This study aims to fill two interrelated knowledge gaps in the extant literature on the association between perceived discrimination and health. First, potential selection bias associated with pre-existing health conditions has rarely been rigorously tested in empirical studies. Second, whether there is a reciprocal relationship between perceived discrimination and health has been underexplored. Using longitudinal data from the Americans' Changing Lives data, waves 3 to 5 (1,058), we test the reciprocity between perceived discrimination and health with a formal mediation analysis technique. We also use the Heckman correction to adjust for the potential selection bias associated with attrition. Our analysis indicates that perceived discrimination is associated with poor self-rated health and depressive symptoms even when previous health conditions are considered. Furthermore, net of other confounders, there is a reciprocal relationship between perceived discrimination and depressive symptoms. However, this reciprocity does not hold for self-rated health. These findings indicate that there is a vicious circle between perceived discrimination and mental health. That is, poor mental health may lead to perceived discrimination, and heightened perceived discrimination may subsequently increase depressive symptoms. Sensitivity tests suggest that this reciprocity may vary by gender and race.