Despite the resultant disutility, some people, in particular, the poor, are engaged in behaviors that carry social stigma. Empirical studies on stigmatized behavior are rare, largely due to the formidable challenges of collecting data on stigmatized goods and services. In this paper, we add to this limited empirical evidence by examining the behavior of "donating" blood plasma in exchange for cash rewards in China. We do so using two primary data sets: the first is a three-wave, census-type household survey that enables us to examine the evolving patterns and determinants of "donating" plasma. The second is data on detailed gift exchange records of all households. The data allow us to define reference groups, measure the intensity of social interactions, and identify peer effects using a novel network structure-based instrumental variable strategy. We find that peer effects influence decisions to "donate" plasma. For example, a one-standard-deviation increase in income from "donating" plasma in the peer group increases the value of own plasma "donation" by 0.15 standard deviations. Families with sons have more incentives to "donate" plasma to offset the escalated costs of getting their sons married in a tight marriage market that favors girls.