Imprecise language and Citizens United polling
Americans oppose the Court's recent decision in Citizens United by a margin of nearly two to one (sixty-four to twenty-seven percent), according to a poll conducted by Greenberg Quinlan Rosner Research and McKinnon Media on behalf of Common Cause, Change Congress, and the Public Campaign Action Fund. Another poll, by ABC News and the Washington Post, recently found that eighty percent of Americans oppose the Citizens ruling, with sixty-five percent "strongly" opposing it. Both polls also found that a broad majority of voters favor some type of congressional response to the decision. These findings have been widely reported, with some proponents of reform relying on them to call for a congressional response to the Citizens decision.
However, in my opinion, both surveys used imprecise language to make defective claims. That the surveys may have misinformed their respondents is a cause for concern, because forty-three percent of respondents in the Greenberg Quinlan survey answered that they knew "not very much" or "nothing" about the case, indicating that the survey itself was their primary source of information about the decision. In this post, I examine the accuracy of the surveys' descriptions of the decision in Citizens. I leave it to those with statistical training to examine the survey's methodology and decide what effect, if any, the wording of the questions may have had on the conclusions.
Philosopher of language Paul Grice once proposed various principles to help explain how successful communication is possible. In so doing, Grice described the assumptions that exist between listeners and speakers in a conversation, and as has been demonstrated, Grice's maxims are useful for analyzing whether a survey's language follows our assumptions about conversation. By following these rules, surveys can avoid misleading or imprecise claims.
Essentially, the maxims dictate that in conversation we should be honest, clear, and provide relevant information, so that a listener can assume that she is sufficiently aware of the facts to form an opinion. Grice's maxims instruct a speaker to communicate as clearly as possible and to avoid ambiguous or obscure words. In the context of a survey, this means that the survey administrator should choose the clearest and most relevant phraseology possible; indeed, as multiple studies have demonstrated, survey respondents assume that administrators have done so. And, according to Grice, a speaker must provide all of the information necessary for the conversation; thus, a survey respondent would assume that she was provided with the complete body of information necessary to form an opinion.
But even when survey administrators follow these maxims, seemingly straightforward words can be misunderstood. For example, one survey asked respondents "how many times do you usually switch from one station to the other, when viewing TV on a weekday evening?" Eighty-seven percent of respondents indicated in a follow-up interview that they understood the word "you" to refer to themselves, but thirteen percent understood the word to indicate another meaning: two percent interpreted "you" as "you and your spouse," while two-and-a-half percent interpreted it as "you and your family." This effect of multiple understandings is most pronounced for words with multiple meanings. And if ostensibly everyday words can be misunderstood, then it is crucial that descriptions of technical, complicated issues like campaign finance law be clear and unambiguous to avoid confusion or misinterpretation on the part of survey respondents.
The Greenberg Quinlan poll poses two questions about the Citizens case:
Q. 30 Recently, the Supreme Court ruled on a case brought by the group Citizens United that changed campaign finance laws and will allow corporations, unions and other groups to spend money to directly support or oppose specific candidates. How much have you heard or seen about the Supreme Court’s decision in this case — a great deal, some but not a lot, not very much, or nothing at all?
Q.31 Now let me read you a short description of this case. Before the decision, the law barred corporations and unions from spending money to support or oppose candidates. The Supreme Court overturned this previous law and ruled that corporations and unions have the right to spend money to support or oppose specific candidates. Now, after learning a little bit more about this, do you favor or oppose this decision?
The ABC/Washington Post survey poses this question about the Citizens case:
35. Changing topics, do you support or oppose the recent ruling by the Supreme Court that says corporations and unions can spend as much money as they want to help political candidates win elections? Do you feel that way strongly or somewhat?
A respondent might infer different meanings about what it means to "spend money to support a candidate." "Spend[ing]" could refer either to direct contributions to candidates or to independent election expenditures (for example, buying advertisements in support of a candidate). Similarly, a respondent could interpret the first poll's references to "laws" and "the law" to mean state law, federal laws, or both .
Thus, the first possible reading of the survey questions is that the Court in Citizens United overturned laws prohibiting corporations and unions from supporting or opposing a candidate through direct financial contributions. Yet even before Citizens was decided, corporations were permitted to "“ and, as the Center for Political Accountability has documented, did "“ donate directly to candidates in twenty-eight states. (Information on laws in these states and the twenty-two states that prohibit such donations is available here).
The second possible reading of the survey questions is that Citizens United overturned a law prohibiting corporations and unions from either "support[ing] or oppos[ing]" a candidate (Greenberg poll) or "help[ing] political candidates win elections" (ABC poll) by making independent election-related expenditures. But as a supplemental amicus brief filed by the Chamber of Commerce points out (at 8 n.5), twenty-six states and the District of Columbia already permitted corporations to make independent political expenditures during campaigns, including the purchase of television and radio advertisements that directly encouraged a vote for a specific candidate.
The phrase "the law" and the word "spend," as used in these surveys, do not mean what these words mean in conversational English. To "spend," in a conversational sense, seems to mean "to use or expend money." So "spending" on a campaign or election seems to refer either to making a direct donation to a candidate or to providing indirect financial support, such as by purchasing ads on a candidate's behalf. Many states permitted unions and corporations to "spend" in both these senses before Citizens was decided; that the surveys claim or imply otherwise is inaccurate and misleading.
Further, neither of the surveys mentions important distinctions between federal laws, which previously banned corporate contributions, and state laws, which in many cases have permitted it for years. And in all three of the questions, the broad language seems to affirmatively mislead respondents. Although respondents would assume that the survey used accurate, clear language and provided all of the information needed to form an opinion, the survey did neither.
Although the language of these polls is flawed, it is possible to design an improved poll. Future Citizens United polls ought to distinguish between state and federal laws and eschew mistaken categorical claims. Knowing that respondents will apply conversational definitions to words, the polls' creators should use precise language, clarify what types of corporate and union spending are permitted, and accurately contrast the new scope of campaign laws with previous laws.