This is a case study of how inadequate polling can lead Democrats to accept and promote a radical Republican view of reality. This paper compares two polls, one excellent and revealing, the other inadequate, misleading, and counterproductive. The issues raised are framing and value-shifting (where voters shift, depending on the wording of questions, between two contradictory political world-views they really hold, but about different issues). It also discusses how polls can reveal the difference between what words are commonly assumed to mean, versus what they really mean to voters -- and how polls can test this.

It is a truism that poll results can depend on framing. For example, the NY Times reported last month on a NYT/CBS Don't-Ask-Don't-Tell poll on whether "homosexuals" or "gay men and lesbians" should be allowed to serve openly in the military. Seventy-nine percent of Democrats said they support permitting gay men and lesbians to serve openly. Fewer Democrats however, just 43 percent, said they were in favor of allowing homosexuals to serve openly. That's a 36 percent framing shift on the same literal issue, but not surprising since the words evoked very different frames, one about sex and the other about rights.  Newsworthy for the NY Times, but hardly earthshaking.

But a recent poll by David Binder, perhaps the premier California pollster, showed a framing shift of deep import for Democrats -- a shift of 69 percent on the same issue, depending on the framing. It was noteworthy not just because of the size of the framing shift on the main question, but because the shift was systematic. Roughly, around 18 percent of voters showed that their values are not fixed. They think BOTH like liberals and conservatives -- depending on how they understand the issue. With a liberal value-framing, they give liberal answers; with a conservative value-framing, they give conservative answers. What is most striking is that conservatively framed poll questions are all too often written by Democrats thinking they are neutral. The result is a Democratic move to the right for what are thought to be "pragmatic" reasons, but which are actually self-defeating.

Here is the background.

California is the only state with a legislature run by minority rule. Because it takes a 2/3 vote of both houses to either pass a budget or raise revenue via taxation, 33.4 percent of either house can block the entire legislative process until it gets what it wants. At present 63 percent of both houses are Democrats and 37 percent are far-right Republicans who have taken the Grover Norquist pledge not to raise revenue and to shrink government till it can be drowned in a bathtub. They run the legislature by saying no. This has led to gridlock, huge deficits from lack of revenue, and cuts so massive as to threaten the viability of the state.

Unfortunately, most Californians are unaware of the cause of the crisis, blaming "the legislature," when the cause is only 37 percent of  "the legislature," the 37 percent that runs the legislature under minority rule.

I realized last year that the budget crisis was really a democracy crisis, and that a ballot initiative that could be passed by only a majority could eliminate the 2/3 rules, replacing minority rule by majority rule. The idea was to bring democracy to California. Only two words are needed to be changed in the state Constitution, with "two-thirds" becoming "a majority" in two paragraphs, one on the budget and the other on revenue. The changes could be described in a 14-word, single-sentence initiative that went to the heart of the matter -- democracy. It is called The California Democracy Act:

All legislative actions on revenue and budget must be determined by a majority vote.

One would think voters would like the idea of democracy -- and a ballot initiative they could actually understand. And they do. David Binder of DBR Research recently conducted a poll showing that likely voters support it by a 73-to-22 percent margin -- a difference of 51 percent!

There were 800 randomly selected likely voters, with a ±3.5 percent margin of error -- and 53 questions. In short, it was a thorough and responsible poll.

In California, the Attorney General gets to write the "title and summary" -- the description of the initiative that actually appears on the ballot. At present, the Attorney General is Jerry Brown, who is running for Governor. He had announced that he was against getting rid of the 2/3 rule for taxes, though in favor of a majority for budget alone. The result would make Democrats responsible for the budget, but with no extra money to put in it, they would be presiding over the further decline of the state.

When the Democracy Act came across Brown's desk, he personally penned the following title and summary:

Changes the legislative vote requirement necessary to pass the budget, and to raise taxes from two-thirds to a simple majority. Unknown fiscal impact from lowering the legislative vote requirement for spending and tax increases.  In some cases, the content of the annual state budget could change and / or state tax revenues could increase.  Fiscal impact would depend on the composition and actions of future legislatures.

Instead of the original initiative text, Brown's wording would appear on the ballot if it qualified, and would have to appear on all petitions. This wording uses the word "taxes" three times paired with the verbs "raise" and "increase," as well as the conservative phrase for vilifying liberals "spending and tax increases." 

When DBR Research polled voters on both the original initiative text and the Brown title and summary, the results came out as follows:

                                   Support            Oppose      Difference

Original initiative text          73%                  22%               +51%

Brown title and summary              38%                  56%       -18% 

The Brown wording shifted the result by 69 percent! The largest shift Binder had ever seen.

