Brenner, Philip S., and John D. DeLamater
Working paper no. 2013-03
Like that of other socially desirable behaviors, much of the research on physical exercise is based on self-reports which are prone to overreporting. While research has focused on identifying the presence and degree of overreporting, this paper fills an important gap by investigating its causes. The explanation based in impression management will be challenged, using an explanation based in identity theory as an arguably better fitting alternative. Respondents were randomly assigned to one of two conditions: (1) a web instrument using direct survey questions, or (2) a chronological reporting procedure using text messaging. Comparisons to validation data from a reverse record check indicate significantly greater rates of overreporting in the web condition than in the text condition. Results suggest that measurement bias is associated with the importance of the respondents’ exercise identity, prompted by the directness of the conventional survey question. Findings call into question the benefit of self-administration for bias reduction in measurement of normative behaviors.