A Preference for the Similar: Evidence from Peer Mentoring for Smoking Cessation

Although peer effects have been documented in a variety of contexts, little is known about the role played by peer characteristics in mediating those effects. Are people more likely to follow those most similar to them? If so, along what dimensions does similarity matter most? We study this question in the context of peer mentoring for smoking cessation. Peer mentoring delivered by lay coaches forms the basis of popular health promotion programslike Alcoholics Anonymous and Weight Watchers. We offer peer mentoring delivered by former smokers through a text-messaging platform. The platform serves as the basis for a randomized field experiment to test the effectiveness of peer mentoring for smoking cessation. In total, 200 U.S. adult smokers were randomly assigned to a peer mentor or not. Participants in the control group (N = 99) received automated text messages similar to those sent by SmokefreeTXT, a nationwide text-messaging service sponsored by the National Cancer Institute. Those automated messages were sent at a frequency of one to five per day for up to eight weeks. Smokers in the treatment group (N = 101) received a modified version of the same automated messages, but with additional personalized messages from a randomly assigned peer mentor. For each participant, we collect data on participants’ level of engagement with the platform, opt-out decisions, self-reported smoking status at various time intervals and biochemically verified smoking abstinence at 3 months using a saliva cotinine test. We also study smokers’ demand for peer mentors; in particular, we test whether smokers tend to prefer and/or engage more with a mentor more similar to them. To measure similarity, we build several distance measures that combine objective information on mentor/mentee characteristics with rich preference data gathered before assignment to treatment.

Overall, participants in the peer mentor group were more likely than the control group to abstain from smoking 3 months after enrollment. The unadjusted abstinence rate increased from 12.1% to 21.8% for the self-reported measure (p = 0.07) and from 3.0% to 7.9% for the biochemically verified measure (p = 0.13). Controlling for socio-demographics and smoking-related variables, we estimate a treatment effect of 13.9 percentage points (7.9 percentage points, respectively) on self-reported (biochemically verified) abstinence (p < 0.05). Looking at demand for peer mentoring, we find that smokers exhibit a strong preference for being assigned mentors similar to themselves. For characteristics such as age, smoking intensity, reasons for smoking and reasons for quitting, about 50% of participants prefer to be matched with somebody who presents the same characteristics as themselves. While preference for similarity is more strongly expressed for smoking-related characteristics, about 21% of respondents express a preference for being matched with somebody from the same racial group (35% for gender). Together, these preliminary findings suggest the importance of understanding what characteristics of peers may increase engagement and trigger behavior change.