We study the effect of retirement on social capital in urban China. Using three waves of the China Family Panel Studies (2018–2022) and a fuzzy regression discontinuity design based on China’s statutory retirement ages, we document that retirement does not raise or lower social capital uniformly but reallocates it. Retirement causes bridging social capital to fall and bonding social capital to rise. The reallocation is concentrated among men, whose pre-retirement networks are more heavily workplace-centered. As governments around the world postpone statutory retirement ages in response to population ageing, our estimates imply that pension reforms carry social externalities of opposing sign that fiscal accounting omits.