Social sanctions sustain social order by reinforcing widely accepted principles. Political polarization may weaken this mechanism by fragmenting these principles, yet causal effects are hard to identify: observational data cannot separate the effect of polarized preferences from exposure to polarization. We model theoretically and test experimentally the effectiveness of social sanctions in a representative U.S. sample (N = 2400) that exogenously varies environmental polarization. Participants allocate money between politically opposed recipients privately and publicly, and public allocations can be punished by partisan Observers drawn from distributions varying in their degree of polarization. With greater polarization, public allocations become less equitable because participants (correctly) expect punishment even when acting fairly. This shows that polarization causally undermines the disciplining role of social sanctions.