Experimental Design
We use a non-matched pair randomized controlled trial (between-subjects design) to test for callback differences across different personality traits signaled by social media profiles in the shared housing market in Germany. To do so, we use a 2x2 design (conscientiousness vs. agreeableness/neuroticism x high vs. low), additional to a baseline condition without social media information. Therefore, we randomly vary the information that a potential roommate or landlord receives as a source of additional information about an applicant’s personality in an application for a vacant room (correspondence test design). The submitted applications are similar in all aspects except for the parts that signal an applicant's personality through the corresponding social media profile.
Experimental Design Details: The informational variation of our study is provided by social media profiles that signal different personality traits. Therefore, we create four social media profiles on the photo-sharing platform “Instagram” showing images of a female student who agreed to provide us with her personal photos for the purpose of this experiment. Using a selection of these photos and additional photos, e.g., from PsychoFlickr (Segalin et al. 2016), we compile a set of photos and description texts indicating high (low) conscientiousness and high (low) agreeableness/low (high) neuroticism. These photos and description texts were uploaded to the profiles over the course of three years. The name of each profile owner is held constant across all conditions and represents one of the most common German female names in the respective birth cohort. In addition, all other characteristics of the profile owner are held constant, i.e., age, study program, etc. The number of followers and the number of other profiles subscribed to by the fictitious profile are very similar for the conditions high (low) conscientiousness and differ for high (low) agreeableness/low (high) neuroticism, in line with the personality traits. Real students are recruited as followers in order to make the profiles as realistic as possible. To test whether the treatment conditions (fictitious social media profiles) successfully manipulate the relevant personality traits, we conducted a randomized online experiment prior to this RCT in which one of the four social media profiles signaling the respective personality traits was randomly assigned.