Overview. A simple comparison of behavioral data from customers who either did or did not participate in a CRP would suffer from self-selection effects. In experimental terms, participation in a CRP can be interpreted as a treatment, and our research question pertains to whether the treatment causes a certain outcome (recommenders' loyalty). We aim to determine outcome differences (difference in loyalty) between customers in the treatment condition and similar customers without the treatment. However, we cannot know how the treatment group would behave if these participants had not received the treatment (had not participated in a CRP). Unlike in an experimental setting, we cannot randomly assign customers in a real-life setting to participation (i.e., treatment) and nonparticipation (i.e., control) groups. Rather, participants in a CRP have self selected into the participation group, which implies they are at least somewhat satisfied and loyal. Customers who participate in a CRP likely differ substantially from those who do not, beyond their participation, so simply comparing the outcome variables for both groups is not a suitable solution.