| In the past decade, access to the Internet has substantially increased, and now various commercial software packages provide the means for researchers to readily design and create online surveys in a short period of time. The purpose of this dissertation is to demonstrate the utility and cost-effectiveness of conducting methodological experiments on survey question effects such as effects of not sure, don't know, and no opinion, tone of wording: forbid vs. allow, effects of middle response alternative, and response order effects.;Since political scientists rarely collect their own data, due to related cost issues, investigators are unable to easily test hypotheses due to their dependency on secondary data or small voluntary student samples. In this research, several classic experiments on question form, wording, and context were replicated using the Internet as a public opinion laboratory.;This study was conducted in two phases prior to 2000 presidential election using Internet technology at the University of Cincinnati. Respondents were randomly assigned to one of the two question forms, and the topics of survey questions in this study were focused on issues in the 2000 presidential election campaign adapted for self-administration on the Internet.;Six experiments were conducted examining the effects of no opinion, don't know, and not sure response alternatives. Except in one case, in which a large number of respondents with a higher level of education expressed opinions about a fictitious issue, the results replicated earlier findings by Schuman and Presser (1981). Two experiments examined the selection of "forbid" vs. "not allow" showing the preference of respondents to choose "not allow" over "forbid." However, we found a lower margin of difference in our experiments in comparison with those of Schuman and Presser (1981). Three experiments examining the effects of a middle response alternative clearly suggested that offering a "middle alternative" results in a significant increase in its selection. Examining the univariate distribution, a statistically significant difference based on response form was found in only one of three cases. The results suggest that middle response alternative effects may not be generalizable to all issues. The bivariate results were similar to findings of Schuman and Presser (1981) and Kalton, Roberts, and Holts (1980) showing that the inferences about the relationship between background variables and attitude items are not affected by the presence or absence of a middle alternative. Finally, three experiments were conducted to examine the response order effects. Only one of the experiments showed a strong primacy effect, with the other two producing less robust results.;Overall, this research confirms that this new mode of data collection utilizing the Internet technology can be effectively utilized to create and/or replicate various methodological experiments in an efficient and cost-effective manner as well as stimulate the development of new hypotheses on survey question effects. |
