In the past decades, there has been a resurgence of public bike-sharing systems (BSSs). While it is claimed that social and environmental benefits are associated with the implementation of BSSs, few empirical studies have investigated the actual congestion reduction effect of BSSs on cities. To fill such gap, this paper aims to examine whether the launch of BSSs can reduce citywide congestion. With a panel dataset of 96 urban areas in the US from 2005 to 2014, we employed a difference-in-differences model with two-way fixed-effects panel regression. The results suggested that the introduction of BSSs shows a significant mixed impact on congestion in general: Larger cities get better off but richer cities get worse off. Such results are consistent with both subsample regression with propensity score matching and different congestion measures. Post-hoc analysis reveals that BSSs have a significant positive effect on reducing rush-hour congestion. Finally, implications, limitations, and future work directions are offered.