A comparison of samples and response quality obtained from RDD telephone survey methodology and Internet survey methodology.

摘要

With their response rates declining and costs rising, telephone surveys are increasingly difficult to conduct. At the same time, Internet data collection is emerging as a viable alternative. This dissertation compares a RDD telephone survey with Internet surveys of large-scale national samples conducted before and after the 2000 US. Presidential Election. The same questionnaire was adapted for data collection over the telephone by the Center for Survey Research at Ohio State University, and over the Internet by the two leading commercial firms doing web surveys: Knowledge Networks and Harris Interactive. Knowledge Networks supply computer equipment to national samples recruited through RDD telephone interviews, while Harris Interactive uses volunteer respondents.; Mode comparisons were conducted in terms of sample representativeness, response distributions, survey satisficing, test-retest reliability, predictive validity, and socially desirable responding. In general, it was found that the recruited samples were more representative than the volunteer sample. There was no consistent mode difference in terms of survey satisficing and measurement reliability, but data collected by Internet surveys manifested higher predictive validity than the telephone data. These differences in predictive validity persisted even after controlling for sample differences in demographic attributes and political knowledge.; Nonetheless, the possibility remained that the mode differences were due to other differences between the Internet and telephone samples. In order to ascertain whether a mode difference truly exists, an experiment was conducted in a controlled lab setting, where participants are randomly assigned to answer survey questions on a computer or over an intercom system. Replicating findings from the national surveys, data from the experiment again suggested that predictive validity was higher in the computer mode than the intercom mode. There was some evidence that the mode difference was more pronounced among participants with lower cognitive aptitude, less experience with surveys, less political knowledge, and those who had not received instructions to think carefully: But these moderating effects were never consistently present across all items under investigation, implying that the mode difference was pretty robust despite potential qualifiers.