The Accuracy of Self-Reported Internet Use—A Validation Study Using Client Log Data

ABSTRACT

The vast majority of empirical research on online communication, or media use in general, relies on self-report measures instead of behavioral data. Previous research has shown that the accuracy of these self-report measures can be quite low, and both over- and underreporting of media use are commonplace. This study compares self-reports of Internet use with client log files from a large household sample. Results show that the accuracy of self-reported frequency and duration of Internet use is quite low, and that survey data are only moderately correlated with log file data. Moreover, there are systematic patterns of misreporting, especially overreporting, rather than random deviations from the log files. Self-reports for specific content such as social network sites or video platforms seem to be more accurate and less consistently biased than self-reports of generic frequency or duration of Internet use. The article closes by demonstrating the consequences of biased self-reports and discussing possible solutions to the problem.

Acknowledgements

I thank GfK Germany for generously providing survey and log data from their Media Efficiency Panel. The views and conclusions expressed herein are my own and do not necessarily reflect the views of the GfK. I also thank Marko Bachl, the anonymous reviewers and the editor for helpful comments on the manuscript.

Notes

¹ The original items in German were: “Wie oft nutzen Sie persönlich zuhause das Internet für private Zwecke?“ and „Wie viele Stunden pro Woche nutzen Sie persönlich zuhause das Internet für private Zwecke in etwa?“

² Due to the recoding procedure, participants who used the Internet for more than 15 hours per week or 28 days per month could not fall into the overreporting category, whereas respondents who use the Internet extremely infrequently could never underreport their behavior (Prior, 2009b, p. 135).

³ I’d like to thank several anonymous reviewers as well as the editor for suggesting potential threats to the validity and generalizability of the findings.