ABSTRACT
This chapter proposes that measuring the time it takes survey respondents to answer a question can be a valuable tool in opening the black box of the response formation process. Measuring latency to respond time, it will be argued, offers a deeper understanding of processes typically relegated to the status of functional artifacts of self-report, such as social desirability biasing, to the status of theoretically interesting mental processes. An important caveat to add to this discussion is that the technique described to capture latency data is practical, inexpensive, and accessible for most researchers using Web-based surveys. The question “Did you vote in the last presidential election?” is the focus of the study because a substantial social desirability biasing effect has been repeatedly verified for it. Drawing from the information processing paradigm, the project tests the idea that respondents engaged in social desirability biasing will take longer to respond to questions than those who are not. This is important because the voting question is frequently used as a filter to identify likely voters in upcoming elections. Eliminating respondents who misreport voting is critical to making accurate predictions about the outcome of elections for both proprietary and media-sponsored polls. Further analysis of response latency data taken from four waves of Web-administered cross-sectional surveys administered over a five-year period shows biasing takes place on a much larger set of survey questions than previously recognized. Thus, using response latency suggests social desirability biasing is more than an annoying methodological artifact and should be integrated into a larger theory of the answer generation process in survey research. Further, social desirability may be but one example of how self-report surveys can be augmented by the latency to respond measure.
