We examine how social comparison within endogenously formed peer networks affects different measures of young adult mental health. Using over 2 million smartphone-based communication records from the NetHealth Study, we construct objective social networks among university students and estimate a structural model in which individuals choose their peers and mental health depends on the income of chosen peers. When networks are treated as exogenous, higher average peer income predicts better mental health; once network formation is modelled endogenously, the relationship reverses: exposure to higher-income peers significantly worsens depressive symptoms. Income gaps harm mental health in both directions, consistent with inequality aversion rather than relative deprivation. The reversal survives extending binary links to message-weighted networks, and counterfactual exercises show that policies reducing the volume of messages exchanged with peers improve mental health, increasingly so with the size of the reduction, though the effect of targeted reductions depends on which ties are attenuated. Ignoring endogenous peer selection can thus yield qualitatively misleading conclusions about peer effects on mental health.