Although the public and media like the "racehorse" aspects of pre-election polling, polls are more useful in analyzing the movement and content of public opinion than in predicting election outcomes. While there were few surveys near election day, there were enough to trace the proposal's fate.
Since 1995, polls have reported that growth and traffic congestion are among the main concerns of metro residents. In a series of surveys conducted during 1996 and early 1997, the RTD sales tax increase for light rail and other transit improvements held support levels above 59 percent.
The proposal's support began to decline in August and dropped precipitously in September and October of 1997. Analysis of the campaign's own surveys, which used a consistent question wording and sample screened to identify likely voters, shows that the proposal began with 57 percent support in July. In August, the campaign reported 55 percent, dropping to 52 percent at the end of September. This decline correlates with an increase in conflict among opponents, the campaign and the Board that occurred throughout late summer. The decline also was hastened by reports of opposition from Boulder, Adams and other suburban political leaders. In the final weeks before the election, the drop in support continued, and was reported at 46 percent in one survey and 51 percent in another. The history of tax initiatives shows that pre-election day poll reports of declines are common, and when a tax proposal drops below the mid-50 percent range shortly before an election, it will likely lose. A bandwagon effect motivates some supporters to change their minds or not vote, and late-deciding voters usually vote against a tax proposition.
There are a number of methodological and interpretative issues that affect ballot initiative polls conducted in low-turnout tax elections. Divergence between final election polls and the election result is seldom a question of sampling error the confidence range reported in the media as plus or minus some percentage. The difference is most commonly a problem in sample selection, question wording, order or timing of the survey.
One of the most difficult polling tasks is interviewing people most likely to vote. RTD's turnout of 30 percent means 70 percent of registered voters did not turnout and should not be in the survey sample. Screening techniques to determine likely voters have been developed, but are never perfect. In addition, with more than two-thirds of voters in mailback counties voting by mail, it is also necessary to capture a representative sample of those who voted early from home. Voters must be asked whether they have already voted and, if so, how, which can increase refusal votes.
Voters often report pseudo opinions prior to the crystallization of their final preferences. Early in a campaign, people answer ballot questions with a very narrow base of information usually dominated by proponents' early media coverage. Hence, early polls are often affected by more idealistic responses based on little information. Voters toward the end of campaigns begin to hear both sides of issues and tend to focus more on the specifics of the proposals. Being in favor of congestion relief or light rail becomes tempered by information about the proposal's size and tax impact on voters. From the GTR polls in early spring to those late in the campaign, the movement of public opinion reflected the public's absorption of campaign information.
Question order is critical in ballot issue polls. If a tax survey question is preceded by questions that place the tax in a more favorable frame of reference, in this case, questions about growth and congestion, the results will be biased. Some poll questions reported in the media are the results of a single question added to a survey for an unrelated poll. Along with problems inherent in a sample having been selected for a different purpose, the placement of the question can be tainted by preceding questions. In addition, careful polling seldom depends on a single voting question, but rather a series of questions that probe basic attitudes and arguments surrounding the proposal.
A substantial number of voters do not decide how they will vote or even if they are going to vote until quite late in a campaign. While that does not account for the error in one final poll reported prior to the RTD election, it has accounted for polling misses in many elections where consolidating opposition or changing turnout altered the results from the final poll to election day.
Polling is both an art and a science. Even if the pollster's methodology is impeccable, survey results must be interpreted. The same results can lead to different interpretations and provide different political insights into the likely direction of public opinion. When support for a tax initiative is declining and final polls show support at or below 50 percent, the history of similar situations strongly suggests defeat.
Even though Colorado's voter base is growing, its off-year election turnout was as low as expected. The 29 percent turnout of registered voters was within the range of off-year elections in 1993 and 1995. Although there were 150,000 more registered voters in the metro area in 1997 than 1995, only 35,000 more voters turned out. The Guide the Ride campaign hoped low turnout would assist in targeting and getting out supporters. However, when a proposal and its campaign lose credibility, fewer voters are seldom an advantage. The campaign assumed it needed 200,000 "yes" votes, but only received 160,000.
Ballot issue campaigns often believe that low voter turnout is to their advantage. Fewer voters mean resources can be concentrated on fewer political decision makers. The goal is to locate supporters and bring them out to vote. The difficulty with the low turnout strategy is that the most frequent voters, especially with the convenience of mailback balloting, tend to be the most attentive to news and election information, and can be among the most resistant to tax increases and the most receptive to the culture of opposition. Low turnout strategies try to dominate the information that targeted voters receive. But, as this campaign demonstrates, it is nearly impossible to isolate voters from the currents of general media and hard news.
Colorado's introduction of mailback voting and the extension of voting over a month has dramatically changed campaign strategies. In the RTD election, 57 percent of voters reported voting by mail, and most of those did so weeks before November 4, the day for voting in precincts. Not only must advertising start early, but mailback voting means that late election events don't affect voters who have already sent back ballots. Denver's large snowstorm thought to help proponents was not a factor for mailback voters.
There was considerable news and editorial comment about the positive impact of mailback voting on turnout. In fact, the percentage of turnout declined or registered only small increases in this election. Some editorials cited that more than 40 percent of voters turned out in Adams, Arapahoe and Jefferson counties, but the increased percentages were mostly an artifact of county clerks reporting much smaller numbers of total registered voters. In an effort to reduce costs and increase efficiency, clerks only mailed mailback ballots to frequent voters and new registrants. After comparing actual voter differences between 1995 and 1997 using the Secretary of State's registration figures, the change in turnout was very modest.
Mailback voting is, however, convenient for voters and has increased turnout for most counties using it.
Denver is the lowest turnout jurisdiction in the metro area during this election and could benefit from mailback voting. Unfortunately, the Denver election commission has long resisted the change. It argues Denver voters were not prepared for the complexity of the system.
|©2000 Ciruli Associates.
All rights reserved.