While a large body of literature investigates the bidirectional relationship between retirement and health, few have analyzed the mechanism through which retirement affects health which will provide important policy instrument insights. Using three waves of National Social Life, Health, and Aging Project, we examine the mediating role of the social network in the relationship between retirement and health in USA. We address the endogeneity and reverse causality through panel instrumental fixed-effect methods. We apply both single and parallel mediation analyses to identify the potential mechanism by which social network characteristics mediate the impact of retirement on health. Findings reveal that retirement adversely affects physical and mental health outcomes, and a considerable portion of these effects are explained by social network changes post-retirement. Specifically, 58% of reduction in the probability of reporting good physical health and 4.5% of increment in chances of having depression symptoms post-retirement can be explained by shrinkage in the size of social network in retirees. Using parallel mediation identification to account for dependencies among social network features, we find that social network size induces 79.5% reduction in probability of reporting good physical health and 18.6% increase in probability of having depression in retirees as compared to non-retirees. Findings in this paper suggest that investing in social network of the elderly can buffer the adverse health effect of retirement and can be an effective policy target for promoting healthy aging.