Methodological Issues in Special Election Polling

In a recent article for Public Opinion Pros I wrote about the methodological challenges confronted by California pollsters during the 2003 governor’s recall election. The November special election raises a number of thorny issues for those of us involved in conducting independent public opinion polls on the eight state propositions, especially the definition of “likely voters,” because of the rarity of special elections, the sheer number of ballot measures, and the complexity of the ballot language for each measure.

One clue to voter turnout in November is the percentages of registered voters participating in the four initiative-only special elections (48 percent in 1973, 37 percent in 1979, 36 percent in 1993), which were relatively low in comparison to the governor’s recall in 2003 (61 percent). We are currently monitoring in our monthly polling the level of enthusiasm about the special election itself, attention to news about the special election, and interest in the specific propositions on the ballot. All of the indicators now point to a low voter turnout similar to the 1979 and 1993 special elections. Will it favor the governor or his Democratic opponents?

At this point, it is too early to tell, as partisan interests may be swayed by advertising on the governor’s three initiatives or the other five measures on the ballot. For now, we are using a likely voter screen that we would apply to our preprimary election polling, thus anticipating a smaller electorate that tends to represent older voters, longer-term residents, higher socioeconomic status, and the party faithful.  

How many of the eight initiatives can we ask about in an RDD survey? We tested the eight propositions using paper and pencil ballots in focus groups before our first preelection survey and found that respondents grew weary in the process. We have typically asked about three state propositions in our preelection polls that include candidate races. We are experimenting with asking about five ballot measures for the special election. We believe even this is pushing the envelope, given our interests in exploring the attitudes behind support for ballot choices.

Still, we have decided not to ask about all of the eight measures through split-sample and half-sample techniques. This is because we like to start with a full sample of about 1,000 likely voters, with a goal of having large subgroups to analyze key political, regional, and demographic trends. We will ask about most of the eight initiatives over the course of three preelection surveys, repeating the three governor’s initiatives in each wave, and offering time trends on as many other measures as possible.

How should we ask about the wordy and complex initiatives on the ballot? Our approach is to read the full ballot title and label, including fiscal cost statements, as they will appear to voters in the voting booth or on absentee ballots. Others may choose to ask if voters have heard of the measure, and if they plan to vote yes or no, before reading the ballot information to the subgroup who have not made up their minds. Still others may choose to reword the measure in plainer English and with better grammar for improved clarity. We choose to repeat the exact wording on the ballot for every respondent. We also include follow-up questions on each proposition in order to explore voters’ attitudes toward the basic concept underlying the initiative and their reasons for supporting or opposing the ballot measure.