The effects of discussion and question wording on self and proxy reports of behavioral frequencies

Marketing research surveys often rely on one person to report about the consumption and purchase behavior of other household members. We report the results of an experiment examining how the level of discussion among household members and the wording of a recall question affect the accuracy of reports about the frequency of another household member's behavior. The findings suggest two important implications. First, asking respondents for a count versus an estimate of the behavior affects both the level of reporting for others as well as the accuracy of such reports. Second, screening potential respondents on their level of discussion on a topic with their partner or family member may help reduce reporting error.