Publication Type: Paper/Poster Proposal
Abstract: With the growth of serious response problems in cold random-digit-dialed national samples, numerous analysts have turned to online samples to improve response rates and facilitate timely data collection. This paper reports on the use of Knowledge Networks’ national adult sample to continue some of the time series items previously collected from telephone RDD samples.

Attention is focused on responses asking a respondent to rate his or her level of interest in selected public policy issues and his or her level of understanding of the same issues. These items have been reported extensively in the NSF’s Science and Engineering Indicators series and have been used to estimate the proportion of adults attentive to science and technology and other issue clusters. Recent replications of this work using Knowledge Networks’ adult panel indicate that respondents tend to rate their interest in and knowledge of the same public policy issues significantly lower in online data collections than in earlier telephone interviews.

It is hypothesized that there is a mode effect. The almost 20 years of telephone data are remarkably stable, with almost all variations following major events such as the Persian Gulf War in 1991. The across-the-board reduction in self-assessed knowledge levels in 2003 and 2004 suggest that the absence of an interviewer may have removed a source of implicit social pressure to inflate the level of interest and that the presentation of a “menu” of policy issues may have provided a different context than the sequential presentation of issues in telephone interviews.

A series of descriptive and analytic techniques (including structural equation models) are used to assess the mode hypothesis.