Chen Quin Lam, Department of Statistics, The Ohio State University
Handling Undecided Voters: Using Missing Data Methods in Election Forecasting
The primary aim of this study is to compare the methodologies used in the political science literature and sample survey literature in forecasting election outcomes in the presence of incomplete data. Following a study using data from the Buckeye State Poll conducted in Ohio, where equal allocation of undecided respondents seems to improve forecasting, this article examines alternative missing-data methods used in sample survey literature to deal with the allocation of the undecided respondents. After fitting these models, we compare their effectiveness in forecasting election outcomes. Maximum likelihood estimation is used to fit most of these models and it was found that some of them do reasonably well in forecasting election outcomes.
Mary Outwater, Department of Political Science, The Ohio State University
Are Higher Response Rates Better? An Investigation Into the Usefulness of Convincing Reluctant Respondents to Participate in Telephone Surveys
One of the purposes of survey research is to present a reasonable estimation of public opinion. It is commonly assumed that the best way to accomplish this is to gain the highest response rate possible. The question raised here is whether this philosophy holds in matters non-political. Brehm (1993) investigated this question with regard to political issues and found little attitude difference among those who readily agreed to participate in a survey and those who initially refused but were then later persuaded, but little is known regarding such differences in other areas.
This research expands this area of study by looking at the differences in attitudes between respondents who agree to participate in a telephone survey on the first attempt, and those who initially refuse but agree during a second attempt at a later date. The focus here is on attitudes dealing with economic matters, such as how people think the economy is doing and by how much inflation will affect prices in the near future. These types of questions form the basis of many economic indicators and are often cited in all forms of news media. The ultimate goal of this project is to help determine how much effort should be placed into converting initial refusals into ultimate completed interviews. Such attempts are generally time-consuming for the survey organizations that engage in them and therefore evaluating the usefulness of the practice is of great interest.
