Addressing the challenges of measuring self-reported media use: Using contingent feedback to increase data quality in web and face-to-face survey modes

Although passive log-file data can be useful to determine household digital media consumption, many researchers rely on self-reported media measures because samples are less expensive to recruit and maintain, and self-reported measures can estimate a broad range of media use as well as other attitudinal and behavioral information. Unfortunately, digital media use is often difficult for respondents to report. First, some digital media consumption may be considered sensitive, leading respondents to answer in a socially desirable manner. Second, the pervasiveness of digital media makes it difficult for respondents to recall their behavior, leading to estimations and guesses. Third, as the names of digital media services evolve, respondents may be asked to respond to effortful open questions that inquire into uses for media rather than be asked to respond to less onerous closed questions about use of specific services. This dissertation examines whether or not the mode of survey data collection can address these issues, thereby increasing the validity of self-reported digital media use estimates. Specifically, I assert that the ability of web surveys to provide contingent feedback, similar to face-to-face interviews, can decrease social desirability bias and satisficing and increase the quality of responses to open media use questions. In order to test my hypotheses, I conducted a survey mode experiment with three conditions: face-to-face, standard web and experimental web. The face-to-face and experimental web condition use contingent feedback in the form of probes and instructions in an attempt to decrease social desirability bias and satisficing while increasing the number of responses to open questions.;The findings indicate that probes and instructions in the experimental web mode decrease social desirability bias and satisficing in closed questions but do not increase the number of answers to open questions. Furthermore, the contingent feedback appears to increase data quality to a greater extent in the face-to-face condition than in the experimental web condition. The findings lend support to the notion that the quality of self-reported digital media use measures can be improved by the use of probes, instructions and other types of contingent feedback in both interviewer and self-administered modes.