We conduct an experiment where subjects are matched in groups of three and vote on a moral transgression. Analyzing different voting rules, the frequency of votes for the moral transgression increases with the number of votes required for it. This effect persists when considering pivotal votes only, which eliminates opportunities to save on own moral costs and to rely instead on sufficiently many votes for the transgression by other group members. A series of novel treatments allows us to identify guilt sharing and preferences for consensual voting as empirically relevant and independent drivers of voting behavior.