The Validity Of Response Latency Measurement In Cati Surveys

There is agreement among researchers that the measurement of raw response times from surveys requires a treatment for baseline speed of respondents and interviewers. There remains disagreement about the appropriate procedure to transform raw response times to response latency and there is no study that evaluates the different approaches with regards to their potential measurement error, which may have an affect on modeling and causal inference. This article fills this gap by examining the measurement validity of different ways raw response times are treated for baseline speed. This study uses data from three separate CATI surveys conducted in Germany in the years 2012 and 2013 where raw response times where measured manually by interviewers. We apply a Multitrait-Multimethod approach to examine four different operationalizations for response latencies and their relationship to attitude accessibility and indicators of response quality. We conclude with remarks about the levels of validity and potential biases when latencies are used as independent variables.