But this was not mere wording. I had expected a large shift, but the neural theory behind my cognitive linguistics research had made a deeper prediction: Many voters have both conservative and liberal value-systems in their brain circuitry, linking each value-system to different issues. Each value-system, when activated, shuts down the other, and each can be activated by language. The prediction was that this shift was systematic, tied to value-based ideas -- not just a matter of one wording or another.

A second prediction was made from long experience. After a strong attack from the right, a liberal poll advantage on an initiative can be expected to drop by around 10 percent.

Brilliantly, the DBR poll tested both for the systematic effect and simulated the effect of a right wing attack. The systematic effect was tested by a battery of pro-arguments followed by a battery of con-arguments, each in distinct wording. The pro-arguments were given first, followed by the battery of con-arguments. Right after the con arguments, the original wording and the attorney general's title and summary were tested again.

                                   Support            Oppose      Difference

Original initiative text          62 %                  34 %             +28 %

Brown title and summary              43 %                  52 %       -9 %

                                                       37 % shift

As predicted, in the face of con-arguments, the 73 - 33 percent advantage for the original initiative dropped to a 62 - 34 percent advantage, a loss of 11 points, but still a 28-point advantage. The attorney general's wording also suffered a loss after the pro-arguments, going from 38-to-56 percent before the arguments to 43-to-52 percent after the arguments, a 9 percent drop for the attorney general's language, about as expected. The total shift after the arguments, from +28 to -9 is 37 percent.

The current explanation of the shift is as follows. There are two political value-systems that voters have, call them Pro and Con. (You might think them as Progressive and Conservative, though no overall views are tested in the poll.) About 40-to-45 percent have a consistently Pro-worldview. About 35-to-40 percent have a consistently Con worldview.  About 18 percent have BOTH worldviews, and the understanding provided by language can trigger one or the other, resulting in a shift.

Now things get really interesting. The DBR poll found a way to test this explanation.  The respondents to the poll were asked if they found the pro- and con-arguments convincing or unconvincing. On the battery of pro-arguments, an average of 57 percent found the pro-arguments convincing and 38 percent found them unconvincing.

On the battery of con-arguments, 57 percent found the con-arguments convincing and 41 percent found them unconvincing. The same high percentage -- 57% on average -- who were convinced by the pro-arguments were also convinced by the con arguments! As in the shift found in the support for the initiatives, the wording resulted in a shift of about the same magnitude. On the pro- and con-arguments,  it was 35 percent -- well within the ±3.5%  margin of error.

                             Convincing            Unconvincing      Difference

Pro-initiative arguments          57 %                      38 %                    +19 %

Con-initiative arguments                57 %                      41 %            -16 %

                                                             35 % shift

This result fits the explanation given above: About 40-to-45 percent are consistently Pro and about 35-to-40 percent consistently Con, with about 18 percent having both Pro and Con worldviews -- and shifting, depending on how language leads them to understand the issue. A large majority of voters stay the same, but a value-shift of about 18 percent of the voters makes for a huge "public opinion discrepancy" of around 36 percent.

What is public opinion on the initiative? It depends on what the initiative is taken as saying. Is it about democracy and majority rule or is it about raising taxes? Overall, public opinion is very favorable on one understanding and very unfavorable on the other. 

Is there a fact of the matter? Is one understanding more true than the other?

At this point, the DBR Research poll gets even more interesting. When a voter hears "raise taxes", he or she usually understands the phrase as meaning "raise my taxes." In short, there appears to be a difference between what the words say and what the voter taking the poll understands. Technically, plugging a tax loophole previously given to certain corporations can be seen as "raising taxes" since those corporations would now be paying their fair share instead of a previously reduced amount. Charging oil companies for the oil they take out of the ground in California is called an "oil severance tax." But such actions would not be "raising taxes" on any individual.

This raises the question of whether the attorney general's title and summary was misleading. When it said "raise taxes", were most voters misled into thinking it meant raising their taxes?

The DBR poll found a way to test this. It asked the following question:

Some experts on the state budget say that enough money to solve the budget crisis can be raised without raising taxes on those in the lower or middle income brackets.  Instead, tax loopholes for corporations can be closed and a fee can be assessed to oil companies for extracting their oil from the land.  Do you support or oppose solving the budget crisis by closing tax loopholes on corporations and charging oil companies an extraction fee without raising taxes on lower and middle income Californians?

The response:     Support - 62 %         Oppose - 34%

In short, most Californians, those hurting most in the lower and middle income groups, are not opposed to raising taxes in general. They just think they are already paying fair taxes.  What does this mean for the shifts we have seen toward the Attorney General's title and summary, which says that the initiative is about "raising taxes"? It means that most voters are misled by the language into thinking that the initiative is about raising their taxes.

