Homophily in Advice Seeking

How to cite

Gallen, Yana; Wasserman, Melanie. 2021. "Homophily in Advice Seeking." AEA RCT Registry. AEARCTR-0008674. https://doi.org/10.1257/rct.8674-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Yana Gallen* , Melanie Wasserman

* Primary investigator

Registry ID
AEARCTR-0008674
DOI
https://doi.org/10.1257/rct.8674-1.0
First registered
Last updated
Status
On going
Trial dates
2021-11-12 – 2022-02-12
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/8674

Abstract

This paper investigates homophily in advice seeking. Using administrative data from an online college student mentoring platform, we document that female students are 28 percent more likely to reach out to female mentors relative to male students, conditional on various observable characteristics. This preference may be driven by differences in the advice given by female mentors, or it may result from female students using mentor gender as a proxy for other unobservable mentor characteristics. If this preference persists even when female mentors are scarce or have less desirable characteristics than their male counterparts, then it may result in less mentoring time or lower mentorship quality for female students. However, if there are large benefits from homophily in the quality of advice received, then female students may be better off making this tradeoff. Ultimately, it is theoretically ambiguous whether homophily hurts or improves the quality of mentorship received.

Even though female students prefer to interact with female mentors, the patterns from the online mentoring platform provide little evidence that female mentors offer higher quality mentorship. Female mentors are less likely than male mentors to respond to messages from both male and female students. Furthermore, relative to male mentors, female mentors’ responses to female students are shorter. These descriptive patterns call into question whether the benefits of homophily outweigh its costs.

To causally isolate homophily by gender in advice seeking, and shed light on its costs and benefits to students, we adapt a common methodology for estimating willingness-to-pay for product and job attributes. Specifically, we estimate college students’ preferences for mentor characteristics by implementing a hypothetical choice preference elicitation in a setting that incentivizes truthful responses. Students are shown pairs of hypothetical mentors’ profiles and asked to choose which mentor they prefer. We investigate whether female students' preference for female mentors is primarily driven by demand for advice that only female professionals can provide, or if it is because students use mentor gender as a proxy for other desirable characteristics, such as approachability. Specifically, in the preference elicitation, alongside the hypothetical mentor profiles, we provide information from a prior mentee (whose gender is randomized) on their experience with the mentor, including the extent to which the mentor was knowledgable about job opportunities, easy to talk to/friendly, and gave personalized advice.

Intervention

We estimate college students’ preferences for mentor characteristics by implementing a hypothetical choice preference elicitation in a setting that incentivizes truthful responses (namely, receiving personalized help with finding a mentor based on responses). Students are shown pairs of hypothetical mentors’ profiles and asked to choose which mentor they prefer. The characteristics visible to students are: mentor first name, mentor occupation, mentor year of college graduation, whether the mentor is a first generation college goer, mentor availability, and (when shown) the rating of the mentor by a previous mentee (indicated by first name) on 3 dimensions (knowledgable about job opportunities, easy to talk to/friendly, and gave personalized advice). The gender of mentors and mentees is indicated by their first name. We investigate whether female students' preference for female mentors is primarily driven by demand for advice that only female professionals can provide, or if it is because students use mentor gender as a proxy for other desirable characteristics, such as approachability. Specifically, in the preference elicitation, alongside the hypothetical mentor profiles, we provide information from a prior mentee (whose gender is randomized) on their experience with the mentor, including the extent to which the mentor was helpful with finding an internship/job, easy to talk to/friendly, and gave personalized advice. We will randomize 50% of student participants to see pairs of profiles with ratings (ratings treatment) and 50% to see pairs of profiles without ratings (no ratings treatment). All other profile characteristics, and the ratings themselves, are randomized in each profile comparison.

Experimental design

The survey is conducted online through Qualtrics. Upon opening the survey link, students are randomized into the ratings or no ratings condition. The students are then shown 30 pairs of mentor profiles and asked to choose their preferred mentor within each pair. The characteristics of each mentor profile as well as the ratings and gender of the previous mentee are random and independently drawn in each profile. In order to incentivize truthful responses, we tell students at the beginning of the study that their responses will be used to provide them with personalized advice on how to find mentors. After completing the profile comparisons, students are asked background questions about their demographic characteristics, advice-seeking behavior, and stated preferences for mentor characteristics. Within two weeks of completing the survey, students relieve personalized advice and a five dollar deposit into their student account.

Design details

Randomization method
Randomization is done by a computer (javascript in qualtrics at the time the survey is opened)
Randomization unit
The student is the level of randomization
Clusters
500 female students
Observations
15000 student-mentor pairs
Arms
250 students in no ratings treatment, 250 students in ratings treatment. All other profile characteristics randomly and independently assigned and expected to be balanced across treatments.
Primary outcomes
We test whether female students have a preference for female mentors in the treatment with ratings and without ratings. For men, our previous work suggests that there is no same-gender preference in mentoring so we do not anticipate finding such a preference in either treatment in this experiment.
Secondary outcomes
We are also interested in homophily by whether a student/mentor is a first generation college goer and how these preferences are affected by ratings. The tests will be the same as above for gender. In addition to estimating the preference for female mentors, we also estimate students' average willingness to pay for female mentors in terms of other mentor characteristics, in particular mentor availability and occupation match.

Cited by

No citation data available for this registration yet.

Keywords