Voting early, voting often
By NICK PANAGAKIS
The questions that pollsters ask to find out how people would vote have changed little in 50 years. The last major development was in 1940 when the phrase "if the election were held today" was added because some voters were reluctant to state their intentions weeks or months before actually entering a voting booth. At about the same time, pollsters began to report results on how the race looked at a point in time and not as a prediction of how people would vote in the future.
The objective was to simulate behavior, to approximate the decision process on Election Day. Because voters can't choose "undecided" in the voting booth, respondents were not specifically given the option of being undecided. A choice between candidates was "forced." Undecided was an answer which had to be volunteered by respondents.
In 1981 we developed a question that allows respondents to cast multiple "votes" for candidates. We continue to ask the standard voting question. Then afterwards poll respondents are asked to allocate 10 votes among the field of candidates.
How respondents divide their votes measures the internal dynamics of the complicated decision of choosing a candidate. Respondents who give all or most of the 10 votes to one candidate are counted as a voter for that candidate. If a voter's feelings are mixed, he says so by spreading his votes among candidates. How many of the 10 votes are given a particular candidate determines whether a voter strongly or softly supports that candidate. Importantly, respondents can give an equal number of votes to two candidates. Thus, the answers to this question measure the level of undecided voters more precisely than the standard voting question.
Providing an accurate measure of undecided voters is especially important in incumbent races because most undecided voters end up voting for the challenger. (See "Pulse" column, Illinois Issues. April 1989). We have seen cases of incumbents losing an election where the standard voting question in our own and other polls showed them ahead or approaching the threshold level of 50 percent, while the 10-vote allocation showed less support. Examples include Illinois U.S. Sen. Charles H. Percy in 1984 and Wisconsin Gov. Tony Earl in 1986.
The objective was to simulate behavior, to approximate the decision process on Election Day. Because voters can't choose 'undecided' in the voting booth . . .
In primary races with multiple candidates the 10-vote allocation question also identifies which candidates most undecided voters will choose between. In the 1983 Chicago mayoral Democratic primary, most undecided voters were deciding between incumbent Mayor Jane Byrne and successful challenger Harold Washington, which meant that the third candidate. Richard M. Daley, would have difficulty catching up by Election Day. This year in our early March Chicago Tribune poll on the Democratic primary for Cook County board president, Richard Phelan and R. Eugene Pincham were in a statistical tie,
30/May 1990/Illinois Issues
as they were in most other polls. But Phelan was far more likely to share undecided votes with one of two other candidates. Phelan ended up winning.
Ten-vote allocation also allows cross-tabulation of all other poll data by strong and soft voters for each candidate and by undecideds. This identifies compelling perceptions and issues based on the differences between strong and soft voters for a candidate. Soft voters can reveal doubts about their choice based on where they differ from strong voters. Together, soft and undecided voters are the persuadable voters, those who will determine the outcome.
How the 10 votes are spread also provides a measure of intensity. The ratio of strong to soft voters shows how committed voters are to their candidate. This ratio can be good news or bad news, depending on who the candidate is. The norm is 2 to 1; anything more or less is meaningful depending on the situation.
An early poll showing an incumbent out in front but with a low ratio of strong to soft voters means trouble ahead. This was the case, as later polls would show, when Missouri's U.S. Sen. John Danforth nearly lost reelection in 1982. This was also true for Wisconsin Gov. Earl in 1986. For comfort an incumbent should either have a high strong voter ratio or an increasing ratio as his campaign gets underway. On the other hand, a trailing challenger who is gaining on the incumbent should expect a low ratio as he adds new voters to his base of support.
In the 1990 Cook County board president primary, where there was no incumbent, Pincham had a higher strong voter ratio than his opponents but came in second. To us this meant that voters for his opponents were not ready to commit themselves, but they did after Phelan turned it into a two-man race in the final week.
Allowing respondents to vote often in early polls has clarified both the level and structure of candidate preference in a way that helps get more meaningful results.
Nick Panagakis is president of Market Shares Corporation, a marketing and public opinion research firm headquartered in Mount Prospect. A member of the National Council on Public Polls, Panagakis is best known for preelection and exit polls conducted for the news media in Illinois, Missouri and Wisconsin.
May 1990/Illinois Issues/31