For this reason, I have resubmitted the California Democracy Act, asking Attorney General Brown for a new title and summary, one that does not mislead the voters.

Do most voters really care about democracy? Hardened Democratic political leaders told me they didn't believe it. They thought voters only cared about their pocketbooks. So DBR Research tested this as well. The poll asked voters if they agreed or disagreed, as follows:

In a democracy, a majority of legislators should be able to pass everyday legislation.

Agree - 71 %      Disagree  - 24 %

In a democracy, a minority of legislators should be able to block everyday legislation.

Agree - 25 %      Disagree  - 68 %

In short, voters do care overwhelmingly about democracy.

The DBR Research poll is remarkable, and brilliant in many ways. But to see its true significance, one should compare it to other polls, supposedly on the same issue.

In the spring of 2009, when I first thought of this initiative and started discussing it in public, I was told over and over that polls were taken and that my initiative didn't poll. I heard it first from a state senator, then from a powerful official in the State Democratic Party, then from the political directors of various unions who had spoken with that party official.  They were against my initiative on the grounds that it couldn't win, supposedly because it didn't poll. Perhaps the most influential of these polls was one by someone I will call the Other Pollster, taken just after I had submitted the California Democracy Act to the Attorney General. 

(Incidentally, I am not identifying the individuals involved because the issue is not about individuals. As we shall see, the other pollster, the party official, and the political directors were acting normally, all too normally.)

Here is what the Other Pollster, in his summary, referred to as the "direct question."

Would you favor or oppose allowing the state legislature to increase taxes by a majority vote rather than the current two-thirds vote requirement?

Favor - 35 percent    Oppose - 62  percent 

Notice the assumptions built into the question: "allowing the state legislature to raise taxes". Again, the "raise taxes" will be heard as "raise your taxes" and "allow" suggests that the legislature will want, be able to, and will raise your taxes."  

The Other Pollster also asked a slightly different version of this question: 

Regarding taxes and government, would you prefer less government and

lower taxes, or SLIGHTLY HIGHER TAXES FOR BETTER GOVERNMENT SERVICES ?

LESS GOVT/LOWER TAXES. ........ 59 %  BETTER GOVT/HIGHER TAXES. ..... 41 % 

The results are what we would expect. 

The Other Pollster was also asked by the party official to see if the California Democracy Act had any serious support.  

The question the Other Pollster asked on the poll embedded my initiative language into the linguistic frame, "Some people say .........  Do you agree or disagree with this viewpoint?"  It was the only question embedded into this particular linguistic frame. 

Notice that this frame presents a contrast between "some people" and "you," introducing a bias against whatever is in the "..." .   In addition, "some people" indicates a minority opinion, which introduces a second bias.  Third, he referred to it as "this viewpoint," distancing it from the person taking the poll (it is only a "viewpoint") -- a third bias.  

Here is his question and result: 

Some people say that "all state legislative actions on revenue and

budget issues should be determined by a majority vote." Do you agree

or disagree with this viewpoint ?

AGREE. ............ 51       DISAGREE. .......... 43

Even in that triply-biased frame, the original initiative language about majority rule came out ahead by 8 percent, while the language about raising the respondent's taxes came out between 27 and 18 percent behind -- shifts of 35 to 26 percent.  

The Other Pollster noted the shift, but concluded: 

"the question of simply lowering the two-thirds budget approval threshold to a majority vote, without any conditions, was asked two ways: 

•  
  • 35% of voters supported, and 62% opposed, the direct question of "allowing the state legislature to increase taxes by a majority vote, rather than the current two-thirds vote requirement."

•  
  • 51% of voters agreed, and 43% disagreed, with the "Lakoff" question which read: "All state legislative actions on revenues and budget issues should be determined by a majority vote?"

Neither one of these 2 concepts meets the initial 60% voter support threshold needed to withstand the onslaught from a well-funded opposition campaign. 

The difference between the "Lakoff question" and the "direct question" can largely be explained by recognizing that the Lakoff question which read: "all state legislative actions on revenues and budget issues shall be determined by a majority vote" (51% support), did not fully convey the real consequences to voters that the Lakoff language would mean: "allowing the state legislature to increase taxes by a majority vote rather than the current two-thirds vote requirement" (35% support).

On subjects like taxes, it can be dangerous to assume that voters can be moved to vote differently from their true beliefs by using cleverly crafted language." 

First, the Other Pollster does not mention the question he actually asked, using the some-people-say frame. Second, he assumes that the "direct question" is the one that does not mention democracy or majority vote, but rather the one that assumes that "the legislature" wants to,  would be able to, and would increase the respondent's taxes. This is misleading, not "direct," for reasons discussed above. He calls this the "true belief" of the voters. Third, he suggests that asking about democracy and majority rule is "cleverly crafted language" to "move voters to vote differently from their true beliefs."  

