People often make judgments about the ethicality of others' behaviors and then decide how harshly to punish such behaviors. When they make these judgments and decisions, sometimes the victims of the unethical behavior are identifiable, and sometimes they are not. In addition, in our uncertain world, sometimes an unethical action causes harm, and sometimes it does not. We argue that a rational assessment of ethicality should not depend on the identifiability of the victim of wrongdoing or the actual harm caused. Yet in four laboratory studies, we show that these factors have a systematic effect on how people judge the ethicality of the perpetrator of an unethical action. Specifically, we find that identifiability of the victim of wrongdoing and information about the outcome of wrongdoing influence both ethical judgments and decisions to punish wrongdoers. Our studies show that people judge behavior as more unethical when (1) identifiable versus statistical victims are involved and (2) the behavior leads to a negative rather than a positive outcome. We also find that people's willingness to punish wrongdoers is consistent with their judgments, and we offer preliminary evidence on how to reduce these biases.