It is well established that ensemble coding is regulated by physical similarity and variance in a set of stimuli. For example, observers are more accurate at judging the mean size of objects in a set if the overall size variance in the set is small. However, sometimes similarity among set members can be purely subjective. For example, faces from another race tend to look more similar than faces from one’s own race. Very little is known about whether such subjective similarity also regulates ensemble coding in the same manner as objective similarity. To investigate this question, we had British and Chinese participants view sets of four faces that were of either own-race or other-race, own-gender or other-gender. After viewing each set the task was to judge whether a test face was presented in the set. Our results showed that, as demonstrated in prior research, participants often mistook a morphed set average to be a member of the set. Critically, this tendency to average a face set was not stronger for other-race faces. Hence contrary to objective similarity, subjectively perceived similarity in the other-race faces does not facilitate ensemble coding. The results in our British group also replicated de Fockert and Gautrey’s (Psychonomic Bulletin & Review 20 (3), 468-473, 2013) own-gender effect, where observers showed more averaging for own-gender faces. However, our Chinese subjects displayed the same level of averaging for both genders. This suggests a cultural difference in ensemble coding, where the own-gender bias may be overridden by a stronger tendency to employ ensemble coding in Chinese participants.