How does referral recruitment contribute to job segregation, and what can organizations do about it?Current theory on network effects in the labor market emphasizes the job-seeker perspective, focusing on the segregated nature of job-seekers' information and contact networks, and leaves little role for organizational influence. But employee referrals are necessarily initiated from within a firm by referrers. We argue that referrer behavior is the missing link that can help organizations manage the segregating effects of referring.Adopting the referrer's perspective of the process, we develop a computational model which integrates a set of empirically documented referrer behavior mechanisms gleaned from extant organizational case studies.Using this model, we compare the segregating effects of referring when these behaviors are inactive to the effects when the behaviors are active. We show that referrer behaviors substantially boost the segregating effects of referring. This impact of referrer behavior presents an opportunity for organizations. Contrary to popular wisdom, we show that organizational policies designed to influence referrer behaviors can mitigate most if not all of the segregating effects of referring. , and is associated with the imposition of mandated policy changes on organizational defendants in EEO lawsuits (regarding sex discrimination cases, see Hirsh [2008]). In light of these perceived organizational and societal perils, the numerous calls to reduce or even eliminate recruitment via employee 1 Although referral recruitment is viewed to have associations with both the race and sex segregation of jobs, in this paper we focus on sex segregation. The simplifying choice to focus on sex segregation was a pragmatic one: modeling the dynamics of a dichotomous variable such as sex is more straightforward than a multi-valued (and potentially multidimensional) variable as race. We recognize this choice as a limitation. Sex and race are both important and omnipresent social markers of individuals in organizational contexts (Ashforth and Humphrey 1995;Heilman 1995; Hirschfield 1999), and these particular markers can interact in complex ways (Chafetz 1997), especially in the context of the labor market (Browne and Misra 2003;Robinson et al. 2005). This simplification neither makes any assumptions of the relative importance of one type of segregation over another, nor reflects a view that the homogenous treatment of groups by race or sex, and not both simultaneously, is unproblematic. We plan to expand future versions of the model to include race and sex simultaneously.