If you take the Other Pollster's poll and his description of the results at face value, you might very well think that the California Democracy Act "does not poll" when it, in fact, polls 73 percent on the first pass and 62 percent right after a barrage of right-wing attacks.  

Why does the Other Pollster's poll and poll description look that way, and what does it say about the Democratic leadership that commissioned it and believes the Other Pollster's description of his results.  

What Does All This Mean?

Polls have come to matter, in at least four ways. 

First, the issues matter. The issue here is the future of California and whether a minority of ultra conservatives will continue to bankrupt the state government purposely to keep it from meeting desperate public needs. In short, the issue is as serious as any issue in public life. And the question "Does it poll?" becomes literally a matter of life and death for many people, and of impoverishment and suffering for others. 

Second, what the Other Pollster calls the "direct questions" and "true beliefs" are the radical conservative ideas about taxes that conservatives have put forth misleadingly year after year. Here Democrats have been so whipped for so long that they accept conservative framings as simply "true beliefs." What happens when those Democrats are confronted with a question about simple democracy and majority rule, rather than the minority rule that they and the majority of citizens have been suffering under? They cave. When such Democrats see a statement that they actually believe in and wish would happen, they see it as only "cleverly crafted language." The Democratic leadership in California has come to believe a false Republican view of reality, to own it and promote it, and to help make it real. Through polls. 

Third, it is rare for polls to discuss what DBR called "the 33%-percent discrepancy group," that is, the people who have TWO distinct value systems applied to different ideas (e.g., democracy vs. additional taxes on them), and shift depending on the ideas expressed in the language of the poll. These voters need to be studied, isolated as a culturally important demographic group, and taken into account in future polls. This may involve admitting that there may not be such a thing as overall "fixed public opinion" that includes this significantly large group.  Polls should be detecting public understanding -- and studying voters with dual value-systems is crucial if the value-shifters are to be identified and understood.  

Fourth, the word "taxes" is not neutral or objective.  It has been hijacked by the right. By virtue of their communications system, they have changed the framing of the word to mean, according to radical conservative doctrine, "money that individuals have earned without government help that is taken out of their pockets by the government and given to people who haven't earned it and don't deserve it." For many voters, "taxes" has come to be a word defined by the Con ideological worldview, able to activate that worldview in the approximately 18% of voters who switch, depending on language. The last thing Democrats -- or independents -- should be doing is using language that activates a Con worldview and whose effect is to create a shift to the right. It is unfair. In this case it goes against democratic principles. And politically, it is shooting oneself in the foot. 

It is for this reason that I have chosen the word "revenue." "Revenue" is a neutral word in that it has no such doctrinal meaning. It is a word that comes from business. To run a business, you need revenue; and the same is true of running a government. It is just false to think that the use of the word "taxes" is neutral or objective. In the poll questions cited, that right-wing doctrinal meaning is sneaked in, misleadingly. 

Finally, these results show the effectiveness of the radical conservative communication system operating 24/7 using the same effective framing year after year. It operates on an unconscious level, slowly changing the brains of those engaged (on either side) of the discourse that the conservatives define.  Their communication system is so effective, and Democratic leaders have to deal with it so often, they too can get taken in.  

This poll revealed that, in California on this issue, 18% of the likely voters were value-shifters, that is, they seem to have BOTH worldviews. Given that Democrats have 63% of the seats in the legislature at present, that means that the 18 percent has been voting in the Democratic column, either as Democrats or independents. But if they have BOTH worldviews, that means they are susceptible to conservative arguments in conservative language, and could shift, as happened in the case of Scott Brown's election in Massachusetts. Democrats cannot take value-shifters for granted. They have to identify them and convince them using value-based language of their own.  

The results of this poll goes AGAINST the idea that such voters are "in the middle" and that one can appeal to them by moving to the right. The use of the language of the right can move them to think like conservatives, and hence to vote like conservatives.  

I am a cognitive scientist and a linguist, and have been applying what has been learned in those disciplines to our politics. I have been arguing over the past decade and a half that progressives need to build a communication system of their own to (1) express the values they really believe in, to (2) to communicate the truth, (3) to use their own values-based language to show the moral significance of those truths, and (4) avoid communicating conservative beliefs they do not hold, especially by avoiding the language of conservatism.  The poll results just discussed reflect the failure of progressives to do so.  

Pollsters have an awesome responsibility. I see the DBR Research poll as a model for carrying out that responsibility. And I have chosen to discuss that poll at length because of the general lessons it has to teach.