The Visual Narrative II: Visual Stereotypes & Social Networks – A Field Experiment on Social Media

2024-05-10

2024-07-15

Acceptance rate

Acceptance rate conditional on variation in social media information signaling ethnicity and visual stereotypes. The acceptance rate measures whether a user accepts the friend/follow request (in case of a private profile) and/or whether a user re-follows the fictitious profile that sent the request (in case of private and/or public profiles).

Profile visits, impressions, and reach of profiles/posts

Profile visits, reach, and impressions of profiles and/or posts are measured on an aggregated level using the data provided on the platform.

To examine the effect of ethnicity and visual stereotypes on the formation of personal social networks, we use four fictitious Instagram profiles. These profiles were created for a randomized controlled trial (AEARCTR-0011322) that conducted a correspondence test in the shared housing market. Half of the fictitious social media profiles contain photos that refer to common Turkish stereotypes in addition to general, non-ethnically specific images. Further, half of the profiles feature a German-sounding male name, while the other half features a Turkish-sounding male name. The present study aims to measure the rate of acceptance of friend/follower requests on social networking sites using these fictitious profiles. The study will use the 'discover friends' suggestions made by the SNS. We collect friend/contact suggestions from the social networking site (180 to 325 per profile in order to achieve a power of .95; see section “Sample size by treatment arms” below) and send a request to follow these profiles. These suggestions are primarily based on existing networks, resulting in a subject pool of students that closely resembles the subject pool of the RCT (AEARCTR-0011322). In cases where a subject is suggested more than once, we randomly select one of the corresponding fictitious profiles to send a friend/follow request to the respective user. Data is collected on the decision of subjects regarding contact requests, including accept or decline (in case of private profiles), or re-follow (in case of private and/or public profiles). Additionally, we collect data on the degree of connection between a person and our profiles, as measured by the number of connections with a suggested profile. Furthermore, we collect publicly available data on target profiles, such as gender, number of contacts, posts, and biographical information.

Not available

n/a; the sample is rather “ad hoc” (convenience sample) since we do not know how the algorithm of the social networking site composes the suggestions. If a profile suggests the same user as a follower or friend as another profile, the user will be randomly selected to be contacted by one of the corresponding fictitious profiles.

Individual level (if applicable, see above)

No

n/a

720-1,300

The study uses two by two = four profiles (two using a German-, two a Turkish-sounding male name, two feature visual Turkish stereotypes, two lack Turkish stereotypical images). A small feasibility test of the experimental design was conducted, albeit without varying ethnicity and visual stereotypes. Therefore, it is difficult to estimate a realistic minimum detectable effect size. Given these constraints, we estimate that we need approximately 180-325 subjects for friend/follow requests per condition to achieve a power of .95 (depending on effect size). Additionally, depending on the contact suggestions provided by the platform, we may require slightly more observations to ensure sufficient variation in the ‘closeness’ of the suggested contacts, thus preventing subjects from becoming aware of the experiment.