Abstract: At the beginning of its appearance, recommender system made use of content-based filtering method where recommendation was based on similarity between keywords in item description and in user profile. The unability to differentiate item quality in this method was overcome by collaborative filtering method by facilitating users to rate items they consumed. However, this rating-based collaborative filtering suffered from cold-start problem. This problem was in turn solved by the use of personality traits due to its advantages. Initially, personality was predicted explicitly by means of questionnaire. Since this technique was regarded burdensome and time consuming, this led to the application of implicit personality elicitation technique. The commonly used techniques is Personality Elicitation from Text (PET) which predicts personality from what someone write in his/her social media account. When applied to a personality-based recommender system, the obligation to have social media account and to write status with certain length are the weaknesses of PET. In order to cope with such drawbacks, personality prediction based on demographic data is proposed.