선거 예측조사의 정확도와 그 영향요인 연구: 17대 총선 예측조사 결과의 사후 분석

Abstract

There has been much progress in the broadcaster's election forecasting. Especially the exit poll, introduced actively from 16th general election, has been considered as an alternative of telephone survey which has proved to have many kinds of problems. But the performance of the exit poll was far behind satisfaction not only in the 17th general election but also in the 16th. The result raised the following question "Is the exit poll superior to the telephone survey in aspect of accuracy?" The purpose of this study is to compare the accuracy of the exit poll with the telephone survey. Also this paper aims at finding out the factors that affect the accuracy of the exit poll. The result shows that the exit poll is almost similar to the telephone survey in forecasting error, but more serious than the telephone survey in forecasting bias. Considering that the main reason of failure of the forecasting in the 17th general election was attributed to forecasting bias which overestimated government party, this result points that current selection criterion of the exit poll in the competing areas has serious problems. And Forecasting error and forecasting bias in exit poll are attributed to non-sampling process rather than sampling process. This result suggests that it is necessary to investigate the sensitive reactions of the interviewee in order to make a substantial progress in the election forecasting survey.