mere prelude to a convincing GOP win in November.  However, what stand out to me more than the results of this poll, especially given Rasmussen’s tendency to lean right in its results to most polls, are the cognitive dissonance and the intentional value-shifting (based on question framing) visible in the toplines.
Regarding the cognitive dissonance, keep the Senatorial-race results in the back of your mind as you read the responses to questions 8 through 10.

8* How confident are you that Congress knows what it’s doing when it comes to addressing the country’s current economic problems?
2% Very confident
17% Somewhat confident
28% Not very confident
49% Not at all confident
4% Not sure
9* How confident are you that your representatives in Congress are actually representing your best interests?
4% Very confident
17% Somewhat confident
35% Not very confident
41% Not at all confident
3% Not sure
10* Would a group of people selected at random from the phone book do a better job addressing the nation’s problems than the current Congress?
48% Yes
34% No
18% Not sure

That’s right, 77% of respondents were either “not very confident” or “not confident at all” that Congress knew how to fix the economy, 76% were “not very confident” or “not confident at all” that Congress was working in their best interests, and 48% thought that a group of people selected completely at random from “the phone book” would do a better job than Congress is doing.  Yet, in the very same poll, people picked a sitting 4-term Republican congressman — a congressman who voted for TARP and the ARRA; has consistently scored a failing grade from the NEA, the NAACP, and environmental and pro-choice groups; and who voted against creating an office of congressional ethics — 60 to 33 over Bill Halter.
There are two main possibilities behind the discrepancy.¹ Either voters are horribly disillusioned with Congress, but nevertheless think that Boozman is the bee’s knees and that Halter is a real wet blanket, or we are seeing a kind of macro-level value-shifting.
[Because I haven’t seen any Arkansas-centric bloggers discussing the phenomenon, a brief aside on the definition of “value-shifting” is probably necessary.  Technically, it is where voters move between two contradictory opinions that they hold about different issues depending on how a question is presented in a poll.  A great example of this is how support for “off-shore drilling” and support for “energy exploration” vary widely, even when both are referring to the same activity.  The former version triggers a response colored by the negative connotations of “drilling,” often involving environmental or ecological issues; the latter triggers a response that appeals to the positive image of “exploration” as brave and noble as well as the idea that we need to look for new energy sources.  So, when we talk about value-shifting, it is important to note that the contradictory worldviews concern different issues; the question posed, albeit in two different ways or two different contexts, will ostensibly concern the same issue.]
Think of value-shifting this way: the framing of the question, either in its wording or in the context in which it is asked, triggers a reaction by activating an opinion or set of opinion that the listener equates with a specific issue; asking the question in a different way or a different context triggers a wholly different set of opinions because the listener perceives the question as asking about a different issue.
How does this apply to the above Rasmussen poll?  Well, in a poll that purports to be about hypothetical Senate matchups, potential voters are initially asked about their opinion on President Barack Obama.  For whatever reason — and the silliness of this shared opinion is another post for another day — Arkansans are strongly anti-Obama, and they have been since the election.  Starting off with this question triggers a certain set of opinions in the listener, many of which are understandably going to relate to a general feeling about Democrats, regardless of how those people might perceive Lincoln or Halter in a political vacuum.  On top of this, and to reinforce these particular attitudes, when the hypothetical matchups are polled, John Boozman is apparently identified as the Republican candidate, followed by Halter or Lincoln, who are both identified as Democrats.  On top of this deliberate (?) defining of the context, you have the real-life context of Boozman’s somewhat surprising May 18 primary win without a runoff, which only serves to make Boozman look nearly invincible at the moment (even though it could easily have just as much to do with the lack of strong candidates he faced in the primary).
After the hypothetical matchups and a more general question about favorability of the three candidates, respondents were next asked:

5* A proposal has been made to repeal the health care bill and stop it from going into effect. Do you strongly favor, somewhat favor, somewhat oppose or strongly oppose a proposal to repeal the health care bill?

Again, the framing creates a specific bias.  By inserting the first sentence, the question reinforces in the respondent any lingering doubts he or she may have regarding the health care reform bill.  Now, it is not just Joe Voter at home thinking that HCR might not be a great idea; it’s Joe Voter being told that there are other people out there — people with the power to propose an actual repeal — who agree with him.  There is an additional bias created in the wording, “stop it from going into effect,” as well.  By phrasing the question in the negative — there are people who want to repeal this and keep it from happening — the question suggests that “going into effect” is not a good thing.  Between these biases, of course you are going to see overwhelming support for a repeal.
You also create a very strong anti-people-who-made-HCR-happen sentiment and trigger the opinions tied to that stance, which (if you want to create an anti-Congress result in the poll) is a great time to lead in to all of your questions about opinions on Congress.  Then, once they’ve reinforced the anti-Congress sentiment through two more questions, Rasmussen brings the hammer by asking whether some random people selected from the phone book would be better than Congress as currently composed.
All of which is a long-winded way of saying that the results this Rasmussen poll got were, to a large extent, the results they wanted to get, both in terms of Boozman v. Democrat and in the opinions regarding HCR and Congress generally.  Through framing the verbiage of the questions and placing those questions in a specifically drawn context, Rasmussen triggers a certain worldview held by many of the respondents that will shut off any conflicting worldview that might give answers more favorable to the Democrats.
So what, if anything, can we extrapolate from the poll?  If we compare it with the previous polls, the one interesting piece of data to me is the trendline for Halter in these faux elections.  While Lincoln has been in a visible freefall over the past months in her support in hypothetical matchups with Boozman, falling from 35% to 28% since February, Halter has remained very consistent across the board, within 1 or 2 points of 33% every time.  If you discount Boozman’s high number now both as a function of momentum from his May 18 win as well as a function of biased polling, Halter’s consistency suggests that he is almost certainly the stronger candidate against Boozman, as Boozman’s high score in this poll has not effected Halter’s own performance.
A big part of the conventional wisdom behind thinking that either Democrat will definitely lose in November is a theory that all of the beating-up Lincoln and Halter have done to one another since early March has only made Boozman’s path to victory that much easier.  I admit that this conventional wisdom may still turn out to be right, but I would caution anyone from drawing such conclusions from this poll.
¹ Granted, those are not the only possibilities. Maybe the people called equate “Congress” with “Democratic Congress” now that the Democrats are in power. Maybe the people don’t extend anti-incumbency and anti-congressional attitudes toward Representative Boozman when he is running for Senate. Etc. However, I think the two possibilities I listed above are the most likely answers.