Viewpoint: Exit Poll still informative


    By David Magleby

    For the first time in 16 years, the estimates from the 1998 KYBU-Utah Colleges Exit Poll deviated from the actual results by amounts greater than the margin of error would have predicted. This occurred in the three Congressional races.

    Despite press reports to the contrary, David Magleby and J.L. Madrigal did not project a winner in the 2nd Congressional District. The KBYU-Utah Colleges Exit Poll, like the exit polls on KSL and KTVX, saw this race as too close to call.

    Over the eight elections since its inception, the KBYU-Utah Colleges Exit Poll has been consistently within the margin of error in its projections. For reasons that we do not yet understand, the data in the 1998 exit poll was not as accurate at the congressional district level as it has been in the past. Several factors may explain this deviation.

    First, sampling theory informs us that even when the sample and survey are well-designed the sample will be within the margin of error 95 times out of 100, or 19 times out of 20. It is possible that the 1998 sample was our 1 in 20 chance to have a non-representative sample. When doing other types of survey work, it is often possible to return to the field and draw another sample to correct for this statistical artifact. But in exit polling there is only a one-shot opportunity to capture the data.

    Another problem being looked into is an unusually high rejection rate by voters at the voting places. A third possibility could involve the design of the sample. The reality is that when dealing with research in this area, as in so many other areas, you can learn as much from your mistakes as from your successes, and sometimes more. We are now in the process of dissecting our methodology so that in future years we can return to our high level of predictive accuracy.

    To those candidates who won or lost by wider margins than we projected or who were in races that we incorrectly said were too close to call, we apologize if we added anxiety to election night. As always, we did not release any data until after 8:00 p.m., so as not to influence the outcome of the election.

    We thank the hundreds of students from nine Utah colleges and universities who interviewed voters and otherwise participated in the exit poll. We also thank the thousands of Utah voters who completed our questionnaires. There is much more to learn from the data we collected than solely the candidate projections.

    As recent events have demonstrated, elections do matter, and determining what it was that voters were saying can either be left to pundits and journalists or we can be informed by thorough exit polls like the KBYU-Utah Colleges Exit Poll. Some examples of important findings from the 1998 exit poll include:

    More Utah voters saw moral and ethical values and taxes as the most important issues in their vote for the U.S. House than education, social security, or the Clinton scandal.

    Over half of Utah voters wanted President Clinton to resign, and more than 40 percent felt that impeachment would be best for the country if Clinton were not to resign.

    Large majorities of Utah voters think polygamy should be more aggressively prosecuted and that the minimal age for marriage with parental consent should be raised from its present 14 years of age.

    Large majorities of Utah voters favor limits on concealed weapons and gun permits.

    Utah voters were not pleased with the infusion of party and interest group money into Utah elections in 1998, and in nearly all cases over 90 percent of Utah voters favored putting limits on spending in election campaigns.

    More than three-quarters of Utah voters knew what the Y2K (Year 2000) computer issue was, and more than 30 percent believed it will cause major problems.

    These issues are only a subset of the topics that students will be analyzing in the coming weeks an months. The ability to explore how voters felt about the campaign and the important issues in the campaign is the primary reason we do exit polling.

    We will make every effort to correct whatever problems existed in our 1998 survey

    Print Friendly, PDF & Email